MFV: zstd 1.5.7.

Approved by:	re (cperciva)
MFC after:	2 weeks
Relnotes:	yes

(cherry picked from commit c0d9a07101a1e72769ee0619a583f63a078fb391)
(cherry picked from commit 7c9418613a31571e4eceda55a417f7183d26acb5)
(cherry picked from commit 111a42d7d78a88ba6ff7b5d61b0867faccc00276)
(cherry picked from commit ff426ceef657d7c478471e631349c5d1a8b96453)

34 files changed, 4936 insertions, 2958 deletions

diff --git a/sys/contrib/zstd/programs/BUCK b/sys/contrib/zstd/programs/BUCK
deleted file mode 100644
index d2aa6373ff47..000000000000
--- a/sys/contrib/zstd/programs/BUCK
+++ /dev/null
@@ -1,44 +0,0 @@
-cxx_binary(
-    name='zstd',
-    headers=glob(['*.h'], excludes=['datagen.h', 'platform.h', 'util.h']),
-    srcs=glob(['*.c'], excludes=['datagen.c']),
-    deps=[
-        ':datagen',
-        ':util',
-        '//lib:zstd',
-        '//lib:zdict',
-        '//lib:mem',
-        '//lib:xxhash',
-    ],
-    preprocessor_flags=[
-        '-DZSTD_GZCOMPRESS',
-        '-DZSTD_GZDECOMPRESS',
-        '-DZSTD_LZMACOMPRESS',
-        '-DZSTD_LZMADECOMPRES',
-        '-DZSTD_LZ4COMPRESS',
-        '-DZSTD_LZ4DECOMPRES',
-    ],
-    linker_flags=[
-        '-lz',
-        '-llzma',
-        '-llz4',
-    ],
-)
-
-cxx_library(
-    name='datagen',
-    visibility=['PUBLIC'],
-    header_namespace='',
-    exported_headers=['datagen.h'],
-    srcs=['datagen.c'],
-    deps=['//lib:mem'],
-)
-
-
-cxx_library(
-    name='util',
-    visibility=['PUBLIC'],
-    header_namespace='',
-    exported_headers=['util.h', 'platform.h'],
-    deps=['//lib:mem'],
-)
diff --git a/sys/contrib/zstd/programs/Makefile b/sys/contrib/zstd/programs/Makefile
deleted file mode 100644
index f77e1b7f10f8..000000000000
--- a/sys/contrib/zstd/programs/Makefile
+++ /dev/null
@@ -1,435 +0,0 @@
-# ################################################################
-# Copyright (c) Yann Collet, Facebook, Inc.
-# All rights reserved.
-#
-# This source code is licensed under both the BSD-style license (found in the
-# LICENSE file in the root directory of this source tree) and the GPLv2 (found
-# in the COPYING file in the root directory of this source tree).
-# You may select, at your option, one of the above-listed licenses.
-# ##########################################################################
-# zstd : Command Line Utility, supporting gzip-like arguments
-# zstd32 : Same as zstd, but forced to compile in 32-bits mode
-# zstd-nolegacy : zstd without support of decompression of legacy versions
-# zstd-small : minimal zstd without dictionary builder and benchmark
-# zstd-compress : compressor-only version of zstd
-# zstd-decompress : decompressor-only version of zstd
-# ##########################################################################
-
-.PHONY: default
-default: zstd-release
-
-LIBZSTD := ../lib
-
-include $(LIBZSTD)/libzstd.mk
-
-ifeq ($(shell $(CC) -v 2>&1 | $(GREP) -c "gcc version "), 1)
-  ALIGN_LOOP = -falign-loops=32
-else
-  ALIGN_LOOP =
-endif
-
-ZSTDLIB_COMMON_SRC := $(sort $(ZSTD_COMMON_FILES))
-ZSTDLIB_COMPRESS_SRC := $(sort $(ZSTD_COMPRESS_FILES))
-ZSTDLIB_DECOMPRESS_SRC := $(sort $(ZSTD_DECOMPRESS_FILES))
-ZSTDLIB_CORE_SRC := $(sort $(ZSTD_DECOMPRESS_FILES) $(ZSTD_COMMON_FILES) $(ZSTD_COMPRESS_FILES))
-ZDICT_SRC := $(sort $(ZSTD_DICTBUILDER_FILES))
-ZSTDLEGACY_SRC := $(sort $(ZSTD_LEGACY_FILES))
-
-# Sort files in alphabetical order for reproducible builds
-ZSTDLIB_FULL_SRC = $(sort $(ZSTDLIB_CORE_SRC) $(ZSTDLEGACY_SRC) $(ZDICT_SRC))
-ZSTDLIB_LOCAL_SRC = $(notdir $(ZSTDLIB_FULL_SRC))
-ZSTDLIB_LOCAL_OBJ0 := $(ZSTDLIB_LOCAL_SRC:.c=.o)
-ZSTDLIB_LOCAL_OBJ := $(ZSTDLIB_LOCAL_OBJ0:.S=.o)
-
-ZSTD_CLI_SRC := $(sort $(wildcard *.c))
-ZSTD_CLI_OBJ := $(ZSTD_CLI_SRC:.c=.o)
-
-ZSTD_ALL_SRC = $(ZSTDLIB_LOCAL_SRC) $(ZSTD_CLI_SRC)
-ZSTD_ALL_OBJ0 := $(ZSTD_ALL_SRC:.c=.o)
-ZSTD_ALL_OBJ := $(ZSTD_ALL_OBJ0:.S=.o)
-
-# Define *.exe as extension for Windows systems
-ifneq (,$(filter Windows%,$(OS)))
-  EXT =.exe
-  RES64_FILE = windres/zstd64.res
-  RES32_FILE = windres/zstd32.res
-ifneq (,$(filter x86_64%,$(shell $(CC) -dumpmachine)))
-    RES_FILE = $(RES64_FILE)
-else
-    RES_FILE = $(RES32_FILE)
-endif
-else
-  EXT =
-endif
-
-# thread detection
-NO_THREAD_MSG := ==> no threads, building without multithreading support
-HAVE_PTHREAD := $(shell printf '$(NUM_SYMBOL)include <pthread.h>\nint main(void) { return 0; }' > have_pthread.c && $(CC) $(FLAGS) -o have_pthread$(EXT) have_pthread.c -pthread 2> $(VOID) && rm have_pthread$(EXT) && echo 1 || echo 0; rm have_pthread.c)
-HAVE_THREAD := $(shell [ "$(HAVE_PTHREAD)" -eq "1" -o -n "$(filter Windows%,$(OS))" ] && echo 1 || echo 0)
-ifeq ($(HAVE_THREAD), 1)
-  THREAD_MSG := ==> building with threading support
-  THREAD_CPP := -DZSTD_MULTITHREAD
-  THREAD_LD := -pthread
-else
-  THREAD_MSG := $(NO_THREAD_MSG)
-endif
-
-# zlib detection
-NO_ZLIB_MSG := ==> no zlib, building zstd without .gz support
-HAVE_ZLIB ?= $(shell printf '$(NUM_SYMBOL)include <zlib.h>\nint main(void) { return 0; }' > have_zlib.c && $(CC) $(FLAGS) -o have_zlib$(EXT) have_zlib.c -lz 2> $(VOID) && rm have_zlib$(EXT) && echo 1 || echo 0; rm have_zlib.c)
-ifeq ($(HAVE_ZLIB), 1)
-  ZLIB_MSG := ==> building zstd with .gz compression support
-  ZLIBCPP = -DZSTD_GZCOMPRESS -DZSTD_GZDECOMPRESS
-  ZLIBLD = -lz
-else
-  ZLIB_MSG := $(NO_ZLIB_MSG)
-endif
-
-# lzma detection
-NO_LZMA_MSG := ==> no liblzma, building zstd without .xz/.lzma support
-HAVE_LZMA ?= $(shell printf '$(NUM_SYMBOL)include <lzma.h>\nint main(void) { return 0; }' > have_lzma.c && $(CC) $(FLAGS) -o have_lzma$(EXT) have_lzma.c -llzma 2> $(VOID) && rm have_lzma$(EXT) && echo 1 || echo 0; rm have_lzma.c)
-ifeq ($(HAVE_LZMA), 1)
-  LZMA_MSG := ==> building zstd with .xz/.lzma compression support
-  LZMACPP = -DZSTD_LZMACOMPRESS -DZSTD_LZMADECOMPRESS
-  LZMALD = -llzma
-else
-  LZMA_MSG := $(NO_LZMA_MSG)
-endif
-
-# lz4 detection
-NO_LZ4_MSG := ==> no liblz4, building zstd without .lz4 support
-HAVE_LZ4 ?= $(shell printf '$(NUM_SYMBOL)include <lz4frame.h>\n$(NUM_SYMBOL)include <lz4.h>\nint main(void) { return 0; }' > have_lz4.c && $(CC) $(FLAGS) -o have_lz4$(EXT) have_lz4.c -llz4 2> $(VOID) && rm have_lz4$(EXT) && echo 1 || echo 0; rm have_lz4.c)
-ifeq ($(HAVE_LZ4), 1)
-  LZ4_MSG := ==> building zstd with .lz4 compression support
-  LZ4CPP = -DZSTD_LZ4COMPRESS -DZSTD_LZ4DECOMPRESS
-  LZ4LD = -llz4
-else
-  LZ4_MSG := $(NO_LZ4_MSG)
-endif
-
-# explicit backtrace enable/disable for Linux & Darwin
-ifeq ($(BACKTRACE), 0)
-  DEBUGFLAGS += -DBACKTRACE_ENABLE=0
-endif
-ifeq (,$(filter Windows%, $(OS)))
-ifeq ($(BACKTRACE), 1)
-  DEBUGFLAGS += -DBACKTRACE_ENABLE=1
-  DEBUGFLAGS_LD += -rdynamic
-endif
-endif
-
-SET_CACHE_DIRECTORY = \
-   +$(MAKE) --no-print-directory $@ \
-    BUILD_DIR=obj/$(HASH_DIR) \
-    CPPFLAGS="$(CPPFLAGS)" \
-    CFLAGS="$(CFLAGS)" \
-    LDFLAGS="$(LDFLAGS)" \
-    LDLIBS="$(LDLIBS)" \
-    ZSTD_ALL_SRC="$(ZSTD_ALL_SRC)"
-
-
-.PHONY: all
-all: zstd
-
-.PHONY: allVariants
-allVariants: zstd zstd-compress zstd-decompress zstd-small zstd-frugal zstd-nolegacy zstd-dictBuilder
-
-.PHONY: zstd  # must always be run
-zstd : CPPFLAGS += $(THREAD_CPP) $(ZLIBCPP) $(LZMACPP) $(LZ4CPP)
-zstd : LDFLAGS += $(THREAD_LD) $(DEBUGFLAGS_LD)
-zstd : LDLIBS += $(ZLIBLD) $(LZMALD) $(LZ4LD)
-zstd : CPPFLAGS += -DZSTD_LEGACY_SUPPORT=$(ZSTD_LEGACY_SUPPORT)
-ifneq (,$(filter Windows%,$(OS)))
-zstd : $(RES_FILE)
-endif
-
-ifndef BUILD_DIR
-# generate BUILD_DIR from flags
-
-zstd:
-	$(SET_CACHE_DIRECTORY)
-
-else
-# BUILD_DIR is defined
-
-ZSTD_OBJ := $(addprefix $(BUILD_DIR)/, $(ZSTD_ALL_OBJ))
-$(BUILD_DIR)/zstd : $(ZSTD_OBJ)
-	@echo "$(THREAD_MSG)"
-	@echo "$(ZLIB_MSG)"
-	@echo "$(LZMA_MSG)"
-	@echo "$(LZ4_MSG)"
-	@echo LINK $@
-	$(CC) $(FLAGS) $^ $(LDLIBS) -o $@$(EXT)
-
-ifeq ($(HAVE_HASH),1)
-SRCBIN_HASH = $(shell cat $(BUILD_DIR)/zstd$(EXT) 2> $(VOID) | $(HASH) | cut -f 1 -d " ")
-DSTBIN_HASH = $(shell cat zstd$(EXT) 2> $(VOID) | $(HASH) | cut -f 1 -d " ")
-BIN_ISDIFFERENT = $(if $(filter $(SRCBIN_HASH),$(DSTBIN_HASH)),0,1)
-else
-BIN_ISDIFFERENT = 1
-endif
-
-zstd : $(BUILD_DIR)/zstd
-	if [ $(BIN_ISDIFFERENT) -eq 1 ]; then \
-		cp -f $<$(EXT) $@$(EXT); \
-		echo zstd build completed; \
-	else \
-		echo zstd already built; \
-	fi
-
-endif  # BUILD_DIR
-
-
-.PHONY: zstd-release
-zstd-release: DEBUGFLAGS := -DBACKTRACE_ENABLE=0
-zstd-release: DEBUGFLAGS_LD :=
-zstd-release: zstd
-
-zstd32 : CPPFLAGS += $(THREAD_CPP)
-zstd32 : LDFLAGS  += $(THREAD_LD)
-zstd32 : CPPFLAGS += -DZSTD_LEGACY_SUPPORT=$(ZSTD_LEGACY_SUPPORT)
-ifneq (,$(filter Windows%,$(OS)))
-zstd32 : $(RES32_FILE)
-endif
-zstd32 : $(ZSTDLIB_FULL_SRC) $(ZSTD_CLI_SRC)
-	$(CC) -m32 $(FLAGS) $^ -o $@$(EXT)
-
-## zstd-nolegacy: same scope as zstd, with just support of legacy formats removed
-zstd-nolegacy : LDFLAGS += $(THREAD_LD) $(ZLIBLD) $(LZMALD) $(LZ4LD) $(DEBUGFLAGS_LD)
-zstd-nolegacy : CPPFLAGS += -UZSTD_LEGACY_SUPPORT -DZSTD_LEGACY_SUPPORT=0
-zstd-nolegacy : $(ZSTDLIB_CORE_SRC) $(ZDICT_SRC) $(ZSTD_CLI_OBJ)
-	$(CC) $(FLAGS) $^ -o $@$(EXT) $(LDFLAGS)
-
-.PHONY: zstd-nomt
-zstd-nomt : THREAD_CPP :=
-zstd-nomt : THREAD_LD  :=
-zstd-nomt : THREAD_MSG := - multi-threading disabled
-zstd-nomt : zstd
-
-.PHONY: zstd-nogz
-zstd-nogz : ZLIBCPP :=
-zstd-nogz : ZLIBLD  :=
-zstd-nogz : ZLIB_MSG := - gzip support is disabled
-zstd-nogz : zstd
-
-.PHONY: zstd-noxz
-zstd-noxz : LZMACPP :=
-zstd-noxz : LZMALD  :=
-zstd-noxz : LZMA_MSG := - xz/lzma support is disabled
-zstd-noxz : zstd
-
-## zstd-dll: zstd executable linked to dynamic library libzstd (must have same version)
-.PHONY: zstd-dll
-zstd-dll : LDFLAGS+= -L$(LIBZSTD)
-zstd-dll : LDLIBS += -lzstd
-zstd-dll : ZSTDLIB_LOCAL_SRC = xxhash.c
-zstd-dll : zstd
-
-
-## zstd-pgo: zstd executable optimized with PGO.
-.PHONY: zstd-pgo
-zstd-pgo :
-	$(MAKE) clean
-	$(MAKE) zstd MOREFLAGS=-fprofile-generate
-	./zstd -b19i1 $(PROFILE_WITH)
-	./zstd -b16i1 $(PROFILE_WITH)
-	./zstd -b9i2 $(PROFILE_WITH)
-	./zstd -b $(PROFILE_WITH)
-	./zstd -b7i2 $(PROFILE_WITH)
-	./zstd -b5 $(PROFILE_WITH)
-	$(RM) zstd *.o
-	case $(CC) in *clang*) if ! [ -e default.profdata ]; then llvm-profdata merge -output=default.profdata default*.profraw; fi ;; esac
-	$(MAKE) zstd MOREFLAGS=-fprofile-use
-
-## zstd-small: minimal target, supporting only zstd compression and decompression. no bench. no legacy. no other format.
-zstd-small: CFLAGS = -Os -s
-zstd-frugal zstd-small: $(ZSTDLIB_CORE_SRC) zstdcli.c util.c timefn.c fileio.c
-	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT -DZSTD_NOTRACE -UZSTD_LEGACY_SUPPORT -DZSTD_LEGACY_SUPPORT=0 $^ -o $@$(EXT)
-
-zstd-decompress: $(ZSTDLIB_COMMON_SRC) $(ZSTDLIB_DECOMPRESS_SRC) zstdcli.c util.c timefn.c fileio.c
-	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT -DZSTD_NOCOMPRESS -DZSTD_NOTRACE -UZSTD_LEGACY_SUPPORT -DZSTD_LEGACY_SUPPORT=0 $^ -o $@$(EXT)
-
-zstd-compress: $(ZSTDLIB_COMMON_SRC) $(ZSTDLIB_COMPRESS_SRC) zstdcli.c util.c timefn.c fileio.c
-	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODICT -DZSTD_NODECOMPRESS -DZSTD_NOTRACE -UZSTD_LEGACY_SUPPORT -DZSTD_LEGACY_SUPPORT=0 $^ -o $@$(EXT)
-
-## zstd-dictBuilder: executable supporting dictionary creation and compression (only)
-zstd-dictBuilder: $(ZSTDLIB_COMMON_SRC) $(ZSTDLIB_COMPRESS_SRC) $(ZDICT_SRC) zstdcli.c util.c timefn.c fileio.c dibio.c
-	$(CC) $(FLAGS) -DZSTD_NOBENCH -DZSTD_NODECOMPRESS -DZSTD_NOTRACE $^ -o $@$(EXT)
-
-zstdmt: zstd
-	ln -sf zstd zstdmt
-
-.PHONY: generate_res
-generate_res: $(RES64_FILE) $(RES32_FILE)
-
-ifneq (,$(filter Windows%,$(OS)))
-RC ?= windres
-# http://stackoverflow.com/questions/708238/how-do-i-add-an-icon-to-a-mingw-gcc-compiled-executable
-$(RES64_FILE): windres/zstd.rc
-	$(RC) -o $@ -I ../lib -I windres -i $< -O coff -F pe-x86-64
-$(RES32_FILE): windres/zstd.rc
-	$(RC) -o $@ -I ../lib -I windres -i $< -O coff -F pe-i386
-endif
-
-.PHONY: clean
-clean:
-	$(RM) core *.o tmp* result* *.gcda dictionary *.zst \
-        zstd$(EXT) zstd32$(EXT) zstd-dll$(EXT) \
-        zstd-compress$(EXT) zstd-decompress$(EXT) \
-        zstd-small$(EXT) zstd-frugal$(EXT) zstd-nolegacy$(EXT) zstd4$(EXT) \
-        zstd-dictBuilder$(EXT) \
-        *.gcda default*.profraw default.profdata have_zlib$(EXT)
-	$(RM) -r obj/*
-	@echo Cleaning completed
-
-MD2ROFF = ronn
-MD2ROFF_FLAGS = --roff --warnings --manual="User Commands" --organization="zstd $(ZSTD_VERSION)"
-
-zstd.1: zstd.1.md ../lib/zstd.h
-	cat $< | $(MD2ROFF) $(MD2ROFF_FLAGS) | sed -n '/^\.\\\".*/!p' > $@
-
-zstdgrep.1: zstdgrep.1.md ../lib/zstd.h
-	cat $< | $(MD2ROFF) $(MD2ROFF_FLAGS) | sed -n '/^\.\\\".*/!p' > $@
-
-zstdless.1: zstdless.1.md ../lib/zstd.h
-	cat $< | $(MD2ROFF) $(MD2ROFF_FLAGS) | sed -n '/^\.\\\".*/!p' > $@
-
-.PHONY: man
-man: zstd.1 zstdgrep.1 zstdless.1
-
-.PHONY: clean-man
-clean-man:
-	$(RM) zstd.1
-	$(RM) zstdgrep.1
-	$(RM) zstdless.1
-
-.PHONY: preview-man
-preview-man: clean-man man
-	man ./zstd.1
-	man ./zstdgrep.1
-	man ./zstdless.1
-
-
-# Generate .h dependencies automatically
-
-DEPFLAGS = -MT $@ -MMD -MP -MF
-
-$(BUILD_DIR)/%.o : %.c $(BUILD_DIR)/%.d | $(BUILD_DIR)
-	@echo CC $@
-	$(COMPILE.c) $(DEPFLAGS) $(BUILD_DIR)/$*.d $(OUTPUT_OPTION) $<
-
-$(BUILD_DIR)/%.o : %.S | $(BUILD_DIR)
-	@echo AS $@
-	$(COMPILE.S) $(OUTPUT_OPTION) $<
-
-MKDIR ?= mkdir
-$(BUILD_DIR): ; $(MKDIR) -p $@
-
-DEPFILES := $(ZSTD_OBJ:.o=.d)
-$(DEPFILES):
-
-include $(wildcard $(DEPFILES))
-
-
-
-#-----------------------------------------------------------------------------
-# make install is validated only for Linux, macOS, BSD, Hurd and Solaris targets
-#-----------------------------------------------------------------------------
-ifneq (,$(filter $(UNAME),Linux Darwin GNU/kFreeBSD GNU OpenBSD FreeBSD NetBSD DragonFly SunOS Haiku AIX))
-
-HAVE_COLORNEVER = $(shell echo a | egrep --color=never a > /dev/null 2> /dev/null && echo 1 || echo 0)
-EGREP_OPTIONS ?=
-ifeq ($HAVE_COLORNEVER, 1)
-  EGREP_OPTIONS += --color=never
-endif
-EGREP = egrep $(EGREP_OPTIONS)
-AWK = awk
-
-# Print a two column output of targets and their description. To add a target description, put a
-# comment in the Makefile with the format "## <TARGET>: <DESCRIPTION>".  For example:
-#
-## list: Print all targets and their descriptions (if provided)
-.PHONY: list
-list:
-	TARGETS=$$($(MAKE) -pRrq -f $(lastword $(MAKEFILE_LIST)) : 2>/dev/null \
-		| $(AWK) -v RS= -F: '/^# File/,/^# Finished Make data base/ {if ($$1 !~ "^[#.]") {print $$1}}' \
-		| $(EGREP) -v  -e '^[^[:alnum:]]' | sort); \
-	{ \
-	    printf "Target Name\tDescription\n"; \
-	    printf "%0.s-" {1..16}; printf "\t"; printf "%0.s-" {1..40}; printf "\n"; \
-	    for target in $$TARGETS; do \
-	        line=$$($(EGREP) "^##[[:space:]]+$$target:" $(lastword $(MAKEFILE_LIST))); \
-	        description=$$(echo $$line | $(AWK) '{i=index($$0,":"); print substr($$0,i+1)}' | xargs); \
-	        printf "$$target\t$$description\n"; \
-	    done \
-	} | column -t -s $$'\t'
-
-
-DESTDIR     ?=
-# directory variables : GNU conventions prefer lowercase
-# see https://www.gnu.org/prep/standards/html_node/Makefile-Conventions.html
-# support both lower and uppercase (BSD), use uppercase in script
-prefix      ?= /usr/local
-PREFIX      ?= $(prefix)
-exec_prefix ?= $(PREFIX)
-bindir      ?= $(exec_prefix)/bin
-BINDIR      ?= $(bindir)
-datarootdir ?= $(PREFIX)/share
-mandir      ?= $(datarootdir)/man
-man1dir     ?= $(mandir)/man1
-
-ifneq (,$(filter $(UNAME),OpenBSD FreeBSD NetBSD DragonFly SunOS))
-  MANDIR  ?= $(PREFIX)/man
-  MAN1DIR ?= $(MANDIR)/man1
-else
-  MAN1DIR ?= $(man1dir)
-endif
-
-ifneq (,$(filter $(UNAME),SunOS))
-  INSTALL ?= ginstall
-else
-  INSTALL ?= install
-endif
-
-INSTALL_PROGRAM ?= $(INSTALL)
-INSTALL_SCRIPT  ?= $(INSTALL_PROGRAM)
-INSTALL_DATA    ?= $(INSTALL) -m 644
-INSTALL_MAN     ?= $(INSTALL_DATA)
-
-.PHONY: install
-install:
-	# generate zstd only if not already present
-	[ -e zstd ] || $(MAKE) zstd-release
-	[ -e $(DESTDIR)$(BINDIR) ] || $(INSTALL) -d -m 755 $(DESTDIR)$(BINDIR)/
-	[ -e $(DESTDIR)$(MAN1DIR) ] || $(INSTALL) -d -m 755 $(DESTDIR)$(MAN1DIR)/
-	@echo Installing binaries
-	$(INSTALL_PROGRAM) zstd$(EXT) $(DESTDIR)$(BINDIR)/zstd$(EXT)
-	ln -sf zstd$(EXT) $(DESTDIR)$(BINDIR)/zstdcat$(EXT)
-	ln -sf zstd$(EXT) $(DESTDIR)$(BINDIR)/unzstd$(EXT)
-	ln -sf zstd$(EXT) $(DESTDIR)$(BINDIR)/zstdmt$(EXT)
-	$(INSTALL_SCRIPT) zstdless $(DESTDIR)$(BINDIR)/zstdless
-	$(INSTALL_SCRIPT) zstdgrep $(DESTDIR)$(BINDIR)/zstdgrep
-	@echo Installing man pages
-	$(INSTALL_MAN) zstd.1 $(DESTDIR)$(MAN1DIR)/zstd.1
-	ln -sf zstd.1 $(DESTDIR)$(MAN1DIR)/zstdcat.1
-	ln -sf zstd.1 $(DESTDIR)$(MAN1DIR)/unzstd.1
-	$(INSTALL_MAN) zstdgrep.1 $(DESTDIR)$(MAN1DIR)/zstdgrep.1
-	$(INSTALL_MAN) zstdless.1 $(DESTDIR)$(MAN1DIR)/zstdless.1
-	@echo zstd installation completed
-
-.PHONY: uninstall
-uninstall:
-	$(RM) $(DESTDIR)$(BINDIR)/zstdgrep
-	$(RM) $(DESTDIR)$(BINDIR)/zstdless
-	$(RM) $(DESTDIR)$(BINDIR)/zstdcat
-	$(RM) $(DESTDIR)$(BINDIR)/unzstd
-	$(RM) $(DESTDIR)$(BINDIR)/zstdmt
-	$(RM) $(DESTDIR)$(BINDIR)/zstd
-	$(RM) $(DESTDIR)$(MAN1DIR)/zstdless.1
-	$(RM) $(DESTDIR)$(MAN1DIR)/zstdgrep.1
-	$(RM) $(DESTDIR)$(MAN1DIR)/zstdcat.1
-	$(RM) $(DESTDIR)$(MAN1DIR)/unzstd.1
-	$(RM) $(DESTDIR)$(MAN1DIR)/zstd.1
-	@echo zstd programs successfully uninstalled
-
-endif
diff --git a/sys/contrib/zstd/programs/README.md b/sys/contrib/zstd/programs/README.md
index 5570f90c3b4f..43ef07a45bec 100644
--- a/sys/contrib/zstd/programs/README.md
+++ b/sys/contrib/zstd/programs/README.md
@@ -129,88 +129,131 @@ CLI includes in-memory compression benchmark module for zstd.
 The benchmark is conducted using given filenames. The files are read into memory and joined together.
 It makes benchmark more precise as it eliminates I/O overhead.
 Multiple filenames can be supplied, as multiple parameters, with wildcards,
-or names of directories can be used as parameters with `-r` option.
+or directory names can be used with `-r` option.
+If no file is provided, the benchmark will use a procedurally generated "lorem ipsum" content.
 
 The benchmark measures ratio, compressed size, compression and decompression speed.
 One can select compression levels starting from `-b` and ending with `-e`.
 The `-i` parameter selects minimal time used for each of tested levels.
 
+The benchmark can also be used to test specific parameters,
+such as number of threads (`-T#`), or advanced parameters (`--zstd=#`), or dictionary compression (`-D DICTIONARY`),
+and many others available on command for regular compression and decompression.
+
 
 ### Usage of Command Line Interface
 The full list of options can be obtained with `-h` or `-H` parameter:
 ```
-Usage :
-      zstd [args] [FILE(s)] [-o file]
+*** Zstandard CLI (64-bit) v1.5.6, by Yann Collet ***
+
+Compress or decompress the INPUT file(s); reads from STDIN if INPUT is `-` or not provided.
+
+Usage: zstd [OPTIONS...] [INPUT... | -] [-o OUTPUT]
+
+Options:
+  -o OUTPUT                     Write output to a single file, OUTPUT.
+  -k, --keep                    Preserve INPUT file(s). [Default]
+  --rm                          Remove INPUT file(s) after successful (de)compression.
+
+  -#                            Desired compression level, where `#` is a number between 1 and 19;
+                                lower numbers provide faster compression, higher numbers yield
+                                better compression ratios. [Default: 3]
+
+  -d, --decompress              Perform decompression.
+  -D DICT                       Use DICT as the dictionary for compression or decompression.
+
+  -f, --force                   Disable input and output checks. Allows overwriting existing files,
+                                receiving input from the console, printing output to STDOUT, and
+                                operating on links, block devices, etc. Unrecognized formats will be
+                                passed-through through as-is.
+
+  -h                            Display short usage and exit.
+  -H, --help                    Display full help and exit.
+  -V, --version                 Display the program version and exit.
+
+Advanced options:
+  -c, --stdout                  Write to STDOUT (even if it is a console) and keep the INPUT file(s).
+
+  -v, --verbose                 Enable verbose output; pass multiple times to increase verbosity.
+  -q, --quiet                   Suppress warnings; pass twice to suppress errors.
+  --trace LOG                   Log tracing information to LOG.
+
+  --[no-]progress               Forcibly show/hide the progress counter. NOTE: Any (de)compressed
+                                output to terminal will mix with progress counter text.
+
+  -r                            Operate recursively on directories.
+  --filelist LIST               Read a list of files to operate on from LIST.
+  --output-dir-flat DIR         Store processed files in DIR.
+  --output-dir-mirror DIR       Store processed files in DIR, respecting original directory structure.
+  --[no-]asyncio                Use asynchronous IO. [Default: Enabled]
+
+  --[no-]check                  Add XXH64 integrity checksums during compression. [Default: Add, Validate]
+                                If `-d` is present, ignore/validate checksums during decompression.
+
+  --                            Treat remaining arguments after `--` as files.
+
+Advanced compression options:
+  --ultra                       Enable levels beyond 19, up to 22; requires more memory.
+  --fast[=#]                    Use to very fast compression levels. [Default: 1]
+  --adapt                       Dynamically adapt compression level to I/O conditions.
+  --long[=#]                    Enable long distance matching with window log #. [Default: 27]
+  --patch-from=REF              Use REF as the reference point for Zstandard's diff engine.
+
+  -T#                           Spawn # compression threads. [Default: 1; pass 0 for core count.]
+  --single-thread               Share a single thread for I/O and compression (slightly different than `-T1`).
+  --auto-threads={physical|logical}
+                                Use physical/logical cores when using `-T0`. [Default: Physical]
+
+  -B#                           Set job size to #. [Default: 0 (automatic)]
+  --rsyncable                   Compress using a rsync-friendly method (`-B` sets block size).
+
+  --exclude-compressed          Only compress files that are not already compressed.
+
+  --stream-size=#               Specify size of streaming input from STDIN.
+  --size-hint=#                 Optimize compression parameters for streaming input of approximately size #.
+  --target-compressed-block-size=#
+                                Generate compressed blocks of approximately # size.
+
+  --no-dictID                   Don't write `dictID` into the header (dictionary compression only).
+  --[no-]compress-literals      Force (un)compressed literals.
+  --[no-]row-match-finder       Explicitly enable/disable the fast, row-based matchfinder for
+                                the 'greedy', 'lazy', and 'lazy2' strategies.
 
-FILE    : a filename
-          with no FILE, or when FILE is - , read standard input
-Arguments :
- -#     : # compression level (1-19, default: 3)
- -d     : decompression
- -D DICT: use DICT as Dictionary for compression or decompression
- -o file: result stored into `file` (only 1 output file)
- -f     : overwrite output without prompting, also (de)compress links
---rm    : remove source file(s) after successful de/compression
- -k     : preserve source file(s) (default)
- -h/-H  : display help/long help and exit
+  --format=zstd                 Compress files to the `.zst` format. [Default]
+  --[no-]mmap-dict              Memory-map dictionary file rather than mallocing and loading all at once
+  --format=gzip                 Compress files to the `.gz` format.
+  --format=xz                   Compress files to the `.xz` format.
+  --format=lzma                 Compress files to the `.lzma` format.
+  --format=lz4                 Compress files to the `.lz4` format.
 
-Advanced arguments :
- -V     : display Version number and exit
- -c     : write to standard output (even if it is the console)
- -v     : verbose mode; specify multiple times to increase verbosity
- -q     : suppress warnings; specify twice to suppress errors too
---no-progress : do not display the progress counter
- -r     : operate recursively on directories
---filelist FILE : read list of files to operate upon from FILE
---output-dir-flat DIR : processed files are stored into DIR
---output-dir-mirror DIR : processed files are stored into DIR respecting original directory structure
---[no-]check : during compression, add XXH64 integrity checksum to frame (default: enabled). If specified with -d, decompressor will ignore/validate checksums in compressed frame (default: validate).
---      : All arguments after "--" are treated as files
+Advanced decompression options:
+  -l                            Print information about Zstandard-compressed files.
+  --test                        Test compressed file integrity.
+  -M#                           Set the memory usage limit to # megabytes.
+  --[no-]sparse                 Enable sparse mode. [Default: Enabled for files, disabled for STDOUT.]
+  --[no-]pass-through           Pass through uncompressed files as-is. [Default: Disabled]
 
-Advanced compression arguments :
---ultra : enable levels beyond 19, up to 22 (requires more memory)
---long[=#]: enable long distance matching with given window log (default: 27)
---fast[=#]: switch to very fast compression levels (default: 1)
---adapt : dynamically adapt compression level to I/O conditions
---patch-from=FILE : specify the file to be used as a reference point for zstd's diff engine
- -T#    : spawns # compression threads (default: 1, 0==# cores)
- -B#    : select size of each job (default: 0==automatic)
---single-thread : use a single thread for both I/O and compression (result slightly different than -T1)
---rsyncable : compress using a rsync-friendly method (-B sets block size)
---exclude-compressed: only compress files that are not already compressed
---stream-size=# : specify size of streaming input from `stdin`
---size-hint=# optimize compression parameters for streaming input of approximately this size
---target-compressed-block-size=# : generate compressed block of approximately targeted size
---no-dictID : don't write dictID into header (dictionary compression only)
---[no-]compress-literals : force (un)compressed literals
---format=zstd : compress files to the .zst format (default)
---format=gzip : compress files to the .gz format
---format=xz : compress files to the .xz format
---format=lzma : compress files to the .lzma format
---format=lz4 : compress files to the .lz4 format
+Dictionary builder:
+  --train                       Create a dictionary from a training set of files.
 
-Advanced decompression arguments :
- -l     : print information about zstd compressed files
---test  : test compressed file integrity
- -M#    : Set a memory usage limit for decompression
---[no-]sparse : sparse mode (default: disabled)
+  --train-cover[=k=#,d=#,steps=#,split=#,shrink[=#]]
+                                Use the cover algorithm (with optional arguments).
+  --train-fastcover[=k=#,d=#,f=#,steps=#,split=#,accel=#,shrink[=#]]
+                                Use the fast cover algorithm (with optional arguments).
 
-Dictionary builder :
---train ## : create a dictionary from a training set of files
---train-cover[=k=#,d=#,steps=#,split=#,shrink[=#]] : use the cover algorithm with optional args
---train-fastcover[=k=#,d=#,f=#,steps=#,split=#,accel=#,shrink[=#]] : use the fast cover algorithm with optional args
---train-legacy[=s=#] : use the legacy algorithm with selectivity (default: 9)
- -o DICT : DICT is dictionary name (default: dictionary)
---maxdict=# : limit dictionary to specified size (default: 112640)
---dictID=# : force dictionary ID to specified value (default: random)
+  --train-legacy[=s=#]          Use the legacy algorithm with selectivity #. [Default: 9]
+  -o NAME                       Use NAME as dictionary name. [Default: dictionary]
+  --maxdict=#                   Limit dictionary to specified size #. [Default: 112640]
+  --dictID=#                    Force dictionary ID to #. [Default: Random]
 
-Benchmark arguments :
- -b#    : benchmark file(s), using # compression level (default: 3)
- -e#    : test all compression levels successively from -b# to -e# (default: 1)
- -i#    : minimum evaluation time in seconds (default: 3s)
- -B#    : cut file into independent blocks of size # (default: no block)
- -S     : output one benchmark result per input file (default: consolidated result)
---priority=rt : set process priority to real-time
+Benchmark options:
+  -b#                           Perform benchmarking with compression level #. [Default: 3]
+  -e#                           Test all compression levels up to #; starting level is `-b#`. [Default: 1]
+  -i#                           Set the minimum evaluation to time # seconds. [Default: 3]
+  -B#                           Cut file into independent chunks of size #. [Default: No chunking]
+  -S                            Output one benchmark result per input file. [Default: Consolidated result]
+  -D dictionary                 Benchmark using dictionary
+  --priority=rt                 Set process priority to real-time.
 ```
 
 ### Passing parameters through Environment Variables
@@ -275,7 +318,7 @@ compression speed (for lower levels) with minimal change in compression ratio.
 
 The below table illustrates this on the [Silesia compression corpus].
 
-[Silesia compression corpus]: http://sun.aei.polsl.pl/~sdeor/index.php?page=silesia
+[Silesia compression corpus]: https://sun.aei.polsl.pl//~sdeor/index.php?page=silesia
 
 | Method | Compression ratio | Compression speed | Decompression speed  |
 |:-------|------------------:|------------------:|---------------------:|
diff --git a/sys/contrib/zstd/programs/benchfn.c b/sys/contrib/zstd/programs/benchfn.c
index 1aadbdd9136c..3e042cf38f82 100644
--- a/sys/contrib/zstd/programs/benchfn.c
+++ b/sys/contrib/zstd/programs/benchfn.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -108,7 +108,6 @@ static BMK_runOutcome_t BMK_setValid_runTime(BMK_runTime_t runTime)
 BMK_runOutcome_t BMK_benchFunction(BMK_benchParams_t p,
                                    unsigned nbLoops)
 {
-    size_t dstSize = 0;
     nbLoops += !nbLoops;   /* minimum nbLoops is 1 */
 
     /* init */
@@ -118,7 +117,8 @@ BMK_runOutcome_t BMK_benchFunction(BMK_benchParams_t p,
     }   }
 
     /* benchmark */
-    {   UTIL_time_t const clockStart = UTIL_getTime();
+    {   size_t dstSize = 0;
+        UTIL_time_t const clockStart = UTIL_getTime();
         unsigned loopNb, blockNb;
         if (p.initFn != NULL) p.initFn(p.initPayload);
         for (loopNb = 0; loopNb < nbLoops; loopNb++) {
@@ -229,9 +229,9 @@ BMK_runOutcome_t BMK_benchTimedFn(BMK_timedFnState_t* cont,
             cont->timeSpent_ns += (unsigned long long)loopDuration_ns;
 
             /* estimate nbLoops for next run to last approximately 1 second */
-            if (loopDuration_ns > (runBudget_ns / 50)) {
+            if (loopDuration_ns > ((double)runBudget_ns / 50)) {
                 double const fastestRun_ns = MIN(bestRunTime.nanoSecPerRun, newRunTime.nanoSecPerRun);
-                cont->nbLoops = (unsigned)(runBudget_ns / fastestRun_ns) + 1;
+                cont->nbLoops = (unsigned)((double)runBudget_ns / fastestRun_ns) + 1;
             } else {
                 /* previous run was too short : blindly increase workload by x multiplier */
                 const unsigned multiplier = 10;
@@ -239,7 +239,7 @@ BMK_runOutcome_t BMK_benchTimedFn(BMK_timedFnState_t* cont,
                 cont->nbLoops *= multiplier;
             }
 
-            if(loopDuration_ns < runTimeMin_ns) {
+            if(loopDuration_ns < (double)runTimeMin_ns) {
                 /* don't report results for which benchmark run time was too small : increased risks of rounding errors */
                 assert(completed == 0);
                 continue;
diff --git a/sys/contrib/zstd/programs/benchfn.h b/sys/contrib/zstd/programs/benchfn.h
index 590f292eaa61..3fc6e0d04551 100644
--- a/sys/contrib/zstd/programs/benchfn.h
+++ b/sys/contrib/zstd/programs/benchfn.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -15,17 +15,12 @@
  * or detecting and returning an error
  */
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
 #ifndef BENCH_FN_H_23876
 #define BENCH_FN_H_23876
 
 /* ===  Dependencies  === */
 #include <stddef.h>   /* size_t */
 
-
 /* ====  Benchmark any function, iterated on a set of blocks  ==== */
 
 /* BMK_runTime_t: valid result return type */
@@ -123,7 +118,7 @@ BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome);
 /* when benchmark failed, it means one invocation of `benchFn` failed.
  * The failure was detected by `errorFn`, operating on return values of `benchFn`.
  * Returns the faulty return value.
- * note : this function will abort() program execution if benchmark did not failed.
+ * note : this function will abort() program execution if benchmark did not fail.
  *        always check if benchmark failed first !
  */
 size_t BMK_extract_errorResult(BMK_runOutcome_t outcome);
@@ -175,9 +170,4 @@ typedef union {
 } BMK_timedFnState_shell;
 BMK_timedFnState_t* BMK_initStatic_timedFnState(void* buffer, size_t size, unsigned total_ms, unsigned run_ms);
 
-
 #endif   /* BENCH_FN_H_23876 */
-
-#if defined (__cplusplus)
-}
-#endif
diff --git a/sys/contrib/zstd/programs/benchzstd.c b/sys/contrib/zstd/programs/benchzstd.c
index fa2659efbbba..f9274a5172c1 100644
--- a/sys/contrib/zstd/programs/benchzstd.c
+++ b/sys/contrib/zstd/programs/benchzstd.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -8,197 +8,295 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
-
 /* **************************************
-*  Tuning parameters
-****************************************/
-#ifndef BMK_TIMETEST_DEFAULT_S   /* default minimum time per test */
-#define BMK_TIMETEST_DEFAULT_S 3
+ *  Tuning parameters
+ ****************************************/
+#ifndef BMK_TIMETEST_DEFAULT_S /* default minimum time per test */
+#    define BMK_TIMETEST_DEFAULT_S 3
 #endif
 
-
 /* *************************************
-*  Includes
-***************************************/
-#include "platform.h"    /* Large Files support */
-#include "util.h"        /* UTIL_getFileSize, UTIL_sleep */
-#include <stdlib.h>      /* malloc, free */
-#include <string.h>      /* memset, strerror */
-#include <stdio.h>       /* fprintf, fopen */
-#include <errno.h>
-#include <assert.h>      /* assert */
+ *  Includes
+ ***************************************/
+/* this must be included first */
+#include "platform.h" /* Large Files support, compiler specifics */
 
-#include "timefn.h"      /* UTIL_time_t */
-#include "benchfn.h"
+/* then following system includes */
+#include <assert.h> /* assert */
+#include <errno.h>
+#include <stdio.h>    /* fprintf, fopen */
+#include <stdlib.h>   /* malloc, free */
+#include <string.h>   /* memset, strerror */
+#include "util.h"     /* UTIL_getFileSize, UTIL_sleep */
 #include "../lib/common/mem.h"
+#include "benchfn.h"
+#include "timefn.h" /* UTIL_time_t */
 #ifndef ZSTD_STATIC_LINKING_ONLY
-#define ZSTD_STATIC_LINKING_ONLY
+#    define ZSTD_STATIC_LINKING_ONLY
 #endif
 #include "../lib/zstd.h"
-#include "datagen.h"     /* RDG_genBuffer */
+#include "datagen.h" /* RDG_genBuffer */
+#include "lorem.h"   /* LOREM_genBuffer */
 #ifndef XXH_INLINE_ALL
-#define XXH_INLINE_ALL
+#    define XXH_INLINE_ALL
 #endif
 #include "../lib/common/xxhash.h"
-#include "benchzstd.h"
 #include "../lib/zstd_errors.h"
-
+#include "benchzstd.h"
 
 /* *************************************
-*  Constants
-***************************************/
+ *  Constants
+ ***************************************/
 #ifndef ZSTD_GIT_COMMIT
-#  define ZSTD_GIT_COMMIT_STRING ""
+#    define ZSTD_GIT_COMMIT_STRING ""
 #else
-#  define ZSTD_GIT_COMMIT_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_GIT_COMMIT)
+#    define ZSTD_GIT_COMMIT_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_GIT_COMMIT)
 #endif
 
-#define TIMELOOP_MICROSEC     (1*1000000ULL) /* 1 second */
-#define TIMELOOP_NANOSEC      (1*1000000000ULL) /* 1 second */
-#define ACTIVEPERIOD_MICROSEC (70*TIMELOOP_MICROSEC) /* 70 seconds */
-#define COOLPERIOD_SEC        10
+#define TIMELOOP_MICROSEC (1 * 1000000ULL)             /* 1 second */
+#define TIMELOOP_NANOSEC (1 * 1000000000ULL)           /* 1 second */
+#define ACTIVEPERIOD_MICROSEC (70 * TIMELOOP_MICROSEC) /* 70 seconds */
+#define COOLPERIOD_SEC 10
 
-#define KB *(1 <<10)
-#define MB *(1 <<20)
-#define GB *(1U<<30)
+#define KB *(1 << 10)
+#define MB *(1 << 20)
+#define GB *(1U << 30)
 
 #define BMK_RUNTEST_DEFAULT_MS 1000
 
-static const size_t maxMemory = (sizeof(size_t)==4)  ?
-                    /* 32-bit */ (2 GB - 64 MB) :
-                    /* 64-bit */ (size_t)(1ULL << ((sizeof(size_t)*8)-31));
-
+static const size_t maxMemory = (sizeof(size_t) == 4)
+        ?
+        /* 32-bit */ (2 GB - 64 MB)
+        :
+        /* 64-bit */ (size_t)(1ULL << ((sizeof(size_t) * 8) - 31));
 
 /* *************************************
-*  console display
-***************************************/
-#define DISPLAY(...)         { fprintf(stderr, __VA_ARGS__); fflush(NULL); }
-#define DISPLAYLEVEL(l, ...) if (displayLevel>=l) { DISPLAY(__VA_ARGS__); }
-/* 0 : no display;   1: errors;   2 : + result + interaction + warnings;   3 : + progression;   4 : + information */
-#define OUTPUT(...)          { fprintf(stdout, __VA_ARGS__); fflush(NULL); }
-#define OUTPUTLEVEL(l, ...)  if (displayLevel>=l) { OUTPUT(__VA_ARGS__); }
-
+ *  console display
+ ***************************************/
+#define DISPLAY(...)                  \
+    {                                 \
+        fprintf(stderr, __VA_ARGS__); \
+        fflush(NULL);                 \
+    }
+#define DISPLAYLEVEL(l, ...)  \
+    if (displayLevel >= l) {  \
+        DISPLAY(__VA_ARGS__); \
+    }
+/* 0 : no display;   1: errors;   2 : + result + interaction + warnings;   3 : +
+ * progression;   4 : + information */
+#define OUTPUT(...)                   \
+    {                                 \
+        fprintf(stdout, __VA_ARGS__); \
+        fflush(NULL);                 \
+    }
+#define OUTPUTLEVEL(l, ...)  \
+    if (displayLevel >= l) { \
+        OUTPUT(__VA_ARGS__); \
+    }
 
 /* *************************************
-*  Exceptions
-***************************************/
+ *  Exceptions
+ ***************************************/
 #ifndef DEBUG
-#  define DEBUG 0
+#    define DEBUG 0
 #endif
-#define DEBUGOUTPUT(...) { if (DEBUG) DISPLAY(__VA_ARGS__); }
+#define DEBUGOUTPUT(...)          \
+    {                             \
+        if (DEBUG)                \
+            DISPLAY(__VA_ARGS__); \
+    }
 
-#define RETURN_ERROR_INT(errorNum, ...)  {               \
-    DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);    \
-    DISPLAYLEVEL(1, "Error %i : ", errorNum);         \
-    DISPLAYLEVEL(1, __VA_ARGS__);                     \
-    DISPLAYLEVEL(1, " \n");                           \
-    return errorNum;                                  \
-}
+#define RETURN_ERROR_INT(errorNum, ...)                \
+    {                                                  \
+        DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__); \
+        DISPLAYLEVEL(1, "Error %i : ", errorNum);      \
+        DISPLAYLEVEL(1, __VA_ARGS__);                  \
+        DISPLAYLEVEL(1, " \n");                        \
+        return errorNum;                               \
+    }
+
+#define CHECK_Z(zf)                                                  \
+    {                                                                \
+        size_t const zerr = zf;                                      \
+        if (ZSTD_isError(zerr)) {                                    \
+            DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);           \
+            DISPLAY("Error : ");                                     \
+            DISPLAY("%s failed : %s", #zf, ZSTD_getErrorName(zerr)); \
+            DISPLAY(" \n");                                          \
+            exit(1);                                                 \
+        }                                                            \
+    }
 
-#define CHECK_Z(zf) {              \
-    size_t const zerr = zf;        \
-    if (ZSTD_isError(zerr)) {      \
-        DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);  \
-        DISPLAY("Error : ");       \
-        DISPLAY("%s failed : %s",  \
-                #zf, ZSTD_getErrorName(zerr));   \
-        DISPLAY(" \n");            \
-        exit(1);                   \
-    }                              \
+#define RETURN_ERROR(errorNum, retType, ...)           \
+    {                                                  \
+        retType r;                                     \
+        memset(&r, 0, sizeof(retType));                \
+        DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__); \
+        DISPLAYLEVEL(1, "Error %i : ", errorNum);      \
+        DISPLAYLEVEL(1, __VA_ARGS__);                  \
+        DISPLAYLEVEL(1, " \n");                        \
+        r.tag = errorNum;                              \
+        return r;                                      \
+    }
+
+static size_t uintSize(unsigned value)
+{
+    size_t size = 1;
+    while (value >= 10) {
+        size++;
+        value /= 10;
+    }
+    return size;
 }
 
-#define RETURN_ERROR(errorNum, retType, ...)  {       \
-    retType r;                                        \
-    memset(&r, 0, sizeof(retType));                   \
-    DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__);    \
-    DISPLAYLEVEL(1, "Error %i : ", errorNum);         \
-    DISPLAYLEVEL(1, __VA_ARGS__);                     \
-    DISPLAYLEVEL(1, " \n");                           \
-    r.tag = errorNum;                                 \
-    return r;                                         \
+/* Note: presume @buffer is large enough */
+static void writeUint_varLen(char* buffer, size_t capacity, unsigned value)
+{
+    int endPos = (int)uintSize(value) - 1;
+    assert(uintSize(value) >= 1);
+    assert(uintSize(value) < capacity); (void)capacity;
+    while (endPos >= 0) {
+        char c = '0' + (char)(value % 10);
+        buffer[endPos--] = c;
+        value /= 10;
+    }
 }
 
+/* replacement for snprintf(), which is not supported by C89.
+ * sprintf() would be the supported one, but it's labelled unsafe:
+ * modern static analyzer will flag sprintf() as dangerous, making it unusable.
+ * formatString_u() replaces snprintf() for the specific case where there is only one %u argument */
+static int formatString_u(char* buffer, size_t buffer_size, const char* formatString, unsigned int value)
+{
+    size_t const valueSize = uintSize(value);
+    size_t written = 0;
+    int i;
+
+    for (i = 0; formatString[i] != '\0' && written < buffer_size - 1; i++) {
+        if (formatString[i] != '%') {
+            buffer[written++] = formatString[i];
+            continue;
+        }
+
+        i++;
+        if (formatString[i] == 'u') {
+            if (written + valueSize >= buffer_size) abort(); /* buffer not large enough */
+            writeUint_varLen(buffer + written, buffer_size - written, value);
+            written += valueSize;
+        } else if (formatString[i] == '%') { /* Check for escaped percent sign */
+            buffer[written++] = '%';
+        } else {
+            abort(); /* unsupported format */
+        }
+    }
+
+    if (written < buffer_size) {
+        buffer[written] = '\0';
+    } else {
+        abort(); /* buffer not large enough */
+    }
+
+    return (int)written;
+}
 
 /* *************************************
-*  Benchmark Parameters
-***************************************/
+ *  Benchmark Parameters
+ ***************************************/
 
-BMK_advancedParams_t BMK_initAdvancedParams(void) {
+BMK_advancedParams_t BMK_initAdvancedParams(void)
+{
     BMK_advancedParams_t const res = {
-        BMK_both, /* mode */
+        BMK_both,               /* mode */
         BMK_TIMETEST_DEFAULT_S, /* nbSeconds */
-        0, /* blockSize */
-        0, /* nbWorkers */
-        0, /* realTime */
-        0, /* additionalParam */
-        0, /* ldmFlag */
-        0, /* ldmMinMatch */
-        0, /* ldmHashLog */
-        0, /* ldmBuckSizeLog */
-        0,  /* ldmHashRateLog */
-        ZSTD_ps_auto, /* literalCompressionMode */
-        0 /* useRowMatchFinder */
+        0,                      /* blockSize */
+        0,               /* targetCBlockSize */
+        0,                      /* nbWorkers */
+        0,                      /* realTime */
+        0,                      /* additionalParam */
+        0,                      /* ldmFlag */
+        0,                      /* ldmMinMatch */
+        0,                      /* ldmHashLog */
+        0,                      /* ldmBuckSizeLog */
+        0,                      /* ldmHashRateLog */
+        ZSTD_ps_auto,           /* literalCompressionMode */
+        0                       /* useRowMatchFinder */
     };
     return res;
 }
 
-
 /* ********************************************************
-*  Bench functions
-**********************************************************/
+ *  Bench functions
+ **********************************************************/
 typedef struct {
     const void* srcPtr;
     size_t srcSize;
-    void*  cPtr;
+    void* cPtr;
     size_t cRoom;
     size_t cSize;
-    void*  resPtr;
+    void* resPtr;
     size_t resSize;
 } blockParam_t;
 
 #undef MIN
 #undef MAX
-#define MIN(a,b)    ((a) < (b) ? (a) : (b))
-#define MAX(a,b)    ((a) > (b) ? (a) : (b))
+#define MIN(a, b) ((a) < (b) ? (a) : (b))
+#define MAX(a, b) ((a) > (b) ? (a) : (b))
 
-static void
-BMK_initCCtx(ZSTD_CCtx* ctx,
-            const void* dictBuffer, size_t dictBufferSize,
-            int cLevel,
-            const ZSTD_compressionParameters* comprParams,
-            const BMK_advancedParams_t* adv)
+static void BMK_initCCtx(
+        ZSTD_CCtx* ctx,
+        const void* dictBuffer,
+        size_t dictBufferSize,
+        int cLevel,
+        const ZSTD_compressionParameters* comprParams,
+        const BMK_advancedParams_t* adv)
 {
     ZSTD_CCtx_reset(ctx, ZSTD_reset_session_and_parameters);
-    if (adv->nbWorkers==1) {
+    if (adv->nbWorkers == 1) {
         CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_nbWorkers, 0));
     } else {
         CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_nbWorkers, adv->nbWorkers));
     }
     CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_compressionLevel, cLevel));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_useRowMatchFinder, adv->useRowMatchFinder));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_enableLongDistanceMatching, adv->ldmFlag));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_useRowMatchFinder, adv->useRowMatchFinder));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_enableLongDistanceMatching, adv->ldmFlag));
     CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmMinMatch, adv->ldmMinMatch));
     CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmHashLog, adv->ldmHashLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmBucketSizeLog, adv->ldmBucketSizeLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_ldmHashRateLog, adv->ldmHashRateLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_windowLog, (int)comprParams->windowLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_hashLog, (int)comprParams->hashLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_chainLog, (int)comprParams->chainLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_searchLog, (int)comprParams->searchLog));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_minMatch, (int)comprParams->minMatch));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_targetLength, (int)comprParams->targetLength));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_literalCompressionMode, (int)adv->literalCompressionMode));
-    CHECK_Z(ZSTD_CCtx_setParameter(ctx, ZSTD_c_strategy, (int)comprParams->strategy));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_ldmBucketSizeLog, adv->ldmBucketSizeLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_ldmHashRateLog, adv->ldmHashRateLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_windowLog, (int)comprParams->windowLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_hashLog, (int)comprParams->hashLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_chainLog, (int)comprParams->chainLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_searchLog, (int)comprParams->searchLog));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_minMatch, (int)comprParams->minMatch));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_targetLength, (int)comprParams->targetLength));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx,
+            ZSTD_c_literalCompressionMode,
+            (int)adv->literalCompressionMode));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_strategy, (int)comprParams->strategy));
+    CHECK_Z(ZSTD_CCtx_setParameter(
+            ctx, ZSTD_c_targetCBlockSize, (int)adv->targetCBlockSize));
     CHECK_Z(ZSTD_CCtx_loadDictionary(ctx, dictBuffer, dictBufferSize));
 }
 
-static void BMK_initDCtx(ZSTD_DCtx* dctx,
-    const void* dictBuffer, size_t dictBufferSize) {
+static void
+BMK_initDCtx(ZSTD_DCtx* dctx, const void* dictBuffer, size_t dictBufferSize)
+{
     CHECK_Z(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_and_parameters));
     CHECK_Z(ZSTD_DCtx_loadDictionary(dctx, dictBuffer, dictBufferSize));
 }
 
-
 typedef struct {
     ZSTD_CCtx* cctx;
     const void* dictBuffer;
@@ -208,9 +306,16 @@ typedef struct {
     const BMK_advancedParams_t* adv;
 } BMK_initCCtxArgs;
 
-static size_t local_initCCtx(void* payload) {
+static size_t local_initCCtx(void* payload)
+{
     BMK_initCCtxArgs* ag = (BMK_initCCtxArgs*)payload;
-    BMK_initCCtx(ag->cctx, ag->dictBuffer, ag->dictBufferSize, ag->cLevel, ag->comprParams, ag->adv);
+    BMK_initCCtx(
+            ag->cctx,
+            ag->dictBuffer,
+            ag->dictBufferSize,
+            ag->cLevel,
+            ag->comprParams,
+            ag->adv);
     return 0;
 }
 
@@ -220,18 +325,20 @@ typedef struct {
     size_t dictBufferSize;
 } BMK_initDCtxArgs;
 
-static size_t local_initDCtx(void* payload) {
+static size_t local_initDCtx(void* payload)
+{
     BMK_initDCtxArgs* ag = (BMK_initDCtxArgs*)payload;
     BMK_initDCtx(ag->dctx, ag->dictBuffer, ag->dictBufferSize);
     return 0;
 }
 
-
 /* `addArgs` is the context */
 static size_t local_defaultCompress(
-                    const void* srcBuffer, size_t srcSize,
-                    void* dstBuffer, size_t dstSize,
-                    void* addArgs)
+        const void* srcBuffer,
+        size_t srcSize,
+        void* dstBuffer,
+        size_t dstSize,
+        void* addArgs)
 {
     ZSTD_CCtx* const cctx = (ZSTD_CCtx*)addArgs;
     return ZSTD_compress2(cctx, dstBuffer, dstSize, srcBuffer, srcSize);
@@ -239,18 +346,24 @@ static size_t local_defaultCompress(
 
 /* `addArgs` is the context */
 static size_t local_defaultDecompress(
-                    const void* srcBuffer, size_t srcSize,
-                    void* dstBuffer, size_t dstCapacity,
-                    void* addArgs)
+        const void* srcBuffer,
+        size_t srcSize,
+        void* dstBuffer,
+        size_t dstCapacity,
+        void* addArgs)
 {
-    size_t moreToFlush = 1;
+    size_t moreToFlush    = 1;
     ZSTD_DCtx* const dctx = (ZSTD_DCtx*)addArgs;
     ZSTD_inBuffer in;
     ZSTD_outBuffer out;
-    in.src = srcBuffer; in.size = srcSize; in.pos = 0;
-    out.dst = dstBuffer; out.size = dstCapacity; out.pos = 0;
+    in.src   = srcBuffer;
+    in.size  = srcSize;
+    in.pos   = 0;
+    out.dst  = dstBuffer;
+    out.size = dstCapacity;
+    out.pos  = 0;
     while (moreToFlush) {
-        if(out.pos == out.size) {
+        if (out.pos == out.size) {
             return (size_t)-ZSTD_error_dstSize_tooSmall;
         }
         moreToFlush = ZSTD_decompressStream(dctx, &out, &in);
@@ -259,10 +372,8 @@ static size_t local_defaultDecompress(
         }
     }
     return out.pos;
-
 }
 
-
 /* ================================================================= */
 /*      Benchmark Zstandard, mem-to-mem scenarios                    */
 /* ================================================================= */
@@ -286,99 +397,145 @@ static BMK_benchOutcome_t BMK_benchOutcome_error(void)
     return b;
 }
 
-static BMK_benchOutcome_t BMK_benchOutcome_setValidResult(BMK_benchResult_t result)
+static BMK_benchOutcome_t BMK_benchOutcome_setValidResult(
+        BMK_benchResult_t result)
 {
     BMK_benchOutcome_t b;
-    b.tag = 0;
+    b.tag                         = 0;
     b.internal_never_use_directly = result;
     return b;
 }
 
-
 /* benchMem with no allocation */
-static BMK_benchOutcome_t
-BMK_benchMemAdvancedNoAlloc(
-                    const void** srcPtrs, size_t* srcSizes,
-                    void** cPtrs, size_t* cCapacities, size_t* cSizes,
-                    void** resPtrs, size_t* resSizes,
-                    void** resultBufferPtr, void* compressedBuffer,
-                    size_t maxCompressedSize,
-                    BMK_timedFnState_t* timeStateCompress,
-                    BMK_timedFnState_t* timeStateDecompress,
+static BMK_benchOutcome_t BMK_benchMemAdvancedNoAlloc(
+        const void** srcPtrs,
+        size_t* srcSizes,
+        void** cPtrs,
+        size_t* cCapacities,
+        size_t* cSizes,
+        void** resPtrs,
+        size_t* resSizes,
+        void** resultBufferPtr,
+        void* compressedBuffer,
+        size_t maxCompressedSize,
+        BMK_timedFnState_t* timeStateCompress,
+        BMK_timedFnState_t* timeStateDecompress,
 
-                    const void* srcBuffer, size_t srcSize,
-                    const size_t* fileSizes, unsigned nbFiles,
-                    const int cLevel,
-                    const ZSTD_compressionParameters* comprParams,
-                    const void* dictBuffer, size_t dictBufferSize,
-                    ZSTD_CCtx* cctx, ZSTD_DCtx* dctx,
-                    int displayLevel, const char* displayName,
-                    const BMK_advancedParams_t* adv)
+        const void* srcBuffer,
+        size_t srcSize,
+        const size_t* fileSizes,
+        unsigned nbFiles,
+        const int cLevel,
+        const ZSTD_compressionParameters* comprParams,
+        const void* dictBuffer,
+        size_t dictBufferSize,
+        ZSTD_CCtx* cctx,
+        ZSTD_DCtx* dctx,
+        int displayLevel,
+        const char* displayName,
+        const BMK_advancedParams_t* adv)
 {
-    size_t const blockSize = ((adv->blockSize>=32 && (adv->mode != BMK_decodeOnly)) ? adv->blockSize : srcSize) + (!srcSize);  /* avoid div by 0 */
+    size_t const blockSize =
+            ((adv->blockSize >= 32 && (adv->mode != BMK_decodeOnly))
+                     ? adv->blockSize
+                     : srcSize)
+            + (!srcSize); /* avoid div by 0 */
     BMK_benchResult_t benchResult;
     size_t const loadedCompressedSize = srcSize;
-    size_t cSize = 0;
-    double ratio = 0.;
+    size_t cSize                      = 0;
+    double ratio                      = 0.;
     U32 nbBlocks;
 
-    assert(cctx != NULL); assert(dctx != NULL);
+    assert(cctx != NULL);
+    assert(dctx != NULL);
 
     /* init */
     memset(&benchResult, 0, sizeof(benchResult));
-    if (strlen(displayName)>17) displayName += strlen(displayName) - 17;   /* display last 17 characters */
-    if (adv->mode == BMK_decodeOnly) {  /* benchmark only decompression : source must be already compressed */
+    if (strlen(displayName) > 17)
+        displayName +=
+                strlen(displayName) - 17; /* display last 17 characters */
+    if (adv->mode == BMK_decodeOnly) {
+        /* benchmark only decompression : source must be already compressed */
         const char* srcPtr = (const char*)srcBuffer;
-        U64 totalDSize64 = 0;
+        U64 totalDSize64   = 0;
         U32 fileNb;
-        for (fileNb=0; fileNb<nbFiles; fileNb++) {
-            U64 const fSize64 = ZSTD_findDecompressedSize(srcPtr, fileSizes[fileNb]);
-            if (fSize64==0) RETURN_ERROR(32, BMK_benchOutcome_t, "Impossible to determine original size ");
+        for (fileNb = 0; fileNb < nbFiles; fileNb++) {
+            U64 const fSize64 =
+                    ZSTD_findDecompressedSize(srcPtr, fileSizes[fileNb]);
+            if (fSize64 == ZSTD_CONTENTSIZE_UNKNOWN) {
+                RETURN_ERROR(
+                        32,
+                        BMK_benchOutcome_t,
+                        "Decompressed size cannot be determined: cannot benchmark");
+            }
+            if (fSize64 == ZSTD_CONTENTSIZE_ERROR) {
+                RETURN_ERROR(
+                        32,
+                        BMK_benchOutcome_t,
+                        "Error while trying to assess decompressed size: data may be invalid");
+            }
             totalDSize64 += fSize64;
             srcPtr += fileSizes[fileNb];
         }
-        {   size_t const decodedSize = (size_t)totalDSize64;
-            assert((U64)decodedSize == totalDSize64);   /* check overflow */
+        {
+            size_t const decodedSize = (size_t)totalDSize64;
+            assert((U64)decodedSize == totalDSize64); /* check overflow */
             free(*resultBufferPtr);
+            if (totalDSize64 > decodedSize) { /* size_t overflow */
+                RETURN_ERROR(
+                        32,
+                        BMK_benchOutcome_t,
+                        "decompressed size is too large for local system");
+            }
             *resultBufferPtr = malloc(decodedSize);
             if (!(*resultBufferPtr)) {
-                RETURN_ERROR(33, BMK_benchOutcome_t, "not enough memory");
-            }
-            if (totalDSize64 > decodedSize) {  /* size_t overflow */
-                free(*resultBufferPtr);
-                RETURN_ERROR(32, BMK_benchOutcome_t, "original size is too large");
+                RETURN_ERROR(
+                        33,
+                        BMK_benchOutcome_t,
+                        "allocation error: not enough memory");
             }
-            cSize = srcSize;
+            cSize   = srcSize;
             srcSize = decodedSize;
-            ratio = (double)srcSize / (double)cSize;
+            ratio   = (double)srcSize / (double)cSize;
         }
     }
 
     /* Init data blocks  */
-    {   const char* srcPtr = (const char*)srcBuffer;
-        char* cPtr = (char*)compressedBuffer;
-        char* resPtr = (char*)(*resultBufferPtr);
+    {
+        const char* srcPtr = (const char*)srcBuffer;
+        char* cPtr         = (char*)compressedBuffer;
+        char* resPtr       = (char*)(*resultBufferPtr);
         U32 fileNb;
-        for (nbBlocks=0, fileNb=0; fileNb<nbFiles; fileNb++) {
-            size_t remaining = fileSizes[fileNb];
-            U32 const nbBlocksforThisFile = (adv->mode == BMK_decodeOnly) ? 1 : (U32)((remaining + (blockSize-1)) / blockSize);
-            U32 const blockEnd = nbBlocks + nbBlocksforThisFile;
-            for ( ; nbBlocks<blockEnd; nbBlocks++) {
+        for (nbBlocks = 0, fileNb = 0; fileNb < nbFiles; fileNb++) {
+            size_t remaining              = fileSizes[fileNb];
+            U32 const nbBlocksforThisFile = (adv->mode == BMK_decodeOnly)
+                    ? 1
+                    : (U32)((remaining + (blockSize - 1)) / blockSize);
+            U32 const blockEnd            = nbBlocks + nbBlocksforThisFile;
+            for (; nbBlocks < blockEnd; nbBlocks++) {
                 size_t const thisBlockSize = MIN(remaining, blockSize);
-                srcPtrs[nbBlocks] = srcPtr;
-                srcSizes[nbBlocks] = thisBlockSize;
-                cPtrs[nbBlocks] = cPtr;
-                cCapacities[nbBlocks] = (adv->mode == BMK_decodeOnly) ? thisBlockSize : ZSTD_compressBound(thisBlockSize);
-                resPtrs[nbBlocks] = resPtr;
-                resSizes[nbBlocks] = (adv->mode == BMK_decodeOnly) ? (size_t) ZSTD_findDecompressedSize(srcPtr, thisBlockSize) : thisBlockSize;
+                srcPtrs[nbBlocks]          = srcPtr;
+                srcSizes[nbBlocks]         = thisBlockSize;
+                cPtrs[nbBlocks]            = cPtr;
+                cCapacities[nbBlocks]      = (adv->mode == BMK_decodeOnly)
+                             ? thisBlockSize
+                             : ZSTD_compressBound(thisBlockSize);
+                resPtrs[nbBlocks]          = resPtr;
+                resSizes[nbBlocks]         = (adv->mode == BMK_decodeOnly)
+                                ? (size_t)ZSTD_findDecompressedSize(
+                                srcPtr, thisBlockSize)
+                                : thisBlockSize;
                 srcPtr += thisBlockSize;
                 cPtr += cCapacities[nbBlocks];
                 resPtr += thisBlockSize;
                 remaining -= thisBlockSize;
                 if (adv->mode == BMK_decodeOnly) {
-                    cSizes[nbBlocks] = thisBlockSize;
+                    cSizes[nbBlocks]  = thisBlockSize;
                     benchResult.cSize = thisBlockSize;
-    }   }   }   }
+                }
+            }
+        }
+    }
 
     /* warming up `compressedBuffer` */
     if (adv->mode == BMK_decodeOnly) {
@@ -387,233 +544,330 @@ BMK_benchMemAdvancedNoAlloc(
         RDG_genBuffer(compressedBuffer, maxCompressedSize, 0.10, 0.50, 1);
     }
 
+    if (!UTIL_support_MT_measurements() && adv->nbWorkers > 1) {
+        OUTPUTLEVEL(
+                2,
+                "Warning : time measurements may be incorrect in multithreading mode... \n")
+    }
+
     /* Bench */
-    {   U64 const crcOrig = (adv->mode == BMK_decodeOnly) ? 0 : XXH64(srcBuffer, srcSize, 0);
-#       define NB_MARKS 4
+    {
+        U64 const crcOrig = (adv->mode == BMK_decodeOnly)
+                ? 0
+                : XXH64(srcBuffer, srcSize, 0);
+#define NB_MARKS 4
         const char* marks[NB_MARKS] = { " |", " /", " =", " \\" };
-        U32 markNb = 0;
-        int compressionCompleted = (adv->mode == BMK_decodeOnly);
-        int decompressionCompleted = (adv->mode == BMK_compressOnly);
+        U32 markNb                  = 0;
+        int compressionCompleted    = (adv->mode == BMK_decodeOnly);
+        int decompressionCompleted  = (adv->mode == BMK_compressOnly);
         BMK_benchParams_t cbp, dbp;
         BMK_initCCtxArgs cctxprep;
         BMK_initDCtxArgs dctxprep;
 
-        cbp.benchFn = local_defaultCompress;   /* ZSTD_compress2 */
-        cbp.benchPayload = cctx;
-        cbp.initFn = local_initCCtx;   /* BMK_initCCtx */
-        cbp.initPayload = &cctxprep;
-        cbp.errorFn = ZSTD_isError;
-        cbp.blockCount = nbBlocks;
-        cbp.srcBuffers = srcPtrs;
-        cbp.srcSizes = srcSizes;
-        cbp.dstBuffers = cPtrs;
+        cbp.benchFn       = local_defaultCompress; /* ZSTD_compress2 */
+        cbp.benchPayload  = cctx;
+        cbp.initFn        = local_initCCtx; /* BMK_initCCtx */
+        cbp.initPayload   = &cctxprep;
+        cbp.errorFn       = ZSTD_isError;
+        cbp.blockCount    = nbBlocks;
+        cbp.srcBuffers    = srcPtrs;
+        cbp.srcSizes      = srcSizes;
+        cbp.dstBuffers    = cPtrs;
         cbp.dstCapacities = cCapacities;
-        cbp.blockResults = cSizes;
+        cbp.blockResults  = cSizes;
 
-        cctxprep.cctx = cctx;
-        cctxprep.dictBuffer = dictBuffer;
+        cctxprep.cctx           = cctx;
+        cctxprep.dictBuffer     = dictBuffer;
         cctxprep.dictBufferSize = dictBufferSize;
-        cctxprep.cLevel = cLevel;
-        cctxprep.comprParams = comprParams;
-        cctxprep.adv = adv;
+        cctxprep.cLevel         = cLevel;
+        cctxprep.comprParams    = comprParams;
+        cctxprep.adv            = adv;
 
-        dbp.benchFn = local_defaultDecompress;
-        dbp.benchPayload = dctx;
-        dbp.initFn = local_initDCtx;
-        dbp.initPayload = &dctxprep;
-        dbp.errorFn = ZSTD_isError;
-        dbp.blockCount = nbBlocks;
-        dbp.srcBuffers = (const void* const *) cPtrs;
-        dbp.srcSizes = cSizes;
-        dbp.dstBuffers = resPtrs;
+        dbp.benchFn       = local_defaultDecompress;
+        dbp.benchPayload  = dctx;
+        dbp.initFn        = local_initDCtx;
+        dbp.initPayload   = &dctxprep;
+        dbp.errorFn       = ZSTD_isError;
+        dbp.blockCount    = nbBlocks;
+        dbp.srcBuffers    = (const void* const*)cPtrs;
+        dbp.srcSizes      = cSizes;
+        dbp.dstBuffers    = resPtrs;
         dbp.dstCapacities = resSizes;
-        dbp.blockResults = NULL;
+        dbp.blockResults  = NULL;
 
-        dctxprep.dctx = dctx;
-        dctxprep.dictBuffer = dictBuffer;
+        dctxprep.dctx           = dctx;
+        dctxprep.dictBuffer     = dictBuffer;
         dctxprep.dictBufferSize = dictBufferSize;
 
-        OUTPUTLEVEL(2, "\r%70s\r", "");   /* blank line */
+        OUTPUTLEVEL(2, "\r%70s\r", ""); /* blank line */
         assert(srcSize < UINT_MAX);
-        OUTPUTLEVEL(2, "%2s-%-17.17s :%10u -> \r", marks[markNb], displayName, (unsigned)srcSize);
+        OUTPUTLEVEL(
+                2,
+                "%2s-%-17.17s :%10u -> \r",
+                marks[markNb],
+                displayName,
+                (unsigned)srcSize);
 
         while (!(compressionCompleted && decompressionCompleted)) {
             if (!compressionCompleted) {
-                BMK_runOutcome_t const cOutcome = BMK_benchTimedFn( timeStateCompress, cbp);
+                BMK_runOutcome_t const cOutcome =
+                        BMK_benchTimedFn(timeStateCompress, cbp);
 
                 if (!BMK_isSuccessful_runOutcome(cOutcome)) {
-                    return BMK_benchOutcome_error();
+                    RETURN_ERROR(30, BMK_benchOutcome_t, "compression error");
                 }
 
-                {   BMK_runTime_t const cResult = BMK_extract_runTime(cOutcome);
-                    cSize = cResult.sumOfReturn;
+                {
+                    BMK_runTime_t const cResult = BMK_extract_runTime(cOutcome);
+                    cSize                       = cResult.sumOfReturn;
                     ratio = (double)srcSize / (double)cSize;
-                    {   BMK_benchResult_t newResult;
-                        newResult.cSpeed = (U64)((double)srcSize * TIMELOOP_NANOSEC / cResult.nanoSecPerRun);
+                    {
+                        BMK_benchResult_t newResult;
+                        newResult.cSpeed =
+                                (U64)((double)srcSize * TIMELOOP_NANOSEC
+                                      / cResult.nanoSecPerRun);
                         benchResult.cSize = cSize;
                         if (newResult.cSpeed > benchResult.cSpeed)
                             benchResult.cSpeed = newResult.cSpeed;
-                }   }
+                    }
+                }
 
-                {   int const ratioAccuracy = (ratio < 10.) ? 3 : 2;
+                {
+                    int const ratioDigits = 1 + (ratio < 100.) + (ratio < 10.);
                     assert(cSize < UINT_MAX);
-                    OUTPUTLEVEL(2, "%2s-%-17.17s :%10u ->%10u (x%5.*f), %6.*f MB/s \r",
-                            marks[markNb], displayName,
-                            (unsigned)srcSize, (unsigned)cSize,
-                            ratioAccuracy, ratio,
-                            benchResult.cSpeed < (10 * MB_UNIT) ? 2 : 1, (double)benchResult.cSpeed / MB_UNIT);
+                    OUTPUTLEVEL(
+                            2,
+                            "%2s-%-17.17s :%10u ->%10u (x%5.*f), %6.*f MB/s \r",
+                            marks[markNb],
+                            displayName,
+                            (unsigned)srcSize,
+                            (unsigned)cSize,
+                            ratioDigits,
+                            ratio,
+                            benchResult.cSpeed < (10 * MB_UNIT) ? 2 : 1,
+                            (double)benchResult.cSpeed / MB_UNIT);
                 }
-                compressionCompleted = BMK_isCompleted_TimedFn(timeStateCompress);
+                compressionCompleted =
+                        BMK_isCompleted_TimedFn(timeStateCompress);
             }
 
-            if(!decompressionCompleted) {
-                BMK_runOutcome_t const dOutcome = BMK_benchTimedFn(timeStateDecompress, dbp);
+            if (!decompressionCompleted) {
+                BMK_runOutcome_t const dOutcome =
+                        BMK_benchTimedFn(timeStateDecompress, dbp);
 
-                if(!BMK_isSuccessful_runOutcome(dOutcome)) {
-                    return BMK_benchOutcome_error();
+                if (!BMK_isSuccessful_runOutcome(dOutcome)) {
+                    RETURN_ERROR(30, BMK_benchOutcome_t, "decompression error");
                 }
 
-                {   BMK_runTime_t const dResult = BMK_extract_runTime(dOutcome);
-                    U64 const newDSpeed = (U64)((double)srcSize * TIMELOOP_NANOSEC / dResult.nanoSecPerRun);
+                {
+                    BMK_runTime_t const dResult = BMK_extract_runTime(dOutcome);
+                    U64 const newDSpeed =
+                            (U64)((double)srcSize * TIMELOOP_NANOSEC
+                                  / dResult.nanoSecPerRun);
                     if (newDSpeed > benchResult.dSpeed)
                         benchResult.dSpeed = newDSpeed;
                 }
 
-                {   int const ratioAccuracy = (ratio < 10.) ? 3 : 2;
-                    OUTPUTLEVEL(2, "%2s-%-17.17s :%10u ->%10u (x%5.*f), %6.*f MB/s, %6.1f MB/s\r",
-                            marks[markNb], displayName,
-                            (unsigned)srcSize, (unsigned)cSize,
-                            ratioAccuracy, ratio,
-                            benchResult.cSpeed < (10 * MB_UNIT) ? 2 : 1, (double)benchResult.cSpeed / MB_UNIT,
+                {
+                    int const ratioDigits = 1 + (ratio < 100.) + (ratio < 10.);
+                    OUTPUTLEVEL(
+                            2,
+                            "%2s-%-17.17s :%10u ->%10u (x%5.*f), %6.*f MB/s, %6.1f MB/s\r",
+                            marks[markNb],
+                            displayName,
+                            (unsigned)srcSize,
+                            (unsigned)cSize,
+                            ratioDigits,
+                            ratio,
+                            benchResult.cSpeed < (10 * MB_UNIT) ? 2 : 1,
+                            (double)benchResult.cSpeed / MB_UNIT,
                             (double)benchResult.dSpeed / MB_UNIT);
                 }
-                decompressionCompleted = BMK_isCompleted_TimedFn(timeStateDecompress);
+                decompressionCompleted =
+                        BMK_isCompleted_TimedFn(timeStateDecompress);
             }
-            markNb = (markNb+1) % NB_MARKS;
-        }   /* while (!(compressionCompleted && decompressionCompleted)) */
+            markNb = (markNb + 1) % NB_MARKS;
+        } /* while (!(compressionCompleted && decompressionCompleted)) */
 
         /* CRC Checking */
-        {   const BYTE* resultBuffer = (const BYTE*)(*resultBufferPtr);
-            U64 const crcCheck = XXH64(resultBuffer, srcSize, 0);
-            if ((adv->mode == BMK_both) && (crcOrig!=crcCheck)) {
+        {
+            const BYTE* resultBuffer = (const BYTE*)(*resultBufferPtr);
+            U64 const crcCheck       = XXH64(resultBuffer, srcSize, 0);
+            if ((adv->mode == BMK_both) && (crcOrig != crcCheck)) {
                 size_t u;
                 DISPLAY("!!! WARNING !!! %14s : Invalid Checksum : %x != %x   \n",
-                        displayName, (unsigned)crcOrig, (unsigned)crcCheck);
-                for (u=0; u<srcSize; u++) {
+                        displayName,
+                        (unsigned)crcOrig,
+                        (unsigned)crcCheck);
+                for (u = 0; u < srcSize; u++) {
                     if (((const BYTE*)srcBuffer)[u] != resultBuffer[u]) {
                         unsigned segNb, bNb, pos;
                         size_t bacc = 0;
                         DISPLAY("Decoding error at pos %u ", (unsigned)u);
                         for (segNb = 0; segNb < nbBlocks; segNb++) {
-                            if (bacc + srcSizes[segNb] > u) break;
+                            if (bacc + srcSizes[segNb] > u)
+                                break;
                             bacc += srcSizes[segNb];
                         }
                         pos = (U32)(u - bacc);
                         bNb = pos / (128 KB);
-                        DISPLAY("(sample %u, block %u, pos %u) \n", segNb, bNb, pos);
-                        {   size_t const lowest = (u>5) ? 5 : u;
+                        DISPLAY("(sample %u, block %u, pos %u) \n",
+                                segNb,
+                                bNb,
+                                pos);
+                        {
+                            size_t const lowest = (u > 5) ? 5 : u;
                             size_t n;
                             DISPLAY("origin: ");
-                            for (n=lowest; n>0; n--)
-                                DISPLAY("%02X ", ((const BYTE*)srcBuffer)[u-n]);
+                            for (n = lowest; n > 0; n--)
+                                DISPLAY("%02X ",
+                                        ((const BYTE*)srcBuffer)[u - n]);
                             DISPLAY(" :%02X:  ", ((const BYTE*)srcBuffer)[u]);
-                            for (n=1; n<3; n++)
-                                DISPLAY("%02X ", ((const BYTE*)srcBuffer)[u+n]);
+                            for (n = 1; n < 3; n++)
+                                DISPLAY("%02X ",
+                                        ((const BYTE*)srcBuffer)[u + n]);
                             DISPLAY(" \n");
                             DISPLAY("decode: ");
-                            for (n=lowest; n>0; n--)
-                                DISPLAY("%02X ", resultBuffer[u-n]);
+                            for (n = lowest; n > 0; n--)
+                                DISPLAY("%02X ", resultBuffer[u - n]);
                             DISPLAY(" :%02X:  ", resultBuffer[u]);
-                            for (n=1; n<3; n++)
-                                DISPLAY("%02X ", resultBuffer[u+n]);
+                            for (n = 1; n < 3; n++)
+                                DISPLAY("%02X ", resultBuffer[u + n]);
                             DISPLAY(" \n");
                         }
                         break;
                     }
-                    if (u==srcSize-1) {  /* should never happen */
+                    if (u == srcSize - 1) { /* should never happen */
                         DISPLAY("no difference detected\n");
                     }
-                }   /* for (u=0; u<srcSize; u++) */
-            }   /* if ((adv->mode == BMK_both) && (crcOrig!=crcCheck)) */
-        }   /* CRC Checking */
+                } /* for (u=0; u<srcSize; u++) */
+            }     /* if ((adv->mode == BMK_both) && (crcOrig!=crcCheck)) */
+        }         /* CRC Checking */
 
-        if (displayLevel == 1) {   /* hidden display mode -q, used by python speed benchmark */
+        if (displayLevel
+            == 1) { /* hidden display mode -q, used by python speed benchmark */
             double const cSpeed = (double)benchResult.cSpeed / MB_UNIT;
             double const dSpeed = (double)benchResult.dSpeed / MB_UNIT;
             if (adv->additionalParam) {
-                OUTPUT("-%-3i%11i (%5.3f) %6.2f MB/s %6.1f MB/s  %s (param=%d)\n", cLevel, (int)cSize, ratio, cSpeed, dSpeed, displayName, adv->additionalParam);
+                OUTPUT("-%-3i%11i (%5.3f) %6.2f MB/s %6.1f MB/s  %s (param=%d)\n",
+                       cLevel,
+                       (int)cSize,
+                       ratio,
+                       cSpeed,
+                       dSpeed,
+                       displayName,
+                       adv->additionalParam);
             } else {
-                OUTPUT("-%-3i%11i (%5.3f) %6.2f MB/s %6.1f MB/s  %s\n", cLevel, (int)cSize, ratio, cSpeed, dSpeed, displayName);
+                OUTPUT("-%-3i%11i (%5.3f) %6.2f MB/s %6.1f MB/s  %s\n",
+                       cLevel,
+                       (int)cSize,
+                       ratio,
+                       cSpeed,
+                       dSpeed,
+                       displayName);
             }
         }
 
         OUTPUTLEVEL(2, "%2i#\n", cLevel);
-    }   /* Bench */
+    } /* Bench */
 
-    benchResult.cMem = (1ULL << (comprParams->windowLog)) + ZSTD_sizeof_CCtx(cctx);
+    benchResult.cMem =
+            (1ULL << (comprParams->windowLog)) + ZSTD_sizeof_CCtx(cctx);
     return BMK_benchOutcome_setValidResult(benchResult);
 }
 
-BMK_benchOutcome_t BMK_benchMemAdvanced(const void* srcBuffer, size_t srcSize,
-                        void* dstBuffer, size_t dstCapacity,
-                        const size_t* fileSizes, unsigned nbFiles,
-                        int cLevel, const ZSTD_compressionParameters* comprParams,
-                        const void* dictBuffer, size_t dictBufferSize,
-                        int displayLevel, const char* displayName, const BMK_advancedParams_t* adv)
+BMK_benchOutcome_t BMK_benchMemAdvanced(
+        const void* srcBuffer,
+        size_t srcSize,
+        void* dstBuffer,
+        size_t dstCapacity,
+        const size_t* fileSizes,
+        unsigned nbFiles,
+        int cLevel,
+        const ZSTD_compressionParameters* comprParams,
+        const void* dictBuffer,
+        size_t dictBufferSize,
+        int displayLevel,
+        const char* displayName,
+        const BMK_advancedParams_t* adv)
 
 {
-    int const dstParamsError = !dstBuffer ^ !dstCapacity;  /* must be both NULL or none */
+    int const dstParamsError =
+            !dstBuffer ^ !dstCapacity; /* must be both NULL or none */
 
-    size_t const blockSize = ((adv->blockSize>=32 && (adv->mode != BMK_decodeOnly)) ? adv->blockSize : srcSize) + (!srcSize) /* avoid div by 0 */ ;
-    U32 const maxNbBlocks = (U32) ((srcSize + (blockSize-1)) / blockSize) + nbFiles;
+    size_t const blockSize =
+            ((adv->blockSize >= 32 && (adv->mode != BMK_decodeOnly))
+                     ? adv->blockSize
+                     : srcSize)
+            + (!srcSize) /* avoid div by 0 */;
+    U32 const maxNbBlocks =
+            (U32)((srcSize + (blockSize - 1)) / blockSize) + nbFiles;
 
     /* these are the blockTable parameters, just split up */
-    const void ** const srcPtrs = (const void**)malloc(maxNbBlocks * sizeof(void*));
+    const void** const srcPtrs =
+            (const void**)malloc(maxNbBlocks * sizeof(void*));
     size_t* const srcSizes = (size_t*)malloc(maxNbBlocks * sizeof(size_t));
 
-
-    void ** const cPtrs = (void**)malloc(maxNbBlocks * sizeof(void*));
-    size_t* const cSizes = (size_t*)malloc(maxNbBlocks * sizeof(size_t));
+    void** const cPtrs        = (void**)malloc(maxNbBlocks * sizeof(void*));
+    size_t* const cSizes      = (size_t*)malloc(maxNbBlocks * sizeof(size_t));
     size_t* const cCapacities = (size_t*)malloc(maxNbBlocks * sizeof(size_t));
 
-    void ** const resPtrs = (void**)malloc(maxNbBlocks * sizeof(void*));
+    void** const resPtrs   = (void**)malloc(maxNbBlocks * sizeof(void*));
     size_t* const resSizes = (size_t*)malloc(maxNbBlocks * sizeof(size_t));
 
-    BMK_timedFnState_t* timeStateCompress = BMK_createTimedFnState(adv->nbSeconds * 1000, BMK_RUNTEST_DEFAULT_MS);
-    BMK_timedFnState_t* timeStateDecompress = BMK_createTimedFnState(adv->nbSeconds * 1000, BMK_RUNTEST_DEFAULT_MS);
+    BMK_timedFnState_t* timeStateCompress = BMK_createTimedFnState(
+            adv->nbSeconds * 1000, BMK_RUNTEST_DEFAULT_MS);
+    BMK_timedFnState_t* timeStateDecompress = BMK_createTimedFnState(
+            adv->nbSeconds * 1000, BMK_RUNTEST_DEFAULT_MS);
 
     ZSTD_CCtx* const cctx = ZSTD_createCCtx();
     ZSTD_DCtx* const dctx = ZSTD_createDCtx();
 
-    const size_t maxCompressedSize = dstCapacity ? dstCapacity : ZSTD_compressBound(srcSize) + (maxNbBlocks * 1024);
+    const size_t maxCompressedSize = dstCapacity
+            ? dstCapacity
+            : ZSTD_compressBound(srcSize) + (maxNbBlocks * 1024);
 
-    void* const internalDstBuffer = dstBuffer ? NULL : malloc(maxCompressedSize);
+    void* const internalDstBuffer =
+            dstBuffer ? NULL : malloc(maxCompressedSize);
     void* const compressedBuffer = dstBuffer ? dstBuffer : internalDstBuffer;
 
-    BMK_benchOutcome_t outcome = BMK_benchOutcome_error();  /* error by default */
+    BMK_benchOutcome_t outcome =
+            BMK_benchOutcome_error(); /* error by default */
 
     void* resultBuffer = srcSize ? malloc(srcSize) : NULL;
 
-    int allocationincomplete = !srcPtrs || !srcSizes || !cPtrs ||
-        !cSizes || !cCapacities || !resPtrs || !resSizes ||
-        !timeStateCompress || !timeStateDecompress ||
-        !cctx || !dctx ||
-        !compressedBuffer || !resultBuffer;
-
+    int const allocationincomplete = !srcPtrs || !srcSizes || !cPtrs || !cSizes
+            || !cCapacities || !resPtrs || !resSizes || !timeStateCompress
+            || !timeStateDecompress || !cctx || !dctx || !compressedBuffer
+            || !resultBuffer;
 
     if (!allocationincomplete && !dstParamsError) {
-        outcome = BMK_benchMemAdvancedNoAlloc(srcPtrs, srcSizes,
-                                            cPtrs, cCapacities, cSizes,
-                                            resPtrs, resSizes,
-                                            &resultBuffer,
-                                            compressedBuffer, maxCompressedSize,
-                                            timeStateCompress, timeStateDecompress,
-                                            srcBuffer, srcSize,
-                                            fileSizes, nbFiles,
-                                            cLevel, comprParams,
-                                            dictBuffer, dictBufferSize,
-                                            cctx, dctx,
-                                            displayLevel, displayName, adv);
+        outcome = BMK_benchMemAdvancedNoAlloc(
+                srcPtrs,
+                srcSizes,
+                cPtrs,
+                cCapacities,
+                cSizes,
+                resPtrs,
+                resSizes,
+                &resultBuffer,
+                compressedBuffer,
+                maxCompressedSize,
+                timeStateCompress,
+                timeStateDecompress,
+                srcBuffer,
+                srcSize,
+                fileSizes,
+                nbFiles,
+                cLevel,
+                comprParams,
+                dictBuffer,
+                dictBufferSize,
+                cctx,
+                dctx,
+                displayLevel,
+                displayName,
+                adv);
     }
 
     /* clean up */
@@ -634,104 +888,167 @@ BMK_benchOutcome_t BMK_benchMemAdvanced(const void* srcBuffer, size_t srcSize,
     free(resPtrs);
     free(resSizes);
 
-    if(allocationincomplete) {
-        RETURN_ERROR(31, BMK_benchOutcome_t, "allocation error : not enough memory");
+    if (allocationincomplete) {
+        RETURN_ERROR(
+                31, BMK_benchOutcome_t, "allocation error : not enough memory");
     }
 
-    if(dstParamsError) {
+    if (dstParamsError) {
         RETURN_ERROR(32, BMK_benchOutcome_t, "Dst parameters not coherent");
     }
     return outcome;
 }
 
-BMK_benchOutcome_t BMK_benchMem(const void* srcBuffer, size_t srcSize,
-                        const size_t* fileSizes, unsigned nbFiles,
-                        int cLevel, const ZSTD_compressionParameters* comprParams,
-                        const void* dictBuffer, size_t dictBufferSize,
-                        int displayLevel, const char* displayName) {
-
+BMK_benchOutcome_t BMK_benchMem(
+        const void* srcBuffer,
+        size_t srcSize,
+        const size_t* fileSizes,
+        unsigned nbFiles,
+        int cLevel,
+        const ZSTD_compressionParameters* comprParams,
+        const void* dictBuffer,
+        size_t dictBufferSize,
+        int displayLevel,
+        const char* displayName)
+{
     BMK_advancedParams_t const adv = BMK_initAdvancedParams();
-    return BMK_benchMemAdvanced(srcBuffer, srcSize,
-                                NULL, 0,
-                                fileSizes, nbFiles,
-                                cLevel, comprParams,
-                                dictBuffer, dictBufferSize,
-                                displayLevel, displayName, &adv);
+    return BMK_benchMemAdvanced(
+            srcBuffer,
+            srcSize,
+            NULL,
+            0,
+            fileSizes,
+            nbFiles,
+            cLevel,
+            comprParams,
+            dictBuffer,
+            dictBufferSize,
+            displayLevel,
+            displayName,
+            &adv);
 }
 
-static BMK_benchOutcome_t BMK_benchCLevel(const void* srcBuffer, size_t benchedSize,
-                            const size_t* fileSizes, unsigned nbFiles,
-                            int cLevel, const ZSTD_compressionParameters* comprParams,
-                            const void* dictBuffer, size_t dictBufferSize,
-                            int displayLevel, const char* displayName,
-                            BMK_advancedParams_t const * const adv)
+/* @return: 0 on success, !0 if error */
+static int BMK_benchCLevels(
+        const void* srcBuffer,
+        size_t benchedSize,
+        const size_t* fileSizes,
+        unsigned nbFiles,
+        int startCLevel, int endCLevel,
+        const ZSTD_compressionParameters* comprParams,
+        const void* dictBuffer,
+        size_t dictBufferSize,
+        int displayLevel,
+        const char* displayName,
+        BMK_advancedParams_t const* const adv)
 {
+    int level;
     const char* pch = strrchr(displayName, '\\'); /* Windows */
-    if (!pch) pch = strrchr(displayName, '/');    /* Linux */
-    if (pch) displayName = pch+1;
+    if (!pch)
+        pch = strrchr(displayName, '/'); /* Linux */
+    if (pch)
+        displayName = pch + 1;
+
+    if (endCLevel > ZSTD_maxCLevel()) {
+        DISPLAYLEVEL(1, "Invalid Compression Level \n");
+        return 15;
+    }
+    if (endCLevel < startCLevel) {
+        DISPLAYLEVEL(1, "Invalid Compression Level Range \n");
+        return 15;
+    }
 
     if (adv->realTime) {
         DISPLAYLEVEL(2, "Note : switching to real-time priority \n");
         SET_REALTIME_PRIORITY;
     }
 
-    if (displayLevel == 1 && !adv->additionalParam)   /* --quiet mode */
+    if (displayLevel == 1 && !adv->additionalParam) /* --quiet mode */
         OUTPUT("bench %s %s: input %u bytes, %u seconds, %u KB blocks\n",
-                ZSTD_VERSION_STRING, ZSTD_GIT_COMMIT_STRING,
-                (unsigned)benchedSize, adv->nbSeconds, (unsigned)(adv->blockSize>>10));
+               ZSTD_VERSION_STRING,
+               ZSTD_GIT_COMMIT_STRING,
+               (unsigned)benchedSize,
+               adv->nbSeconds,
+               (unsigned)(adv->blockSize >> 10));
 
-    return BMK_benchMemAdvanced(srcBuffer, benchedSize,
-                                NULL, 0,
-                                fileSizes, nbFiles,
-                                cLevel, comprParams,
-                                dictBuffer, dictBufferSize,
-                                displayLevel, displayName, adv);
+    for (level = startCLevel; level <= endCLevel; level++) {
+        BMK_benchOutcome_t res = BMK_benchMemAdvanced(
+            srcBuffer,
+            benchedSize,
+            NULL,
+            0,
+            fileSizes,
+            nbFiles,
+            level,
+            comprParams,
+            dictBuffer,
+            dictBufferSize,
+            displayLevel,
+            displayName,
+            adv);
+        if (!BMK_isSuccessful_benchOutcome(res)) return 1;
+    }
+    return 0;
 }
 
-BMK_benchOutcome_t BMK_syntheticTest(int cLevel, double compressibility,
-                          const ZSTD_compressionParameters* compressionParams,
-                          int displayLevel, const BMK_advancedParams_t* adv)
+int BMK_syntheticTest(
+        double compressibility,
+        int startingCLevel, int endCLevel,
+        const ZSTD_compressionParameters* compressionParams,
+        int displayLevel,
+        const BMK_advancedParams_t* adv)
 {
-    char name[20] = {0};
-    size_t const benchedSize = 10000000;
-    void* srcBuffer;
-    BMK_benchOutcome_t res;
-
-    if (cLevel > ZSTD_maxCLevel()) {
-        RETURN_ERROR(15, BMK_benchOutcome_t, "Invalid Compression Level");
-    }
+    char nameBuff[20]        = { 0 };
+    const char* name         = nameBuff;
+    size_t const benchedSize = adv->blockSize ? adv->blockSize : 10000000;
 
     /* Memory allocation */
-    srcBuffer = malloc(benchedSize);
-    if (!srcBuffer) RETURN_ERROR(21, BMK_benchOutcome_t, "not enough memory");
+    void* const srcBuffer = malloc(benchedSize);
+    if (!srcBuffer) {
+        DISPLAYLEVEL(1, "allocation error : not enough memory \n");
+        return 16;
+    }
 
     /* Fill input buffer */
-    RDG_genBuffer(srcBuffer, benchedSize, compressibility, 0.0, 0);
+    if (compressibility < 0.0) {
+        LOREM_genBuffer(srcBuffer, benchedSize, 0);
+        name = "Lorem ipsum";
+    } else {
+        RDG_genBuffer(srcBuffer, benchedSize, compressibility, 0.0, 0);
+        formatString_u(
+                nameBuff,
+                sizeof(nameBuff),
+                "Synthetic %u%%",
+                (unsigned)(compressibility * 100));
+    }
 
     /* Bench */
-    snprintf (name, sizeof(name), "Synthetic %2u%%", (unsigned)(compressibility*100));
-    res = BMK_benchCLevel(srcBuffer, benchedSize,
-                    &benchedSize /* ? */, 1 /* ? */,
-                    cLevel, compressionParams,
-                    NULL, 0,  /* dictionary */
-                    displayLevel, name, adv);
-
-    /* clean up */
-    free(srcBuffer);
-
-    return res;
+    {   int res = BMK_benchCLevels(
+                srcBuffer,
+                benchedSize,
+                &benchedSize,
+                1,
+                startingCLevel, endCLevel,
+                compressionParams,
+                NULL,
+                0, /* dictionary */
+                displayLevel,
+                name,
+                adv);
+        free(srcBuffer);
+        return res;
+    }
 }
 
-
-
 static size_t BMK_findMaxMem(U64 requiredMem)
 {
     size_t const step = 64 MB;
-    BYTE* testmem = NULL;
+    BYTE* testmem     = NULL;
 
     requiredMem = (((requiredMem >> 26) + 1) << 26);
     requiredMem += step;
-    if (requiredMem > maxMemory) requiredMem = maxMemory;
+    if (requiredMem > maxMemory)
+        requiredMem = maxMemory;
 
     do {
         testmem = (BYTE*)malloc((size_t)requiredMem);
@@ -745,133 +1062,197 @@ static size_t BMK_findMaxMem(U64 requiredMem)
 /*! BMK_loadFiles() :
  *  Loads `buffer` with content of files listed within `fileNamesTable`.
  *  At most, fills `buffer` entirely. */
-static int BMK_loadFiles(void* buffer, size_t bufferSize,
-                         size_t* fileSizes,
-                         const char* const * fileNamesTable, unsigned nbFiles,
-                         int displayLevel)
+static int BMK_loadFiles(
+        void* buffer,
+        size_t bufferSize,
+        size_t* fileSizes,
+        const char* const* fileNamesTable,
+        unsigned nbFiles,
+        int displayLevel)
 {
     size_t pos = 0, totalSize = 0;
     unsigned n;
-    for (n=0; n<nbFiles; n++) {
-        U64 fileSize = UTIL_getFileSize(fileNamesTable[n]);  /* last file may be shortened */
-        if (UTIL_isDirectory(fileNamesTable[n])) {
-            DISPLAYLEVEL(2, "Ignoring %s directory...       \n", fileNamesTable[n]);
+    for (n = 0; n < nbFiles; n++) {
+        const char* const filename = fileNamesTable[n];
+        U64 fileSize = UTIL_getFileSize(
+                filename); /* last file may be shortened */
+        if (UTIL_isDirectory(filename)) {
+            DISPLAYLEVEL(
+                    2, "Ignoring %s directory...       \n", filename);
             fileSizes[n] = 0;
             continue;
         }
         if (fileSize == UTIL_FILESIZE_UNKNOWN) {
-            DISPLAYLEVEL(2, "Cannot evaluate size of %s, ignoring ... \n", fileNamesTable[n]);
+            DISPLAYLEVEL(
+                    2,
+                    "Cannot evaluate size of %s, ignoring ... \n",
+                    filename);
             fileSizes[n] = 0;
             continue;
         }
-        {   FILE* const f = fopen(fileNamesTable[n], "rb");
-            if (f==NULL) RETURN_ERROR_INT(10, "impossible to open file %s", fileNamesTable[n]);
-            OUTPUTLEVEL(2, "Loading %s...       \r", fileNamesTable[n]);
-            if (fileSize > bufferSize-pos) fileSize = bufferSize-pos, nbFiles=n;   /* buffer too small - stop after this file */
-            {   size_t const readSize = fread(((char*)buffer)+pos, 1, (size_t)fileSize, f);
-                if (readSize != (size_t)fileSize) RETURN_ERROR_INT(11, "could not read %s", fileNamesTable[n]);
+        if (fileSize > bufferSize - pos) {
+            /* buffer too small - limit quantity loaded */
+            fileSize = bufferSize - pos;
+            nbFiles  = n; /* stop after this file */
+        }
+
+        {   FILE* const f = fopen(filename, "rb");
+            if (f == NULL) {
+                RETURN_ERROR_INT(
+                        10, "cannot open file %s", filename);
+            }
+            OUTPUTLEVEL(2, "Loading %s...       \r", filename);
+            {   size_t const readSize =
+                        fread(((char*)buffer) + pos, 1, (size_t)fileSize, f);
+                if (readSize != (size_t)fileSize) {
+                    fclose(f);
+                    RETURN_ERROR_INT(
+                            11, "invalid read %s", filename);
+                }
                 pos += readSize;
             }
             fileSizes[n] = (size_t)fileSize;
             totalSize += (size_t)fileSize;
             fclose(f);
-    }   }
+        }
+    }
 
-    if (totalSize == 0) RETURN_ERROR_INT(12, "no data to bench");
+    if (totalSize == 0)
+        RETURN_ERROR_INT(12, "no data to bench");
     return 0;
 }
 
-BMK_benchOutcome_t BMK_benchFilesAdvanced(
-                        const char* const * fileNamesTable, unsigned nbFiles,
-                        const char* dictFileName, int cLevel,
-                        const ZSTD_compressionParameters* compressionParams,
-                        int displayLevel, const BMK_advancedParams_t* adv)
+int BMK_benchFilesAdvanced(
+        const char* const* fileNamesTable,
+        unsigned nbFiles,
+        const char* dictFileName,
+        int startCLevel, int endCLevel,
+        const ZSTD_compressionParameters* compressionParams,
+        int displayLevel,
+        const BMK_advancedParams_t* adv)
 {
     void* srcBuffer = NULL;
     size_t benchedSize;
-    void* dictBuffer = NULL;
+    void* dictBuffer      = NULL;
     size_t dictBufferSize = 0;
-    size_t* fileSizes = NULL;
-    BMK_benchOutcome_t res;
+    size_t* fileSizes     = NULL;
+    int res = 1;
     U64 const totalSizeToLoad = UTIL_getTotalFileSize(fileNamesTable, nbFiles);
 
     if (!nbFiles) {
-        RETURN_ERROR(14, BMK_benchOutcome_t, "No Files to Benchmark");
+        DISPLAYLEVEL(1, "No Files to Benchmark");
+        return 13;
     }
 
-    if (cLevel > ZSTD_maxCLevel()) {
-        RETURN_ERROR(15, BMK_benchOutcome_t, "Invalid Compression Level");
+    if (endCLevel > ZSTD_maxCLevel()) {
+        DISPLAYLEVEL(1, "Invalid Compression Level");
+        return 14;
     }
 
     if (totalSizeToLoad == UTIL_FILESIZE_UNKNOWN) {
-        RETURN_ERROR(9, BMK_benchOutcome_t, "Error loading files");
+        DISPLAYLEVEL(1, "Error loading files");
+        return 15;
     }
 
     fileSizes = (size_t*)calloc(nbFiles, sizeof(size_t));
-    if (!fileSizes) RETURN_ERROR(12, BMK_benchOutcome_t, "not enough memory for fileSizes");
+    if (!fileSizes) {
+        DISPLAYLEVEL(1, "not enough memory for fileSizes");
+        return 16;
+    }
 
     /* Load dictionary */
     if (dictFileName != NULL) {
         U64 const dictFileSize = UTIL_getFileSize(dictFileName);
         if (dictFileSize == UTIL_FILESIZE_UNKNOWN) {
-            DISPLAYLEVEL(1, "error loading %s : %s \n", dictFileName, strerror(errno));
+            DISPLAYLEVEL(
+                    1,
+                    "error loading %s : %s \n",
+                    dictFileName,
+                    strerror(errno));
             free(fileSizes);
-            RETURN_ERROR(9, BMK_benchOutcome_t, "benchmark aborted");
+            DISPLAYLEVEL(1, "benchmark aborted");
+            return 17;
         }
         if (dictFileSize > 64 MB) {
             free(fileSizes);
-            RETURN_ERROR(10, BMK_benchOutcome_t, "dictionary file %s too large", dictFileName);
+            DISPLAYLEVEL(1, "dictionary file %s too large", dictFileName);
+            return 18;
         }
         dictBufferSize = (size_t)dictFileSize;
-        dictBuffer = malloc(dictBufferSize);
-        if (dictBuffer==NULL) {
+        dictBuffer     = malloc(dictBufferSize);
+        if (dictBuffer == NULL) {
             free(fileSizes);
-            RETURN_ERROR(11, BMK_benchOutcome_t, "not enough memory for dictionary (%u bytes)",
-                            (unsigned)dictBufferSize);
+            DISPLAYLEVEL(
+                    1,
+                    "not enough memory for dictionary (%u bytes)",
+                    (unsigned)dictBufferSize);
+            return 19;
         }
 
-        {   int const errorCode = BMK_loadFiles(dictBuffer, dictBufferSize,
-                                                fileSizes, &dictFileName /*?*/,
-                                                1 /*?*/, displayLevel);
+        {
+            int const errorCode = BMK_loadFiles(
+                    dictBuffer,
+                    dictBufferSize,
+                    fileSizes,
+                    &dictFileName /*?*/,
+                    1 /*?*/,
+                    displayLevel);
             if (errorCode) {
-                res = BMK_benchOutcome_error();
                 goto _cleanUp;
-        }   }
+            }
+        }
     }
 
     /* Memory allocation & restrictions */
     benchedSize = BMK_findMaxMem(totalSizeToLoad * 3) / 3;
-    if ((U64)benchedSize > totalSizeToLoad) benchedSize = (size_t)totalSizeToLoad;
+    if ((U64)benchedSize > totalSizeToLoad)
+        benchedSize = (size_t)totalSizeToLoad;
     if (benchedSize < totalSizeToLoad)
-        DISPLAY("Not enough memory; testing %u MB only...\n", (unsigned)(benchedSize >> 20));
+        DISPLAY("Not enough memory; testing %u MB only...\n",
+                (unsigned)(benchedSize >> 20));
 
     srcBuffer = benchedSize ? malloc(benchedSize) : NULL;
     if (!srcBuffer) {
         free(dictBuffer);
         free(fileSizes);
-        RETURN_ERROR(12, BMK_benchOutcome_t, "not enough memory");
+        DISPLAYLEVEL(1, "not enough memory for srcBuffer");
+        return 20;
     }
 
     /* Load input buffer */
-    {   int const errorCode = BMK_loadFiles(srcBuffer, benchedSize,
-                                        fileSizes, fileNamesTable, nbFiles,
-                                        displayLevel);
+    {
+        int const errorCode = BMK_loadFiles(
+                srcBuffer,
+                benchedSize,
+                fileSizes,
+                fileNamesTable,
+                nbFiles,
+                displayLevel);
         if (errorCode) {
-            res = BMK_benchOutcome_error();
             goto _cleanUp;
-    }   }
+        }
+    }
 
     /* Bench */
-    {   char mfName[20] = {0};
-        snprintf (mfName, sizeof(mfName), " %u files", nbFiles);
-        {   const char* const displayName = (nbFiles > 1) ? mfName : fileNamesTable[0];
-            res = BMK_benchCLevel(srcBuffer, benchedSize,
-                                fileSizes, nbFiles,
-                                cLevel, compressionParams,
-                                dictBuffer, dictBufferSize,
-                                displayLevel, displayName,
-                                adv);
-    }   }
+    {
+        char mfName[20] = { 0 };
+        formatString_u(mfName, sizeof(mfName), " %u files", nbFiles);
+        {   const char* const displayName =
+                    (nbFiles > 1) ? mfName : fileNamesTable[0];
+            res = BMK_benchCLevels(
+                    srcBuffer,
+                    benchedSize,
+                    fileSizes,
+                    nbFiles,
+                    startCLevel, endCLevel,
+                    compressionParams,
+                    dictBuffer,
+                    dictBufferSize,
+                    displayLevel,
+                    displayName,
+                    adv);
+        }
+    }
 
 _cleanUp:
     free(srcBuffer);
@@ -880,13 +1261,21 @@ _cleanUp:
     return res;
 }
 
-
-BMK_benchOutcome_t BMK_benchFiles(
-                    const char* const * fileNamesTable, unsigned nbFiles,
-                    const char* dictFileName,
-                    int cLevel, const ZSTD_compressionParameters* compressionParams,
-                    int displayLevel)
+int BMK_benchFiles(
+        const char* const* fileNamesTable,
+        unsigned nbFiles,
+        const char* dictFileName,
+        int cLevel,
+        const ZSTD_compressionParameters* compressionParams,
+        int displayLevel)
 {
     BMK_advancedParams_t const adv = BMK_initAdvancedParams();
-    return BMK_benchFilesAdvanced(fileNamesTable, nbFiles, dictFileName, cLevel, compressionParams, displayLevel, &adv);
+    return BMK_benchFilesAdvanced(
+            fileNamesTable,
+            nbFiles,
+            dictFileName,
+            cLevel, cLevel,
+            compressionParams,
+            displayLevel,
+            &adv);
 }
diff --git a/sys/contrib/zstd/programs/benchzstd.h b/sys/contrib/zstd/programs/benchzstd.h
index 11ac85da7f94..4fd0e5a8af43 100644
--- a/sys/contrib/zstd/programs/benchzstd.h
+++ b/sys/contrib/zstd/programs/benchzstd.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -14,10 +14,6 @@
   * and display progress result and final summary
   */
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
 #ifndef BENCH_ZSTD_H_3242387
 #define BENCH_ZSTD_H_3242387
 
@@ -26,7 +22,6 @@ extern "C" {
 #define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_compressionParameters */
 #include "../lib/zstd.h"     /* ZSTD_compressionParameters */
 
-
 /* ===  Constants  === */
 
 #define MB_UNIT 1000000
@@ -81,21 +76,13 @@ BMK_benchResult_t BMK_extract_benchResult(BMK_benchOutcome_t outcome);
  *      2 : + result + interaction + warnings;
  *      3 : + information;
  *      4 : + debug
- * @return:
- *      a variant, which expresses either an error, or a valid result.
- *      Use BMK_isSuccessful_benchOutcome() to check if function was successful.
- *      If yes, extract the valid result with BMK_extract_benchResult(),
- *      it will contain :
- *          .cSpeed: compression speed in bytes per second,
- *          .dSpeed: decompression speed in bytes per second,
- *          .cSize : compressed size, in bytes
- *          .cMem  : memory budget required for the compression context
+ * @return: 0 on success, !0 on error
  */
-BMK_benchOutcome_t BMK_benchFiles(
-                   const char* const * fileNamesTable, unsigned nbFiles,
-                   const char* dictFileName,
-                   int cLevel, const ZSTD_compressionParameters* compressionParams,
-                   int displayLevel);
+int BMK_benchFiles(
+            const char* const * fileNamesTable, unsigned nbFiles,
+            const char* dictFileName,
+            int cLevel, const ZSTD_compressionParameters* compressionParams,
+            int displayLevel);
 
 
 typedef enum {
@@ -108,6 +95,7 @@ typedef struct {
     BMK_mode_t mode;        /* 0: all, 1: compress only 2: decode only */
     unsigned nbSeconds;     /* default timing is in nbSeconds */
     size_t blockSize;       /* Maximum size of each block*/
+    size_t targetCBlockSize;/* Approximative size of compressed blocks */
     int nbWorkers;          /* multithreading */
     unsigned realTime;      /* real time priority */
     int additionalParam;    /* used by python speed benchmark */
@@ -116,7 +104,7 @@ typedef struct {
     int ldmHashLog;
     int ldmBucketSizeLog;
     int ldmHashRateLog;
-    ZSTD_paramSwitch_e literalCompressionMode;
+    ZSTD_ParamSwitch_e literalCompressionMode;
     int useRowMatchFinder;  /* use row-based matchfinder if possible */
 } BMK_advancedParams_t;
 
@@ -126,33 +114,27 @@ BMK_advancedParams_t BMK_initAdvancedParams(void);
 /*! BMK_benchFilesAdvanced():
  *  Same as BMK_benchFiles(),
  *  with more controls, provided through advancedParams_t structure */
-BMK_benchOutcome_t BMK_benchFilesAdvanced(
-                   const char* const * fileNamesTable, unsigned nbFiles,
-                   const char* dictFileName,
-                   int cLevel, const ZSTD_compressionParameters* compressionParams,
-                   int displayLevel, const BMK_advancedParams_t* adv);
+int BMK_benchFilesAdvanced(
+               const char* const * fileNamesTable, unsigned nbFiles,
+               const char* dictFileName,
+               int startCLevel, int endCLevel,
+               const ZSTD_compressionParameters* compressionParams,
+               int displayLevel, const BMK_advancedParams_t* adv);
 
 /*! BMK_syntheticTest() -- called from zstdcli */
-/*  Generates a sample with datagen, using compressibility argument */
-/*  cLevel - compression level to benchmark, errors if invalid
- *  compressibility - determines compressibility of sample
- *  compressionParams - basic compression Parameters
- *  displayLevel - see benchFiles
- *  adv - see advanced_Params_t
- * @return:
- *      a variant, which expresses either an error, or a valid result.
- *      Use BMK_isSuccessful_benchOutcome() to check if function was successful.
- *      If yes, extract the valid result with BMK_extract_benchResult(),
- *      it will contain :
- *          .cSpeed: compression speed in bytes per second,
- *          .dSpeed: decompression speed in bytes per second,
- *          .cSize : compressed size, in bytes
- *          .cMem  : memory budget required for the compression context
+/*  Generates a sample with datagen, using @compressibility argument
+ * @cLevel - compression level to benchmark, errors if invalid
+ * @compressibility - determines compressibility of sample, range [0.0 - 1.0]
+ *        if @compressibility < 0.0, uses the lorem ipsum generator
+ * @compressionParams - basic compression Parameters
+ * @displayLevel - see benchFiles
+ * @adv - see advanced_Params_t
+ * @return: 0 on success, !0 on error
  */
-BMK_benchOutcome_t BMK_syntheticTest(
-                          int cLevel, double compressibility,
-                          const ZSTD_compressionParameters* compressionParams,
-                          int displayLevel, const BMK_advancedParams_t* adv);
+int BMK_syntheticTest(double compressibility,
+                      int startingCLevel, int endCLevel,
+                      const ZSTD_compressionParameters* compressionParams,
+                      int displayLevel, const BMK_advancedParams_t* adv);
 
 
 
@@ -190,8 +172,8 @@ BMK_benchOutcome_t BMK_benchMem(const void* srcBuffer, size_t srcSize,
                         int displayLevel, const char* displayName);
 
 
-/* BMK_benchMemAdvanced() : same as BMK_benchMem()
- * with following additional options :
+/* BMK_benchMemAdvanced() : used by Paramgrill
+ * same as BMK_benchMem() with following additional options :
  * dstBuffer - destination buffer to write compressed output in, NULL if none provided.
  * dstCapacity - capacity of destination buffer, give 0 if dstBuffer = NULL
  * adv = see advancedParams_t
@@ -207,7 +189,3 @@ BMK_benchOutcome_t BMK_benchMemAdvanced(const void* srcBuffer, size_t srcSize,
 
 
 #endif   /* BENCH_ZSTD_H_3242387 */
-
-#if defined (__cplusplus)
-}
-#endif
diff --git a/sys/contrib/zstd/programs/datagen.c b/sys/contrib/zstd/programs/datagen.c
index 3b4f9e5c7b69..ddc690bb1b75 100644
--- a/sys/contrib/zstd/programs/datagen.c
+++ b/sys/contrib/zstd/programs/datagen.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/sys/contrib/zstd/programs/datagen.h b/sys/contrib/zstd/programs/datagen.h
index b76ae2a2225c..461fb716c436 100644
--- a/sys/contrib/zstd/programs/datagen.h
+++ b/sys/contrib/zstd/programs/datagen.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -14,6 +14,10 @@
 
 #include <stddef.h>   /* size_t */
 
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
 void RDG_genStdout(unsigned long long size, double matchProba, double litProba, unsigned seed);
 void RDG_genBuffer(void* buffer, size_t size, double matchProba, double litProba, unsigned seed);
 /*!RDG_genBuffer
@@ -27,4 +31,8 @@ void RDG_genBuffer(void* buffer, size_t size, double matchProba, double litProba
    Same as RDG_genBuffer, but generates data into stdout
 */
 
+#if defined (__cplusplus)
+} /* extern "C" */
+#endif
+
 #endif
diff --git a/sys/contrib/zstd/programs/dibio.c b/sys/contrib/zstd/programs/dibio.c
index d19f954486f0..7ba22d15b97f 100644
--- a/sys/contrib/zstd/programs/dibio.c
+++ b/sys/contrib/zstd/programs/dibio.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -27,10 +27,11 @@
 #include <string.h>         /* memset */
 #include <stdio.h>          /* fprintf, fopen, ftello64 */
 #include <errno.h>          /* errno */
-#include <assert.h>
 
 #include "timefn.h"         /* UTIL_time_t, UTIL_clockSpanMicro, UTIL_getTime */
+#include "../lib/common/debug.h" /* assert */
 #include "../lib/common/mem.h"  /* read */
+#include "../lib/zstd_errors.h"
 #include "dibio.h"
 
 
@@ -127,8 +128,11 @@ static int DiB_loadFiles(
     while ( nbSamplesLoaded < sstSize && fileIndex < nbFiles ) {
         size_t fileDataLoaded;
         S64 const fileSize = DiB_getFileSize(fileNamesTable[fileIndex]);
-        if (fileSize <= 0) /* skip if zero-size or file error */
+        if (fileSize <= 0) {
+            /* skip if zero-size or file error */
+            ++fileIndex;
             continue;
+        }
 
         f = fopen( fileNamesTable[fileIndex], "rb");
         if (f == NULL)
@@ -193,7 +197,8 @@ static U32 DiB_rand(U32* src)
 static void DiB_shuffle(const char** fileNamesTable, unsigned nbFiles) {
     U32 seed = 0xFD2FB528;
     unsigned i;
-    assert(nbFiles >= 1);
+    if (nbFiles == 0)
+        return;
     for (i = nbFiles - 1; i > 0; --i) {
         unsigned const j = DiB_rand(&seed) % (i + 1);
         const char* const tmp = fileNamesTable[j];
@@ -269,21 +274,20 @@ static fileStats DiB_fileStats(const char** fileNamesTable, int nbFiles, size_t
     int n;
     memset(&fs, 0, sizeof(fs));
 
-    // We assume that if chunking is requested, the chunk size is < SAMPLESIZE_MAX
+    /* We assume that if chunking is requested, the chunk size is < SAMPLESIZE_MAX */
     assert( chunkSize <= SAMPLESIZE_MAX );
 
     for (n=0; n<nbFiles; n++) {
       S64 const fileSize = DiB_getFileSize(fileNamesTable[n]);
-      // TODO: is there a minimum sample size? What if the file is 1-byte?
+      /* TODO: is there a minimum sample size? What if the file is 1-byte? */
       if (fileSize == 0) {
         DISPLAYLEVEL(3, "Sample file '%s' has zero size, skipping...\n", fileNamesTable[n]);
         continue;
       }
 
       /* the case where we are breaking up files in sample chunks */
-      if (chunkSize > 0)
-      {
-        // TODO: is there a minimum sample size? Can we have a 1-byte sample?
+      if (chunkSize > 0) {
+        /* TODO: is there a minimum sample size? Can we have a 1-byte sample? */
         fs.nbSamples += (int)((fileSize + chunkSize-1) / chunkSize);
         fs.totalSizeToLoad += fileSize;
       }
@@ -294,7 +298,7 @@ static fileStats DiB_fileStats(const char** fileNamesTable, int nbFiles, size_t
           fs.oneSampleTooLarge |= (fileSize > 2*SAMPLESIZE_MAX);
 
           /* Limit to the first SAMPLESIZE_MAX (128kB) of the file */
-          DISPLAYLEVEL(3, "Sample file '%s' is too large, limiting to %d KB",
+          DISPLAYLEVEL(3, "Sample file '%s' is too large, limiting to %d KB\n",
               fileNamesTable[n], SAMPLESIZE_MAX / (1 KB));
         }
         fs.nbSamples += 1;
@@ -350,7 +354,7 @@ int DiB_trainFromFiles(const char* dictFileName, size_t maxDictSize,
     }
 
     /* Checks */
-    if ((!sampleSizes) || (!srcBuffer) || (!dictBuffer))
+    if ((fs.nbSamples && !sampleSizes) || (!srcBuffer) || (!dictBuffer))
         EXM_THROW(12, "not enough memory for DiB_trainFiles");   /* should not happen */
     if (fs.oneSampleTooLarge) {
         DISPLAYLEVEL(2, "!  Warning : some sample(s) are very large \n");
@@ -379,7 +383,7 @@ int DiB_trainFromFiles(const char* dictFileName, size_t maxDictSize,
         srcBuffer, &loadedSize, sampleSizes, fs.nbSamples, fileNamesTable,
         nbFiles, chunkSize, displayLevel);
 
-    {   size_t dictSize;
+    {   size_t dictSize = ZSTD_error_GENERIC;
         if (params) {
             DiB_fillNoise((char*)srcBuffer + loadedSize, NOISELENGTH);   /* guard band, for end of buffer condition */
             dictSize = ZDICT_trainFromBuffer_legacy(dictBuffer, maxDictSize,
@@ -399,8 +403,7 @@ int DiB_trainFromFiles(const char* dictFileName, size_t maxDictSize,
               dictSize = ZDICT_trainFromBuffer_cover(dictBuffer, maxDictSize, srcBuffer,
                                                      sampleSizes, nbSamplesLoaded, *coverParams);
             }
-        } else {
-            assert(fastCoverParams != NULL);
+        } else if (fastCoverParams != NULL) {
             if (optimize) {
               dictSize = ZDICT_optimizeTrainFromBuffer_fastCover(dictBuffer, maxDictSize,
                                                               srcBuffer, sampleSizes, nbSamplesLoaded,
@@ -415,6 +418,8 @@ int DiB_trainFromFiles(const char* dictFileName, size_t maxDictSize,
               dictSize = ZDICT_trainFromBuffer_fastCover(dictBuffer, maxDictSize, srcBuffer,
                                                         sampleSizes, nbSamplesLoaded, *fastCoverParams);
             }
+        } else {
+            assert(0 /* Impossible */);
         }
         if (ZDICT_isError(dictSize)) {
             DISPLAYLEVEL(1, "dictionary training failed : %s \n", ZDICT_getErrorName(dictSize));   /* should not happen */
diff --git a/sys/contrib/zstd/programs/dibio.h b/sys/contrib/zstd/programs/dibio.h
index 666c1e661800..a96104c36d70 100644
--- a/sys/contrib/zstd/programs/dibio.h
+++ b/sys/contrib/zstd/programs/dibio.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/sys/contrib/zstd/programs/fileio.c b/sys/contrib/zstd/programs/fileio.c
index 5338fa62955b..0ecca40d2abc 100644
--- a/sys/contrib/zstd/programs/fileio.c
+++ b/sys/contrib/zstd/programs/fileio.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -28,6 +28,7 @@
 #include <stdio.h>      /* fprintf, open, fdopen, fread, _fileno, stdin, stdout */
 #include <stdlib.h>     /* malloc, free */
 #include <string.h>     /* strcmp, strlen */
+#include <time.h>       /* clock_t, to measure process time */
 #include <fcntl.h>      /* O_WRONLY */
 #include <assert.h>
 #include <errno.h>      /* errno */
@@ -40,8 +41,12 @@
 #  include <io.h>
 #endif
 
-#include "../lib/common/mem.h"     /* U32, U64 */
 #include "fileio.h"
+#include "fileio_asyncio.h"
+#include "fileio_common.h"
+
+FIO_display_prefs_t g_display_prefs = {2, FIO_ps_auto};
+UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
 
 #define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_magicNumber, ZSTD_frameHeaderSize_max */
 #include "../lib/zstd.h"
@@ -65,6 +70,40 @@
 #  include <lz4.h>
 #endif
 
+char const* FIO_zlibVersion(void)
+{
+#if defined(ZSTD_GZCOMPRESS) || defined(ZSTD_GZDECOMPRESS)
+    return zlibVersion();
+#else
+    return "Unsupported";
+#endif
+}
+
+char const* FIO_lz4Version(void)
+{
+#if defined(ZSTD_LZ4COMPRESS) || defined(ZSTD_LZ4DECOMPRESS)
+    /* LZ4_versionString() added in v1.7.3 */
+#   if LZ4_VERSION_NUMBER >= 10703
+        return LZ4_versionString();
+#   else
+#       define ZSTD_LZ4_VERSION LZ4_VERSION_MAJOR.LZ4_VERSION_MINOR.LZ4_VERSION_RELEASE
+#       define ZSTD_LZ4_VERSION_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_LZ4_VERSION)
+        return ZSTD_LZ4_VERSION_STRING;
+#   endif
+#else
+    return "Unsupported";
+#endif
+}
+
+char const* FIO_lzmaVersion(void)
+{
+#if defined(ZSTD_LZMACOMPRESS) || defined(ZSTD_LZMADECOMPRESS)
+    return lzma_version_string();
+#else
+    return "Unsupported";
+#endif
+}
+
 
 /*-*************************************
 *  Constants
@@ -75,69 +114,17 @@
 #define FNSPACE 30
 
 /* Default file permissions 0666 (modulated by umask) */
+/* Temporary restricted file permissions are used when we're going to
+ * chmod/chown at the end of the operation. */
 #if !defined(_WIN32)
 /* These macros aren't defined on windows. */
 #define DEFAULT_FILE_PERMISSIONS (S_IRUSR|S_IWUSR|S_IRGRP|S_IWGRP|S_IROTH|S_IWOTH)
+#define TEMPORARY_FILE_PERMISSIONS (S_IRUSR|S_IWUSR)
 #else
 #define DEFAULT_FILE_PERMISSIONS (0666)
+#define TEMPORARY_FILE_PERMISSIONS (0600)
 #endif
 
-/*-*************************************
-*  Macros
-***************************************/
-#define KB *(1 <<10)
-#define MB *(1 <<20)
-#define GB *(1U<<30)
-#undef MAX
-#define MAX(a,b) ((a)>(b) ? (a) : (b))
-
-struct FIO_display_prefs_s {
-    int displayLevel;   /* 0 : no display;  1: errors;  2: + result + interaction + warnings;  3: + progression;  4: + information */
-    FIO_progressSetting_e progressSetting;
-};
-
-static FIO_display_prefs_t g_display_prefs = {2, FIO_ps_auto};
-
-#define DISPLAY(...)         fprintf(stderr, __VA_ARGS__)
-#define DISPLAYOUT(...)      fprintf(stdout, __VA_ARGS__)
-#define DISPLAYLEVEL(l, ...) { if (g_display_prefs.displayLevel>=l) { DISPLAY(__VA_ARGS__); } }
-
-static const U64 g_refreshRate = SEC_TO_MICRO / 6;
-static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
-
-#define READY_FOR_UPDATE() ((g_display_prefs.progressSetting != FIO_ps_never) && UTIL_clockSpanMicro(g_displayClock) > g_refreshRate)
-#define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); }
-#define DISPLAYUPDATE(l, ...) {                              \
-        if (g_display_prefs.displayLevel>=l && (g_display_prefs.progressSetting != FIO_ps_never)) { \
-            if (READY_FOR_UPDATE() || (g_display_prefs.displayLevel>=4)) { \
-                DELAY_NEXT_UPDATE();                         \
-                DISPLAY(__VA_ARGS__);                        \
-                if (g_display_prefs.displayLevel>=4) fflush(stderr);       \
-    }   }   }
-
-#undef MIN  /* in case it would be already defined */
-#define MIN(a,b)    ((a) < (b) ? (a) : (b))
-
-
-#define EXM_THROW(error, ...)                                             \
-{                                                                         \
-    DISPLAYLEVEL(1, "zstd: ");                                            \
-    DISPLAYLEVEL(5, "Error defined at %s, line %i : \n", __FILE__, __LINE__); \
-    DISPLAYLEVEL(1, "error %i : ", error);                                \
-    DISPLAYLEVEL(1, __VA_ARGS__);                                         \
-    DISPLAYLEVEL(1, " \n");                                               \
-    exit(error);                                                          \
-}
-
-#define CHECK_V(v, f)                                \
-    v = f;                                           \
-    if (ZSTD_isError(v)) {                           \
-        DISPLAYLEVEL(5, "%s \n", #f);                \
-        EXM_THROW(11, "%s", ZSTD_getErrorName(v));   \
-    }
-#define CHECK(f) { size_t err; CHECK_V(err, f); }
-
-
 /*-************************************
 *  Signal (Ctrl-C trapping)
 **************************************/
@@ -237,7 +224,7 @@ static void ABRThandler(int sig) {
 }
 #endif
 
-void FIO_addAbortHandler()
+void FIO_addAbortHandler(void)
 {
 #if BACKTRACE_ENABLE
     signal(SIGABRT, ABRThandler);
@@ -248,94 +235,6 @@ void FIO_addAbortHandler()
 #endif
 }
 
-
-/*-************************************************************
-* Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW
-***************************************************************/
-#if defined(_MSC_VER) && _MSC_VER >= 1400
-#   define LONG_SEEK _fseeki64
-#   define LONG_TELL _ftelli64
-#elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */
-#  define LONG_SEEK fseeko
-#  define LONG_TELL ftello
-#elif defined(__MINGW32__) && !defined(__STRICT_ANSI__) && !defined(__NO_MINGW_LFS) && defined(__MSVCRT__)
-#   define LONG_SEEK fseeko64
-#   define LONG_TELL ftello64
-#elif defined(_WIN32) && !defined(__DJGPP__)
-#   include <windows.h>
-    static int LONG_SEEK(FILE* file, __int64 offset, int origin) {
-        LARGE_INTEGER off;
-        DWORD method;
-        off.QuadPart = offset;
-        if (origin == SEEK_END)
-            method = FILE_END;
-        else if (origin == SEEK_CUR)
-            method = FILE_CURRENT;
-        else
-            method = FILE_BEGIN;
-
-        if (SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, NULL, method))
-            return 0;
-        else
-            return -1;
-    }
-    static __int64 LONG_TELL(FILE* file) {
-        LARGE_INTEGER off, newOff;
-        off.QuadPart = 0;
-        newOff.QuadPart = 0;
-        SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, &newOff, FILE_CURRENT);
-        return newOff.QuadPart;
-    }
-#else
-#   define LONG_SEEK fseek
-#   define LONG_TELL ftell
-#endif
-
-
-/*-*************************************
-*  Parameters: FIO_prefs_t
-***************************************/
-
-/* typedef'd to FIO_prefs_t within fileio.h */
-struct FIO_prefs_s {
-
-    /* Algorithm preferences */
-    FIO_compressionType_t compressionType;
-    U32 sparseFileSupport;   /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */
-    int dictIDFlag;
-    int checksumFlag;
-    int blockSize;
-    int overlapLog;
-    U32 adaptiveMode;
-    U32 useRowMatchFinder;
-    int rsyncable;
-    int minAdaptLevel;
-    int maxAdaptLevel;
-    int ldmFlag;
-    int ldmHashLog;
-    int ldmMinMatch;
-    int ldmBucketSizeLog;
-    int ldmHashRateLog;
-    size_t streamSrcSize;
-    size_t targetCBlockSize;
-    int srcSizeHint;
-    int testMode;
-    ZSTD_paramSwitch_e literalCompressionMode;
-
-    /* IO preferences */
-    U32 removeSrcFile;
-    U32 overwrite;
-
-    /* Computation resources preferences */
-    unsigned memLimit;
-    int nbWorkers;
-
-    int excludeCompressedFiles;
-    int patchFromMode;
-    int contentSize;
-    int allowBlockDevices;
-};
-
 /*-*************************************
 *  Parameters: FIO_ctx_t
 ***************************************/
@@ -355,6 +254,18 @@ struct FIO_ctx_s {
     size_t totalBytesOutput;
 };
 
+static int FIO_shouldDisplayFileSummary(FIO_ctx_t const* fCtx)
+{
+    return fCtx->nbFilesTotal <= 1 || g_display_prefs.displayLevel >= 3;
+}
+
+static int FIO_shouldDisplayMultipleFileSummary(FIO_ctx_t const* fCtx)
+{
+    int const shouldDisplay = (fCtx->nbFilesProcessed >= 1 && fCtx->nbFilesTotal > 1);
+    assert(shouldDisplay || FIO_shouldDisplayFileSummary(fCtx) || fCtx->nbFilesProcessed == 0);
+    return shouldDisplay;
+}
+
 
 /*-*************************************
 *  Parameters: Initialization
@@ -395,6 +306,8 @@ FIO_prefs_t* FIO_createPreferences(void)
     ret->literalCompressionMode = ZSTD_ps_auto;
     ret->excludeCompressedFiles = 0;
     ret->allowBlockDevices = 0;
+    ret->asyncIO = AIO_supported();
+    ret->passThrough = -1;
     return ret;
 }
 
@@ -443,13 +356,13 @@ void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t comp
 
 void FIO_overwriteMode(FIO_prefs_t* const prefs) { prefs->overwrite = 1; }
 
-void FIO_setSparseWrite(FIO_prefs_t* const prefs, unsigned sparse) { prefs->sparseFileSupport = sparse; }
+void FIO_setSparseWrite(FIO_prefs_t* const prefs, int sparse) { prefs->sparseFileSupport = sparse; }
 
 void FIO_setDictIDFlag(FIO_prefs_t* const prefs, int dictIDFlag) { prefs->dictIDFlag = dictIDFlag; }
 
 void FIO_setChecksumFlag(FIO_prefs_t* const prefs, int checksumFlag) { prefs->checksumFlag = checksumFlag; }
 
-void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, unsigned flag) { prefs->removeSrcFile = (flag>0); }
+void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, int flag) { prefs->removeSrcFile = (flag!=0); }
 
 void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit) { prefs->memLimit = memLimit; }
 
@@ -476,7 +389,7 @@ void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog){
     prefs->overlapLog = overlapLog;
 }
 
-void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, unsigned adapt) {
+void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, int adapt) {
     if ((adapt>0) && (prefs->nbWorkers==0))
         EXM_THROW(1, "Adaptive mode is not compatible with single thread mode \n");
     prefs->adaptiveMode = adapt;
@@ -510,7 +423,7 @@ void FIO_setTestMode(FIO_prefs_t* const prefs, int testMode) {
 
 void FIO_setLiteralCompressionMode(
         FIO_prefs_t* const prefs,
-        ZSTD_paramSwitch_e mode) {
+        ZSTD_ParamSwitch_e mode) {
     prefs->literalCompressionMode = mode;
 }
 
@@ -558,6 +471,25 @@ void FIO_setContentSize(FIO_prefs_t* const prefs, int value)
     prefs->contentSize = value != 0;
 }
 
+void FIO_setAsyncIOFlag(FIO_prefs_t* const prefs, int value) {
+#ifdef ZSTD_MULTITHREAD
+    prefs->asyncIO = value;
+#else
+    (void) prefs;
+    (void) value;
+    DISPLAYLEVEL(2, "Note : asyncio is disabled (lack of multithreading support) \n");
+#endif
+}
+
+void FIO_setPassThroughFlag(FIO_prefs_t* const prefs, int value) {
+    prefs->passThrough = (value != 0);
+}
+
+void FIO_setMMapDict(FIO_prefs_t* const prefs, ZSTD_ParamSwitch_e value)
+{
+    prefs->mmapDict = value;
+}
+
 /* FIO_ctx_t functions */
 
 void FIO_setHasStdoutOutput(FIO_ctx_t* const fCtx, int value) {
@@ -595,7 +527,7 @@ static int FIO_removeFile(const char* path)
         DISPLAYLEVEL(2, "zstd: Refusing to remove non-regular file %s\n", path);
         return 0;
     }
-#if defined(_WIN32) || defined(WIN32)
+#if defined(_WIN32)
     /* windows doesn't allow remove read-only files,
      * so try to make it writable first */
     if (!(statbuf.st_mode & _S_IWRITE)) {
@@ -608,26 +540,26 @@ static int FIO_removeFile(const char* path)
 /** FIO_openSrcFile() :
  *  condition : `srcFileName` must be non-NULL. `prefs` may be NULL.
  * @result : FILE* to `srcFileName`, or NULL if it fails */
-static FILE* FIO_openSrcFile(const FIO_prefs_t* const prefs, const char* srcFileName)
+static FILE* FIO_openSrcFile(const FIO_prefs_t* const prefs, const char* srcFileName, stat_t* statbuf)
 {
-    stat_t statbuf;
     int allowBlockDevices = prefs != NULL ? prefs->allowBlockDevices : 0;
     assert(srcFileName != NULL);
+    assert(statbuf != NULL);
     if (!strcmp (srcFileName, stdinmark)) {
         DISPLAYLEVEL(4,"Using stdin for input \n");
         SET_BINARY_MODE(stdin);
         return stdin;
     }
 
-    if (!UTIL_stat(srcFileName, &statbuf)) {
+    if (!UTIL_stat(srcFileName, statbuf)) {
         DISPLAYLEVEL(1, "zstd: can't stat %s : %s -- ignored \n",
                         srcFileName, strerror(errno));
         return NULL;
     }
 
-    if (!UTIL_isRegularFileStat(&statbuf)
-     && !UTIL_isFIFOStat(&statbuf)
-     && !(allowBlockDevices && UTIL_isBlockDevStat(&statbuf))
+    if (!UTIL_isRegularFileStat(statbuf)
+     && !UTIL_isFIFOStat(statbuf)
+     && !(allowBlockDevices && UTIL_isBlockDevStat(statbuf))
     ) {
         DISPLAYLEVEL(1, "zstd: %s is not a regular file -- ignored \n",
                         srcFileName);
@@ -649,6 +581,8 @@ FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
                 const char* srcFileName, const char* dstFileName,
                 const int mode)
 {
+    int isDstRegFile;
+
     if (prefs->testMode) return NULL;  /* do not open file in test mode */
 
     assert(dstFileName != NULL);
@@ -668,11 +602,16 @@ FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
         return NULL;
     }
 
+    isDstRegFile = UTIL_isRegularFile(dstFileName);  /* invoke once */
     if (prefs->sparseFileSupport == 1) {
         prefs->sparseFileSupport = ZSTD_SPARSE_DEFAULT;
+        if (!isDstRegFile) {
+            prefs->sparseFileSupport = 0;
+            DISPLAYLEVEL(4, "Sparse File Support is disabled when output is not a file \n");
+        }
     }
 
-    if (UTIL_isRegularFile(dstFileName)) {
+    if (isDstRegFile) {
         /* Check if destination file already exists */
 #if !defined(_WIN32)
         /* this test does not work on Windows :
@@ -685,7 +624,7 @@ FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
         if (!prefs->overwrite) {
             if (g_display_prefs.displayLevel <= 1) {
                 /* No interaction possible */
-                DISPLAY("zstd: %s already exists; not overwritten  \n",
+                DISPLAYLEVEL(1, "zstd: %s already exists; not overwritten  \n",
                         dstFileName);
                 return NULL;
             }
@@ -717,44 +656,70 @@ FIO_openDstFile(FIO_ctx_t* fCtx, FIO_prefs_t* const prefs,
 #endif
         if (f == NULL) {
             DISPLAYLEVEL(1, "zstd: %s: %s\n", dstFileName, strerror(errno));
+        } else {
+            /* An increased buffer size can provide a significant performance
+             * boost on some platforms. Note that providing a NULL buf with a
+             * size that's not 0 is not defined in ANSI C, but is defined in an
+             * extension. There are three possibilities here:
+             * 1. Libc supports the extended version and everything is good.
+             * 2. Libc ignores the size when buf is NULL, in which case
+             *    everything will continue as if we didn't call `setvbuf()`.
+             * 3. We fail the call and execution continues but a warning
+             *    message might be shown.
+             * In all cases due execution continues. For now, I believe that
+             * this is a more cost-effective solution than managing the buffers
+             * allocations ourselves (will require an API change).
+             */
+            if (setvbuf(f, NULL, _IOFBF, 1 MB)) {
+                DISPLAYLEVEL(2, "Warning: setvbuf failed for %s\n", dstFileName);
+            }
         }
         return f;
     }
 }
 
-/*! FIO_createDictBuffer() :
- *  creates a buffer, pointed by `*bufferPtr`,
+
+/* FIO_getDictFileStat() :
+ */
+static void FIO_getDictFileStat(const char* fileName, stat_t* dictFileStat) {
+    assert(dictFileStat != NULL);
+    if (fileName == NULL) return;
+
+    if (!UTIL_stat(fileName, dictFileStat)) {
+        EXM_THROW(31, "Stat failed on dictionary file %s: %s", fileName, strerror(errno));
+    }
+
+    if (!UTIL_isRegularFileStat(dictFileStat)) {
+        EXM_THROW(32, "Dictionary %s must be a regular file.", fileName);
+    }
+}
+
+/*  FIO_setDictBufferMalloc() :
+ *  allocates a buffer, pointed by `dict->dictBuffer`,
  *  loads `filename` content into it, up to DICTSIZE_MAX bytes.
  * @return : loaded size
  *  if fileName==NULL, returns 0 and a NULL pointer
  */
-static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName, FIO_prefs_t* const prefs)
+static size_t FIO_setDictBufferMalloc(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
 {
     FILE* fileHandle;
     U64 fileSize;
-    stat_t statbuf;
+    void** bufferPtr = &dict->dictBuffer;
 
     assert(bufferPtr != NULL);
+    assert(dictFileStat != NULL);
     *bufferPtr = NULL;
     if (fileName == NULL) return 0;
 
     DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
 
-    if (!UTIL_stat(fileName, &statbuf)) {
-        EXM_THROW(31, "Stat failed on dictionary file %s: %s", fileName, strerror(errno));
-    }
-
-    if (!UTIL_isRegularFileStat(&statbuf)) {
-        EXM_THROW(32, "Dictionary %s must be a regular file.", fileName);
-    }
-
     fileHandle = fopen(fileName, "rb");
 
     if (fileHandle == NULL) {
         EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
     }
 
-    fileSize = UTIL_getFileSizeStat(&statbuf);
+    fileSize = UTIL_getFileSizeStat(dictFileStat);
     {
         size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
         if (fileSize >  dictSizeMax) {
@@ -774,6 +739,130 @@ static size_t FIO_createDictBuffer(void** bufferPtr, const char* fileName, FIO_p
     return (size_t)fileSize;
 }
 
+#if (PLATFORM_POSIX_VERSION > 0)
+#include <sys/mman.h>
+static void FIO_munmap(FIO_Dict_t* dict)
+{
+    munmap(dict->dictBuffer, dict->dictBufferSize);
+    dict->dictBuffer = NULL;
+    dict->dictBufferSize = 0;
+}
+static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
+{
+    int fileHandle;
+    U64 fileSize;
+    void** bufferPtr = &dict->dictBuffer;
+
+    assert(bufferPtr != NULL);
+    assert(dictFileStat != NULL);
+    *bufferPtr = NULL;
+    if (fileName == NULL) return 0;
+
+    DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
+
+    fileHandle = open(fileName, O_RDONLY);
+
+    if (fileHandle == -1) {
+        EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
+    }
+
+    fileSize = UTIL_getFileSizeStat(dictFileStat);
+    {
+        size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
+        if (fileSize >  dictSizeMax) {
+            EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
+                            fileName,  (unsigned)dictSizeMax);   /* avoid extreme cases */
+        }
+    }
+
+    *bufferPtr = mmap(NULL, (size_t)fileSize, PROT_READ, MAP_PRIVATE, fileHandle, 0);
+    if (*bufferPtr==NULL) EXM_THROW(34, "%s", strerror(errno));
+
+    close(fileHandle);
+    return (size_t)fileSize;
+}
+#elif defined(_MSC_VER) || defined(_WIN32)
+#include <windows.h>
+static void FIO_munmap(FIO_Dict_t* dict)
+{
+    UnmapViewOfFile(dict->dictBuffer);
+    CloseHandle(dict->dictHandle);
+    dict->dictBuffer = NULL;
+    dict->dictBufferSize = 0;
+}
+static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
+{
+    HANDLE fileHandle, mapping;
+    U64 fileSize;
+    void** bufferPtr = &dict->dictBuffer;
+
+    assert(bufferPtr != NULL);
+    assert(dictFileStat != NULL);
+    *bufferPtr = NULL;
+    if (fileName == NULL) return 0;
+
+    DISPLAYLEVEL(4,"Loading %s as dictionary \n", fileName);
+
+    fileHandle = CreateFileA(fileName, GENERIC_READ, FILE_SHARE_READ, NULL, OPEN_EXISTING, FILE_ATTRIBUTE_READONLY, NULL);
+
+    if (fileHandle == INVALID_HANDLE_VALUE) {
+        EXM_THROW(33, "Couldn't open dictionary %s: %s", fileName, strerror(errno));
+    }
+
+    fileSize = UTIL_getFileSizeStat(dictFileStat);
+    {
+        size_t const dictSizeMax = prefs->patchFromMode ? prefs->memLimit : DICTSIZE_MAX;
+        if (fileSize >  dictSizeMax) {
+            EXM_THROW(34, "Dictionary file %s is too large (> %u bytes)",
+                            fileName,  (unsigned)dictSizeMax);   /* avoid extreme cases */
+        }
+    }
+
+    mapping = CreateFileMapping(fileHandle, NULL, PAGE_READONLY, 0, 0, NULL);
+	if (mapping == NULL) {
+        EXM_THROW(35, "Couldn't map dictionary %s: %s", fileName, strerror(errno));
+    }
+
+	*bufferPtr = MapViewOfFile(mapping, FILE_MAP_READ, 0, 0, (DWORD)fileSize); /* we can only cast to DWORD here because dictSize <= 2GB */
+	if (*bufferPtr==NULL) EXM_THROW(36, "%s", strerror(errno));
+
+    dict->dictHandle = fileHandle;
+    return (size_t)fileSize;
+}
+#else
+static size_t FIO_setDictBufferMMap(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat)
+{
+   return FIO_setDictBufferMalloc(dict, fileName, prefs, dictFileStat);
+}
+static void FIO_munmap(FIO_Dict_t* dict) {
+   free(dict->dictBuffer);
+   dict->dictBuffer = NULL;
+   dict->dictBufferSize = 0;
+}
+#endif
+
+static void FIO_freeDict(FIO_Dict_t* dict) {
+    if (dict->dictBufferType == FIO_mallocDict) {
+        free(dict->dictBuffer);
+        dict->dictBuffer = NULL;
+        dict->dictBufferSize = 0;
+    } else if (dict->dictBufferType == FIO_mmapDict)  {
+        FIO_munmap(dict);
+    } else {
+        assert(0); /* Should not reach this case */
+    }
+}
+
+static void FIO_initDict(FIO_Dict_t* dict, const char* fileName, FIO_prefs_t* const prefs, stat_t* dictFileStat, FIO_dictBufferType_t dictBufferType) {
+    dict->dictBufferType = dictBufferType;
+    if (dict->dictBufferType == FIO_mallocDict) {
+        dict->dictBufferSize = FIO_setDictBufferMalloc(dict, fileName, prefs, dictFileStat);
+    } else if (dict->dictBufferType == FIO_mmapDict)  {
+        dict->dictBufferSize = FIO_setDictBufferMMap(dict, fileName, prefs, dictFileStat);
+    } else {
+        assert(0); /* Should not reach this case */
+    }
+}
 
 
 /* FIO_checkFilenameCollisions() :
@@ -785,7 +874,7 @@ int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles) {
 
     filenameTableSorted = (const char**) malloc(sizeof(char*) * nbFiles);
     if (!filenameTableSorted) {
-        DISPLAY("Unable to malloc new str array, not checking for name collisions\n");
+        DISPLAYLEVEL(1, "Allocation error during filename collision checking \n");
         return 1;
     }
 
@@ -802,7 +891,7 @@ int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles) {
     prevElem = filenameTableSorted[0];
     for (u = 1; u < nbFiles; ++u) {
         if (strcmp(prevElem, filenameTableSorted[u]) == 0) {
-            DISPLAY("WARNING: Two files have same filename: %s\n", prevElem);
+            DISPLAYLEVEL(2, "WARNING: Two files have same filename: %s\n", prevElem);
         }
         prevElem = filenameTableSorted[u];
     }
@@ -885,45 +974,89 @@ static void FIO_adjustMemLimitForPatchFromMode(FIO_prefs_t* const prefs,
     FIO_setMemLimit(prefs, (unsigned)maxSize);
 }
 
-/* FIO_removeMultiFilesWarning() :
+/* FIO_multiFilesConcatWarning() :
+ * This function handles logic when processing multiple files with -o or -c, displaying the appropriate warnings/prompts.
  * Returns 1 if the console should abort, 0 if console should proceed.
- * This function handles logic when processing multiple files with -o, displaying the appropriate warnings/prompts.
  *
- * If -f is specified, or there is just 1 file, zstd will always proceed as usual.
- * If --rm is specified, there will be a prompt asking for user confirmation.
- *         If -f is specified with --rm, zstd will proceed as usual
- *         If -q is specified with --rm, zstd will abort pre-emptively
- *         If neither flag is specified, zstd will prompt the user for confirmation to proceed.
- * If --rm is not specified, then zstd will print a warning to the user (which can be silenced with -q).
- * However, if the output is stdout, we will always abort rather than displaying the warning prompt.
+ * If output is stdout or test mode is active, check that `--rm` disabled.
+ *
+ * If there is just 1 file to process, zstd will proceed as usual.
+ * If each file get processed into its own separate destination file, proceed as usual.
+ *
+ * When multiple files are processed into a single output,
+ * display a warning message, then disable --rm if it's set.
+ *
+ * If -f is specified or if output is stdout, just proceed.
+ * If output is set with -o, prompt for confirmation.
  */
-static int FIO_removeMultiFilesWarning(FIO_ctx_t* const fCtx, const FIO_prefs_t* const prefs, const char* outFileName, int displayLevelCutoff)
+static int FIO_multiFilesConcatWarning(const FIO_ctx_t* fCtx, FIO_prefs_t* prefs, const char* outFileName, int displayLevelCutoff)
 {
-    int error = 0;
-    if (fCtx->nbFilesTotal > 1 && !prefs->overwrite) {
-        if (g_display_prefs.displayLevel <= displayLevelCutoff) {
-            if (prefs->removeSrcFile) {
-                DISPLAYLEVEL(1, "zstd: Aborting... not deleting files and processing into dst: %s\n", outFileName);
-                error =  1;
-            }
-        } else {
-            if (!strcmp(outFileName, stdoutmark)) {
-                DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into stdout. \n");
-            } else {
-                DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into a single output file: %s \n", outFileName);
-            }
-            DISPLAYLEVEL(2, "The concatenated output CANNOT regenerate the original directory tree. \n")
-            if (prefs->removeSrcFile) {
-                if (fCtx->hasStdoutOutput) {
-                    DISPLAYLEVEL(1, "Aborting. Use -f if you really want to delete the files and output to stdout\n");
-                    error = 1;
-                } else {
-                    error = g_display_prefs.displayLevel > displayLevelCutoff && UTIL_requireUserConfirmation("This is a destructive operation. Proceed? (y/n): ", "Aborting...", "yY", fCtx->hasStdinInput);
-                }
-            }
-        }
+    if (fCtx->hasStdoutOutput) {
+        if (prefs->removeSrcFile)
+            /* this should not happen ; hard fail, to protect user's data
+             * note: this should rather be an assert(), but we want to be certain that user's data will not be wiped out in case it nonetheless happen */
+            EXM_THROW(43, "It's not allowed to remove input files when processed output is piped to stdout. "
+                "This scenario is not supposed to be possible. "
+                "This is a programming error. File an issue for it to be fixed.");
     }
-    return error;
+    if (prefs->testMode) {
+        if (prefs->removeSrcFile)
+            /* this should not happen ; hard fail, to protect user's data
+             * note: this should rather be an assert(), but we want to be certain that user's data will not be wiped out in case it nonetheless happen */
+            EXM_THROW(43, "Test mode shall not remove input files! "
+                 "This scenario is not supposed to be possible. "
+                 "This is a programming error. File an issue for it to be fixed.");
+        return 0;
+    }
+
+    if (fCtx->nbFilesTotal == 1) return 0;
+    assert(fCtx->nbFilesTotal > 1);
+
+    if (!outFileName) return 0;
+
+    if (fCtx->hasStdoutOutput) {
+        DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into stdout. \n");
+    } else {
+        DISPLAYLEVEL(2, "zstd: WARNING: all input files will be processed and concatenated into a single output file: %s \n", outFileName);
+    }
+    DISPLAYLEVEL(2, "The concatenated output CANNOT regenerate original file names nor directory structure. \n")
+
+    /* multi-input into single output : --rm is not allowed */
+    if (prefs->removeSrcFile) {
+        DISPLAYLEVEL(2, "Since it's a destructive operation, input files will not be removed. \n");
+        prefs->removeSrcFile = 0;
+    }
+
+    if (fCtx->hasStdoutOutput) return 0;
+    if (prefs->overwrite) return 0;
+
+    /* multiple files concatenated into single destination file using -o without -f */
+    if (g_display_prefs.displayLevel <= displayLevelCutoff) {
+        /* quiet mode => no prompt => fail automatically */
+        DISPLAYLEVEL(1, "Concatenating multiple processed inputs into a single output loses file metadata. \n");
+        DISPLAYLEVEL(1, "Aborting. \n");
+        return 1;
+    }
+    /* normal mode => prompt */
+    return UTIL_requireUserConfirmation("Proceed? (y/n): ", "Aborting...", "yY", fCtx->hasStdinInput);
+}
+
+static ZSTD_inBuffer setInBuffer(const void* buf, size_t s, size_t pos)
+{
+    ZSTD_inBuffer i;
+    i.src = buf;
+    i.size = s;
+    i.pos = pos;
+    return i;
+}
+
+static ZSTD_outBuffer setOutBuffer(void* buf, size_t s, size_t pos)
+{
+    ZSTD_outBuffer o;
+    o.dst = buf;
+    o.size = s;
+    o.pos = pos;
+    return o;
 }
 
 #ifndef ZSTD_NOCOMPRESS
@@ -932,16 +1065,12 @@ static int FIO_removeMultiFilesWarning(FIO_ctx_t* const fCtx, const FIO_prefs_t*
  *  Compression
  ************************************************************************/
 typedef struct {
-    FILE* srcFile;
-    FILE* dstFile;
-    void*  srcBuffer;
-    size_t srcBufferSize;
-    void*  dstBuffer;
-    size_t dstBufferSize;
-    void* dictBuffer;
-    size_t dictBufferSize;
+    FIO_Dict_t dict;
     const char* dictFileName;
+    stat_t dictFileStat;
     ZSTD_CStream* cctx;
+    WritePoolCtx_t *writeCtx;
+    ReadPoolCtx_t *readCtx;
 } cRess_t;
 
 /** ZSTD_cycleLog() :
@@ -967,21 +1096,25 @@ static void FIO_adjustParamsForPatchFromMode(FIO_prefs_t* const prefs,
     comprParams->windowLog = MAX(ZSTD_WINDOWLOG_MIN, MIN(ZSTD_WINDOWLOG_MAX, fileWindowLog));
     if (fileWindowLog > ZSTD_cycleLog(cParams.chainLog, cParams.strategy)) {
         if (!prefs->ldmFlag)
-            DISPLAYLEVEL(1, "long mode automatically triggered\n");
+            DISPLAYLEVEL(2, "long mode automatically triggered\n");
         FIO_setLdmFlag(prefs, 1);
     }
     if (cParams.strategy >= ZSTD_btopt) {
-        DISPLAYLEVEL(1, "[Optimal parser notes] Consider the following to improve patch size at the cost of speed:\n");
-        DISPLAYLEVEL(1, "- Use --single-thread mode in the zstd cli\n");
-        DISPLAYLEVEL(1, "- Set a larger targetLength (eg. --zstd=targetLength=4096)\n");
-        DISPLAYLEVEL(1, "- Set a larger chainLog (eg. --zstd=chainLog=%u)\n", ZSTD_CHAINLOG_MAX);
-        DISPLAYLEVEL(1, "Also consider playing around with searchLog and hashLog\n");
+        DISPLAYLEVEL(4, "[Optimal parser notes] Consider the following to improve patch size at the cost of speed:\n");
+        DISPLAYLEVEL(4, "- Set a larger targetLength (e.g. --zstd=targetLength=4096)\n");
+        DISPLAYLEVEL(4, "- Set a larger chainLog (e.g. --zstd=chainLog=%u)\n", ZSTD_CHAINLOG_MAX);
+        DISPLAYLEVEL(4, "- Set a larger LDM hashLog (e.g. --zstd=ldmHashLog=%u)\n", ZSTD_LDM_HASHLOG_MAX);
+        DISPLAYLEVEL(4, "- Set a smaller LDM rateLog (e.g. --zstd=ldmHashRateLog=%u)\n", ZSTD_LDM_HASHRATELOG_MIN);
+        DISPLAYLEVEL(4, "Also consider playing around with searchLog and hashLog\n");
     }
 }
 
 static cRess_t FIO_createCResources(FIO_prefs_t* const prefs,
                                     const char* dictFileName, unsigned long long const maxSrcFileSize,
                                     int cLevel, ZSTD_compressionParameters comprParams) {
+    int useMMap = prefs->mmapDict == ZSTD_ps_enable;
+    int forceNoUseMMap = prefs->mmapDict == ZSTD_ps_disable;
+    FIO_dictBufferType_t dictBufferType;
     cRess_t ress;
     memset(&ress, 0, sizeof(ress));
 
@@ -990,23 +1123,26 @@ static cRess_t FIO_createCResources(FIO_prefs_t* const prefs,
     if (ress.cctx == NULL)
         EXM_THROW(30, "allocation error (%s): can't create ZSTD_CCtx",
                     strerror(errno));
-    ress.srcBufferSize = ZSTD_CStreamInSize();
-    ress.srcBuffer = malloc(ress.srcBufferSize);
-    ress.dstBufferSize = ZSTD_CStreamOutSize();
+
+    FIO_getDictFileStat(dictFileName, &ress.dictFileStat);
 
     /* need to update memLimit before calling createDictBuffer
      * because of memLimit check inside it */
     if (prefs->patchFromMode) {
+        U64 const dictSize = UTIL_getFileSizeStat(&ress.dictFileStat);
         unsigned long long const ssSize = (unsigned long long)prefs->streamSrcSize;
-        FIO_adjustParamsForPatchFromMode(prefs, &comprParams, UTIL_getFileSize(dictFileName), ssSize > 0 ? ssSize : maxSrcFileSize, cLevel);
+        useMMap |= dictSize > prefs->memLimit;
+        FIO_adjustParamsForPatchFromMode(prefs, &comprParams, dictSize, ssSize > 0 ? ssSize : maxSrcFileSize, cLevel);
     }
-    ress.dstBuffer = malloc(ress.dstBufferSize);
-    ress.dictBufferSize = FIO_createDictBuffer(&ress.dictBuffer, dictFileName, prefs);   /* works with dictFileName==NULL */
-    if (!ress.srcBuffer || !ress.dstBuffer)
-        EXM_THROW(31, "allocation error : not enough memory");
+
+    dictBufferType = (useMMap && !forceNoUseMMap) ? FIO_mmapDict : FIO_mallocDict;
+    FIO_initDict(&ress.dict, dictFileName, prefs, &ress.dictFileStat, dictBufferType);   /* works with dictFileName==NULL */
+
+    ress.writeCtx = AIO_WritePool_create(prefs, ZSTD_CStreamOutSize());
+    ress.readCtx = AIO_ReadPool_create(prefs, ZSTD_CStreamInSize());
 
     /* Advanced parameters, including dictionary */
-    if (dictFileName && (ress.dictBuffer==NULL))
+    if (dictFileName && (ress.dict.dictBuffer==NULL))
         EXM_THROW(32, "allocation error : can't create dictBuffer");
     ress.dictFileName = dictFileName;
 
@@ -1056,19 +1192,19 @@ static cRess_t FIO_createCResources(FIO_prefs_t* const prefs,
 #endif
     /* dictionary */
     if (prefs->patchFromMode) {
-        CHECK( ZSTD_CCtx_refPrefix(ress.cctx, ress.dictBuffer, ress.dictBufferSize) );
+        CHECK( ZSTD_CCtx_refPrefix(ress.cctx, ress.dict.dictBuffer, ress.dict.dictBufferSize) );
     } else {
-        CHECK( ZSTD_CCtx_loadDictionary(ress.cctx, ress.dictBuffer, ress.dictBufferSize) );
+        CHECK( ZSTD_CCtx_loadDictionary_byReference(ress.cctx, ress.dict.dictBuffer, ress.dict.dictBufferSize) );
     }
 
     return ress;
 }
 
-static void FIO_freeCResources(const cRess_t* const ress)
+static void FIO_freeCResources(cRess_t* const ress)
 {
-    free(ress->srcBuffer);
-    free(ress->dstBuffer);
-    free(ress->dictBuffer);
+    FIO_freeDict(&(ress->dict));
+    AIO_WritePool_free(ress->writeCtx);
+    AIO_ReadPool_free(ress->readCtx);
     ZSTD_freeCStream(ress->cctx);   /* never fails */
 }
 
@@ -1081,6 +1217,7 @@ FIO_compressGzFrame(const cRess_t* ress,  /* buffers & handlers are used, but no
 {
     unsigned long long inFileSize = 0, outFileSize = 0;
     z_stream strm;
+    IOJob_t *writeJob = NULL;
 
     if (compressionLevel > Z_BEST_COMPRESSION)
         compressionLevel = Z_BEST_COMPRESSION;
@@ -1091,56 +1228,65 @@ FIO_compressGzFrame(const cRess_t* ress,  /* buffers & handlers are used, but no
 
     {   int const ret = deflateInit2(&strm, compressionLevel, Z_DEFLATED,
                         15 /* maxWindowLogSize */ + 16 /* gzip only */,
-                        8, Z_DEFAULT_STRATEGY); /* see http://www.zlib.net/manual.html */
+                        8, Z_DEFAULT_STRATEGY); /* see https://www.zlib.net/manual.html */
         if (ret != Z_OK) {
             EXM_THROW(71, "zstd: %s: deflateInit2 error %d \n", srcFileName, ret);
     }   }
 
+    writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
     strm.next_in = 0;
     strm.avail_in = 0;
-    strm.next_out = (Bytef*)ress->dstBuffer;
-    strm.avail_out = (uInt)ress->dstBufferSize;
+    strm.next_out = (Bytef*)writeJob->buffer;
+    strm.avail_out = (uInt)writeJob->bufferSize;
 
     while (1) {
         int ret;
         if (strm.avail_in == 0) {
-            size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile);
-            if (inSize == 0) break;
-            inFileSize += inSize;
-            strm.next_in = (z_const unsigned char*)ress->srcBuffer;
-            strm.avail_in = (uInt)inSize;
+            AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_CStreamInSize());
+            if (ress->readCtx->srcBufferLoaded == 0) break;
+            inFileSize += ress->readCtx->srcBufferLoaded;
+            strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
+            strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
+        }
+
+        {
+            size_t const availBefore = strm.avail_in;
+            ret = deflate(&strm, Z_NO_FLUSH);
+            AIO_ReadPool_consumeBytes(ress->readCtx, availBefore - strm.avail_in);
         }
-        ret = deflate(&strm, Z_NO_FLUSH);
+
         if (ret != Z_OK)
             EXM_THROW(72, "zstd: %s: deflate error %d \n", srcFileName, ret);
-        {   size_t const cSize = ress->dstBufferSize - strm.avail_out;
+        {   size_t const cSize = writeJob->bufferSize - strm.avail_out;
             if (cSize) {
-                if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize)
-                    EXM_THROW(73, "Write error : cannot write to output file : %s ", strerror(errno));
+                writeJob->usedBufferSize = cSize;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 outFileSize += cSize;
-                strm.next_out = (Bytef*)ress->dstBuffer;
-                strm.avail_out = (uInt)ress->dstBufferSize;
-        }   }
+                strm.next_out = (Bytef*)writeJob->buffer;
+                strm.avail_out = (uInt)writeJob->bufferSize;
+            }   }
         if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
-            DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%% ",
-                            (unsigned)(inFileSize>>20),
-                            (double)outFileSize/inFileSize*100)
+            DISPLAYUPDATE_PROGRESS(
+                    "\rRead : %u MB ==> %.2f%% ",
+                    (unsigned)(inFileSize>>20),
+                    (double)outFileSize/(double)inFileSize*100)
         } else {
-            DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%% ",
-                            (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
-                            (double)outFileSize/inFileSize*100);
+            DISPLAYUPDATE_PROGRESS(
+                    "\rRead : %u / %u MB ==> %.2f%% ",
+                    (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
+                    (double)outFileSize/(double)inFileSize*100);
     }   }
 
     while (1) {
         int const ret = deflate(&strm, Z_FINISH);
-        {   size_t const cSize = ress->dstBufferSize - strm.avail_out;
+        {   size_t const cSize = writeJob->bufferSize - strm.avail_out;
             if (cSize) {
-                if (fwrite(ress->dstBuffer, 1, cSize, ress->dstFile) != cSize)
-                    EXM_THROW(75, "Write error : %s ", strerror(errno));
+                writeJob->usedBufferSize = cSize;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 outFileSize += cSize;
-                strm.next_out = (Bytef*)ress->dstBuffer;
-                strm.avail_out = (uInt)ress->dstBufferSize;
-        }   }
+                strm.next_out = (Bytef*)writeJob->buffer;
+                strm.avail_out = (uInt)writeJob->bufferSize;
+            }   }
         if (ret == Z_STREAM_END) break;
         if (ret != Z_BUF_ERROR)
             EXM_THROW(77, "zstd: %s: deflate error %d \n", srcFileName, ret);
@@ -1151,6 +1297,8 @@ FIO_compressGzFrame(const cRess_t* ress,  /* buffers & handlers are used, but no
             EXM_THROW(79, "zstd: %s: deflateEnd error %d \n", srcFileName, ret);
     }   }
     *readsize = inFileSize;
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
     return outFileSize;
 }
 #endif
@@ -1166,6 +1314,7 @@ FIO_compressLzmaFrame(cRess_t* ress,
     lzma_stream strm = LZMA_STREAM_INIT;
     lzma_action action = LZMA_RUN;
     lzma_ret ret;
+    IOJob_t *writeJob = NULL;
 
     if (compressionLevel < 0) compressionLevel = 0;
     if (compressionLevel > 9) compressionLevel = 9;
@@ -1183,46 +1332,55 @@ FIO_compressLzmaFrame(cRess_t* ress,
             EXM_THROW(83, "zstd: %s: lzma_easy_encoder error %d", srcFileName, ret);
     }
 
+    writeJob =AIO_WritePool_acquireJob(ress->writeCtx);
+    strm.next_out = (BYTE*)writeJob->buffer;
+    strm.avail_out = writeJob->bufferSize;
     strm.next_in = 0;
     strm.avail_in = 0;
-    strm.next_out = (BYTE*)ress->dstBuffer;
-    strm.avail_out = ress->dstBufferSize;
 
     while (1) {
         if (strm.avail_in == 0) {
-            size_t const inSize = fread(ress->srcBuffer, 1, ress->srcBufferSize, ress->srcFile);
-            if (inSize == 0) action = LZMA_FINISH;
+            size_t const inSize = AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_CStreamInSize());
+            if (ress->readCtx->srcBufferLoaded == 0) action = LZMA_FINISH;
             inFileSize += inSize;
-            strm.next_in = (BYTE const*)ress->srcBuffer;
-            strm.avail_in = inSize;
+            strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
+            strm.avail_in = ress->readCtx->srcBufferLoaded;
+        }
+
+        {
+            size_t const availBefore = strm.avail_in;
+            ret = lzma_code(&strm, action);
+            AIO_ReadPool_consumeBytes(ress->readCtx, availBefore - strm.avail_in);
         }
 
-        ret = lzma_code(&strm, action);
 
         if (ret != LZMA_OK && ret != LZMA_STREAM_END)
             EXM_THROW(84, "zstd: %s: lzma_code encoding error %d", srcFileName, ret);
-        {   size_t const compBytes = ress->dstBufferSize - strm.avail_out;
+        {   size_t const compBytes = writeJob->bufferSize - strm.avail_out;
             if (compBytes) {
-                if (fwrite(ress->dstBuffer, 1, compBytes, ress->dstFile) != compBytes)
-                    EXM_THROW(85, "Write error : %s", strerror(errno));
+                writeJob->usedBufferSize = compBytes;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 outFileSize += compBytes;
-                strm.next_out = (BYTE*)ress->dstBuffer;
-                strm.avail_out = ress->dstBufferSize;
+                strm.next_out = (BYTE*)writeJob->buffer;
+                strm.avail_out = writeJob->bufferSize;
         }   }
         if (srcFileSize == UTIL_FILESIZE_UNKNOWN)
-            DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
+            DISPLAYUPDATE_PROGRESS("\rRead : %u MB ==> %.2f%%",
                             (unsigned)(inFileSize>>20),
-                            (double)outFileSize/inFileSize*100)
+                            (double)outFileSize/(double)inFileSize*100)
         else
-            DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
+            DISPLAYUPDATE_PROGRESS("\rRead : %u / %u MB ==> %.2f%%",
                             (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
-                            (double)outFileSize/inFileSize*100);
+                            (double)outFileSize/(double)inFileSize*100);
         if (ret == LZMA_STREAM_END) break;
     }
 
     lzma_end(&strm);
     *readsize = inFileSize;
 
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
+
     return outFileSize;
 }
 #endif
@@ -1248,15 +1406,18 @@ FIO_compressLz4Frame(cRess_t* ress,
     LZ4F_preferences_t prefs;
     LZ4F_compressionContext_t ctx;
 
+    IOJob_t* writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
+
     LZ4F_errorCode_t const errorCode = LZ4F_createCompressionContext(&ctx, LZ4F_VERSION);
     if (LZ4F_isError(errorCode))
         EXM_THROW(31, "zstd: failed to create lz4 compression context");
 
     memset(&prefs, 0, sizeof(prefs));
 
-    assert(blockSize <= ress->srcBufferSize);
+    assert(blockSize <= ress->readCtx->base.jobBufferSize);
 
-    prefs.autoFlush = 1;
+    /* autoflush off to mitigate a bug in lz4<=1.9.3 for compression level 12 */
+    prefs.autoFlush = 0;
     prefs.compressionLevel = compressionLevel;
     prefs.frameInfo.blockMode = LZ4F_blockLinked;
     prefs.frameInfo.blockSizeID = LZ4F_max64KB;
@@ -1264,75 +1425,68 @@ FIO_compressLz4Frame(cRess_t* ress,
 #if LZ4_VERSION_NUMBER >= 10600
     prefs.frameInfo.contentSize = (srcFileSize==UTIL_FILESIZE_UNKNOWN) ? 0 : srcFileSize;
 #endif
-    assert(LZ4F_compressBound(blockSize, &prefs) <= ress->dstBufferSize);
+    assert(LZ4F_compressBound(blockSize, &prefs) <= writeJob->bufferSize);
 
     {
-        size_t readSize;
-        size_t headerSize = LZ4F_compressBegin(ctx, ress->dstBuffer, ress->dstBufferSize, &prefs);
+        size_t headerSize = LZ4F_compressBegin(ctx, writeJob->buffer, writeJob->bufferSize, &prefs);
         if (LZ4F_isError(headerSize))
             EXM_THROW(33, "File header generation failed : %s",
                             LZ4F_getErrorName(headerSize));
-        if (fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile) != headerSize)
-            EXM_THROW(34, "Write error : %s (cannot write header)", strerror(errno));
+        writeJob->usedBufferSize = headerSize;
+        AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
         outFileSize += headerSize;
 
         /* Read first block */
-        readSize  = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile);
-        inFileSize += readSize;
+        inFileSize += AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
 
         /* Main Loop */
-        while (readSize>0) {
-            size_t const outSize = LZ4F_compressUpdate(ctx,
-                                        ress->dstBuffer, ress->dstBufferSize,
-                                        ress->srcBuffer, readSize, NULL);
+        while (ress->readCtx->srcBufferLoaded) {
+            size_t inSize = MIN(blockSize, ress->readCtx->srcBufferLoaded);
+            size_t const outSize = LZ4F_compressUpdate(ctx, writeJob->buffer, writeJob->bufferSize,
+                                                       ress->readCtx->srcBuffer, inSize, NULL);
             if (LZ4F_isError(outSize))
                 EXM_THROW(35, "zstd: %s: lz4 compression failed : %s",
                             srcFileName, LZ4F_getErrorName(outSize));
             outFileSize += outSize;
             if (srcFileSize == UTIL_FILESIZE_UNKNOWN) {
-                DISPLAYUPDATE(2, "\rRead : %u MB ==> %.2f%%",
+                DISPLAYUPDATE_PROGRESS("\rRead : %u MB ==> %.2f%%",
                                 (unsigned)(inFileSize>>20),
-                                (double)outFileSize/inFileSize*100)
+                                (double)outFileSize/(double)inFileSize*100)
             } else {
-                DISPLAYUPDATE(2, "\rRead : %u / %u MB ==> %.2f%%",
+                DISPLAYUPDATE_PROGRESS("\rRead : %u / %u MB ==> %.2f%%",
                                 (unsigned)(inFileSize>>20), (unsigned)(srcFileSize>>20),
-                                (double)outFileSize/inFileSize*100);
+                                (double)outFileSize/(double)inFileSize*100);
             }
 
             /* Write Block */
-            {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, outSize, ress->dstFile);
-                if (sizeCheck != outSize)
-                    EXM_THROW(36, "Write error : %s", strerror(errno));
-            }
+            writeJob->usedBufferSize = outSize;
+            AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
 
             /* Read next block */
-            readSize  = fread(ress->srcBuffer, (size_t)1, (size_t)blockSize, ress->srcFile);
-            inFileSize += readSize;
+            AIO_ReadPool_consumeBytes(ress->readCtx, inSize);
+            inFileSize += AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
         }
-        if (ferror(ress->srcFile)) EXM_THROW(37, "Error reading %s ", srcFileName);
 
         /* End of Stream mark */
-        headerSize = LZ4F_compressEnd(ctx, ress->dstBuffer, ress->dstBufferSize, NULL);
+        headerSize = LZ4F_compressEnd(ctx, writeJob->buffer, writeJob->bufferSize, NULL);
         if (LZ4F_isError(headerSize))
             EXM_THROW(38, "zstd: %s: lz4 end of file generation failed : %s",
                         srcFileName, LZ4F_getErrorName(headerSize));
 
-        {   size_t const sizeCheck = fwrite(ress->dstBuffer, 1, headerSize, ress->dstFile);
-            if (sizeCheck != headerSize)
-                EXM_THROW(39, "Write error : %s (cannot write end of stream)",
-                            strerror(errno));
-        }
+        writeJob->usedBufferSize = headerSize;
+        AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
         outFileSize += headerSize;
     }
 
     *readsize = inFileSize;
     LZ4F_freeCompressionContext(ctx);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
 
     return outFileSize;
 }
 #endif
 
-
 static unsigned long long
 FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
                       FIO_prefs_t* const prefs,
@@ -1341,8 +1495,8 @@ FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
                       int compressionLevel, U64* readsize)
 {
     cRess_t const ress = *ressPtr;
-    FILE* const srcFile = ress.srcFile;
-    FILE* const dstFile = ress.dstFile;
+    IOJob_t* writeJob = AIO_WritePool_acquireJob(ressPtr->writeCtx);
+
     U64 compressedfilesize = 0;
     ZSTD_EndDirective directive = ZSTD_e_continue;
     U64 pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;
@@ -1356,6 +1510,9 @@ FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
     unsigned inputPresented = 0;
     unsigned inputBlocked = 0;
     unsigned lastJobID = 0;
+    UTIL_time_t lastAdaptTime = UTIL_getTime();
+    U64 const adaptEveryMicro = REFRESH_RATE;
+
     UTIL_HumanReadableSize_t const file_hrs = UTIL_makeHumanReadableSize(fileSize);
 
     DISPLAYLEVEL(6, "compression using zstd format \n");
@@ -1370,29 +1527,32 @@ FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
       CHECK( ZSTD_CCtx_setPledgedSrcSize(ress.cctx, prefs->streamSrcSize) );
     }
 
-    {
-        int windowLog;
+    {   int windowLog;
         UTIL_HumanReadableSize_t windowSize;
         CHECK(ZSTD_CCtx_getParameter(ress.cctx, ZSTD_c_windowLog, &windowLog));
         if (windowLog == 0) {
-            const ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, fileSize, 0);
-            windowLog = cParams.windowLog;
+            if (prefs->ldmFlag) {
+                /* If long mode is set without a window size libzstd will set this size internally */
+                windowLog = ZSTD_WINDOWLOG_LIMIT_DEFAULT;
+            } else {
+                const ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, fileSize, 0);
+                windowLog = (int)cParams.windowLog;
+            }
         }
         windowSize = UTIL_makeHumanReadableSize(MAX(1ULL, MIN(1ULL << windowLog, pledgedSrcSize)));
         DISPLAYLEVEL(4, "Decompression will require %.*f%s of memory\n", windowSize.precision, windowSize.value, windowSize.suffix);
     }
-    (void)srcFileName;
 
     /* Main compression loop */
     do {
         size_t stillToFlush;
         /* Fill input Buffer */
-        size_t const inSize = fread(ress.srcBuffer, (size_t)1, ress.srcBufferSize, srcFile);
-        ZSTD_inBuffer inBuff = { ress.srcBuffer, inSize, 0 };
+        size_t const inSize = AIO_ReadPool_fillBuffer(ress.readCtx, ZSTD_CStreamInSize());
+        ZSTD_inBuffer inBuff = setInBuffer( ress.readCtx->srcBuffer, ress.readCtx->srcBufferLoaded, 0 );
         DISPLAYLEVEL(6, "fread %u bytes from source \n", (unsigned)inSize);
         *readsize += inSize;
 
-        if ((inSize == 0) || (*readsize == fileSize))
+        if ((ress.readCtx->srcBufferLoaded == 0) || (*readsize == fileSize))
             directive = ZSTD_e_end;
 
         stillToFlush = 1;
@@ -1400,9 +1560,10 @@ FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
             || (directive == ZSTD_e_end && stillToFlush != 0) ) {
 
             size_t const oldIPos = inBuff.pos;
-            ZSTD_outBuffer outBuff = { ress.dstBuffer, ress.dstBufferSize, 0 };
+            ZSTD_outBuffer outBuff = setOutBuffer( writeJob->buffer, writeJob->bufferSize, 0 );
             size_t const toFlushNow = ZSTD_toFlushNow(ress.cctx);
             CHECK_V(stillToFlush, ZSTD_compressStream2(ress.cctx, &outBuff, &inBuff, directive));
+            AIO_ReadPool_consumeBytes(ress.readCtx, inBuff.pos - oldIPos);
 
             /* count stats */
             inputPresented++;
@@ -1411,151 +1572,155 @@ FIO_compressZstdFrame(FIO_ctx_t* const fCtx,
 
             /* Write compressed stream */
             DISPLAYLEVEL(6, "ZSTD_compress_generic(end:%u) => input pos(%u)<=(%u)size ; output generated %u bytes \n",
-                            (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos);
+                         (unsigned)directive, (unsigned)inBuff.pos, (unsigned)inBuff.size, (unsigned)outBuff.pos);
             if (outBuff.pos) {
-                size_t const sizeCheck = fwrite(ress.dstBuffer, 1, outBuff.pos, dstFile);
-                if (sizeCheck != outBuff.pos)
-                    EXM_THROW(25, "Write error : %s (cannot write compressed block)",
-                                    strerror(errno));
+                writeJob->usedBufferSize = outBuff.pos;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 compressedfilesize += outBuff.pos;
             }
 
-            /* display notification; and adapt compression level */
-            if (READY_FOR_UPDATE()) {
+            /* adaptive mode : statistics measurement and speed correction */
+            if (prefs->adaptiveMode && UTIL_clockSpanMicro(lastAdaptTime) > adaptEveryMicro) {
+                ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx);
+
+                lastAdaptTime = UTIL_getTime();
+
+                /* check output speed */
+                if (zfp.currentJobID > 1) {  /* only possible if nbWorkers >= 1 */
+
+                    unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced;
+                    unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed;
+                    assert(zfp.produced >= previous_zfp_update.produced);
+                    assert(prefs->nbWorkers >= 1);
+
+                    /* test if compression is blocked
+                        * either because output is slow and all buffers are full
+                        * or because input is slow and no job can start while waiting for at least one buffer to be filled.
+                        * note : exclude starting part, since currentJobID > 1 */
+                    if ( (zfp.consumed == previous_zfp_update.consumed)   /* no data compressed : no data available, or no more buffer to compress to, OR compression is really slow (compression of a single block is slower than update rate)*/
+                        && (zfp.nbActiveWorkers == 0)                       /* confirmed : no compression ongoing */
+                        ) {
+                        DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n")
+                        speedChange = slower;
+                    }
+
+                    previous_zfp_update = zfp;
+
+                    if ( (newlyProduced > (newlyFlushed * 9 / 8))   /* compression produces more data than output can flush (though production can be spiky, due to work unit : (N==4)*block sizes) */
+                        && (flushWaiting == 0)                        /* flush speed was never slowed by lack of production, so it's operating at max capacity */
+                        ) {
+                        DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed);
+                        speedChange = slower;
+                    }
+                    flushWaiting = 0;
+                }
+
+                /* course correct only if there is at least one new job completed */
+                if (zfp.currentJobID > lastJobID) {
+                    DISPLAYLEVEL(6, "compression level adaptation check \n")
+
+                    /* check input speed */
+                    if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) {   /* warm up period, to fill all workers */
+                        if (inputBlocked <= 0) {
+                            DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n");
+                            speedChange = slower;
+                        } else if (speedChange == noChange) {
+                            unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested;
+                            unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed;
+                            unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced;
+                            unsigned long long newlyFlushed  = zfp.flushed  - previous_zfp_correction.flushed;
+                            previous_zfp_correction = zfp;
+                            assert(inputPresented > 0);
+                            DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n",
+                                            inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100,
+                                            (unsigned)newlyIngested, (unsigned)newlyConsumed,
+                                            (unsigned)newlyFlushed, (unsigned)newlyProduced);
+                            if ( (inputBlocked > inputPresented / 8)     /* input is waiting often, because input buffers is full : compression or output too slow */
+                                && (newlyFlushed * 33 / 32 > newlyProduced)  /* flush everything that is produced */
+                                && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */
+                            ) {
+                                DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n",
+                                                newlyIngested, newlyConsumed, newlyProduced, newlyFlushed);
+                                speedChange = faster;
+                            }
+                        }
+                        inputBlocked = 0;
+                        inputPresented = 0;
+                    }
+
+                    if (speedChange == slower) {
+                        DISPLAYLEVEL(6, "slower speed , higher compression \n")
+                        compressionLevel ++;
+                        if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel();
+                        if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel;
+                        compressionLevel += (compressionLevel == 0);   /* skip 0 */
+                        ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
+                    }
+                    if (speedChange == faster) {
+                        DISPLAYLEVEL(6, "faster speed , lighter compression \n")
+                        compressionLevel --;
+                        if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel;
+                        compressionLevel -= (compressionLevel == 0);   /* skip 0 */
+                        ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
+                    }
+                    speedChange = noChange;
+
+                    lastJobID = zfp.currentJobID;
+                }  /* if (zfp.currentJobID > lastJobID) */
+            } /* if (prefs->adaptiveMode && UTIL_clockSpanMicro(lastAdaptTime) > adaptEveryMicro) */
+
+            /* display notification */
+            if (SHOULD_DISPLAY_PROGRESS() && READY_FOR_UPDATE()) {
                 ZSTD_frameProgression const zfp = ZSTD_getFrameProgression(ress.cctx);
                 double const cShare = (double)zfp.produced / (double)(zfp.consumed + !zfp.consumed/*avoid div0*/) * 100;
                 UTIL_HumanReadableSize_t const buffered_hrs = UTIL_makeHumanReadableSize(zfp.ingested - zfp.consumed);
                 UTIL_HumanReadableSize_t const consumed_hrs = UTIL_makeHumanReadableSize(zfp.consumed);
                 UTIL_HumanReadableSize_t const produced_hrs = UTIL_makeHumanReadableSize(zfp.produced);
 
+                DELAY_NEXT_UPDATE();
+
                 /* display progress notifications */
+                DISPLAY_PROGRESS("\r%79s\r", "");    /* Clear out the current displayed line */
                 if (g_display_prefs.displayLevel >= 3) {
-                    DISPLAYUPDATE(3, "\r(L%i) Buffered :%6.*f%4s - Consumed :%6.*f%4s - Compressed :%6.*f%4s => %.2f%% ",
-                                compressionLevel,
-                                buffered_hrs.precision, buffered_hrs.value, buffered_hrs.suffix,
-                                consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix,
-                                produced_hrs.precision, produced_hrs.value, produced_hrs.suffix,
-                                cShare );
-                } else if (g_display_prefs.displayLevel >= 2 || g_display_prefs.progressSetting == FIO_ps_always) {
+                    /* Verbose progress update */
+                    DISPLAY_PROGRESS(
+                            "(L%i) Buffered:%5.*f%s - Consumed:%5.*f%s - Compressed:%5.*f%s => %.2f%% ",
+                            compressionLevel,
+                            buffered_hrs.precision, buffered_hrs.value, buffered_hrs.suffix,
+                            consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix,
+                            produced_hrs.precision, produced_hrs.value, produced_hrs.suffix,
+                            cShare );
+                } else {
                     /* Require level 2 or forcibly displayed progress counter for summarized updates */
-                    DISPLAYLEVEL(1, "\r%79s\r", "");    /* Clear out the current displayed line */
                     if (fCtx->nbFilesTotal > 1) {
                         size_t srcFileNameSize = strlen(srcFileName);
                         /* Ensure that the string we print is roughly the same size each time */
                         if (srcFileNameSize > 18) {
                             const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
-                            DISPLAYLEVEL(1, "Compress: %u/%u files. Current: ...%s ",
+                            DISPLAY_PROGRESS("Compress: %u/%u files. Current: ...%s ",
                                         fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName);
                         } else {
-                            DISPLAYLEVEL(1, "Compress: %u/%u files. Current: %*s ",
+                            DISPLAY_PROGRESS("Compress: %u/%u files. Current: %*s ",
                                         fCtx->currFileIdx+1, fCtx->nbFilesTotal, (int)(18-srcFileNameSize), srcFileName);
                         }
                     }
-                    DISPLAYLEVEL(1, "Read:%6.*f%4s ", consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix);
+                    DISPLAY_PROGRESS("Read:%6.*f%4s ", consumed_hrs.precision, consumed_hrs.value, consumed_hrs.suffix);
                     if (fileSize != UTIL_FILESIZE_UNKNOWN)
-                        DISPLAYLEVEL(2, "/%6.*f%4s", file_hrs.precision, file_hrs.value, file_hrs.suffix);
-                    DISPLAYLEVEL(1, " ==> %2.f%%", cShare);
-                    DELAY_NEXT_UPDATE();
+                        DISPLAY_PROGRESS("/%6.*f%4s", file_hrs.precision, file_hrs.value, file_hrs.suffix);
+                    DISPLAY_PROGRESS(" ==> %2.f%%", cShare);
                 }
-
-                /* adaptive mode : statistics measurement and speed correction */
-                if (prefs->adaptiveMode) {
-
-                    /* check output speed */
-                    if (zfp.currentJobID > 1) {  /* only possible if nbWorkers >= 1 */
-
-                        unsigned long long newlyProduced = zfp.produced - previous_zfp_update.produced;
-                        unsigned long long newlyFlushed = zfp.flushed - previous_zfp_update.flushed;
-                        assert(zfp.produced >= previous_zfp_update.produced);
-                        assert(prefs->nbWorkers >= 1);
-
-                        /* test if compression is blocked
-                         * either because output is slow and all buffers are full
-                         * or because input is slow and no job can start while waiting for at least one buffer to be filled.
-                         * note : exclude starting part, since currentJobID > 1 */
-                        if ( (zfp.consumed == previous_zfp_update.consumed)   /* no data compressed : no data available, or no more buffer to compress to, OR compression is really slow (compression of a single block is slower than update rate)*/
-                          && (zfp.nbActiveWorkers == 0)                       /* confirmed : no compression ongoing */
-                          ) {
-                            DISPLAYLEVEL(6, "all buffers full : compression stopped => slow down \n")
-                            speedChange = slower;
-                        }
-
-                        previous_zfp_update = zfp;
-
-                        if ( (newlyProduced > (newlyFlushed * 9 / 8))   /* compression produces more data than output can flush (though production can be spiky, due to work unit : (N==4)*block sizes) */
-                          && (flushWaiting == 0)                        /* flush speed was never slowed by lack of production, so it's operating at max capacity */
-                          ) {
-                            DISPLAYLEVEL(6, "compression faster than flush (%llu > %llu), and flushed was never slowed down by lack of production => slow down \n", newlyProduced, newlyFlushed);
-                            speedChange = slower;
-                        }
-                        flushWaiting = 0;
-                    }
-
-                    /* course correct only if there is at least one new job completed */
-                    if (zfp.currentJobID > lastJobID) {
-                        DISPLAYLEVEL(6, "compression level adaptation check \n")
-
-                        /* check input speed */
-                        if (zfp.currentJobID > (unsigned)(prefs->nbWorkers+1)) {   /* warm up period, to fill all workers */
-                            if (inputBlocked <= 0) {
-                                DISPLAYLEVEL(6, "input is never blocked => input is slower than ingestion \n");
-                                speedChange = slower;
-                            } else if (speedChange == noChange) {
-                                unsigned long long newlyIngested = zfp.ingested - previous_zfp_correction.ingested;
-                                unsigned long long newlyConsumed = zfp.consumed - previous_zfp_correction.consumed;
-                                unsigned long long newlyProduced = zfp.produced - previous_zfp_correction.produced;
-                                unsigned long long newlyFlushed  = zfp.flushed  - previous_zfp_correction.flushed;
-                                previous_zfp_correction = zfp;
-                                assert(inputPresented > 0);
-                                DISPLAYLEVEL(6, "input blocked %u/%u(%.2f) - ingested:%u vs %u:consumed - flushed:%u vs %u:produced \n",
-                                                inputBlocked, inputPresented, (double)inputBlocked/inputPresented*100,
-                                                (unsigned)newlyIngested, (unsigned)newlyConsumed,
-                                                (unsigned)newlyFlushed, (unsigned)newlyProduced);
-                                if ( (inputBlocked > inputPresented / 8)     /* input is waiting often, because input buffers is full : compression or output too slow */
-                                  && (newlyFlushed * 33 / 32 > newlyProduced)  /* flush everything that is produced */
-                                  && (newlyIngested * 33 / 32 > newlyConsumed) /* input speed as fast or faster than compression speed */
-                                ) {
-                                    DISPLAYLEVEL(6, "recommend faster as in(%llu) >= (%llu)comp(%llu) <= out(%llu) \n",
-                                                    newlyIngested, newlyConsumed, newlyProduced, newlyFlushed);
-                                    speedChange = faster;
-                                }
-                            }
-                            inputBlocked = 0;
-                            inputPresented = 0;
-                        }
-
-                        if (speedChange == slower) {
-                            DISPLAYLEVEL(6, "slower speed , higher compression \n")
-                            compressionLevel ++;
-                            if (compressionLevel > ZSTD_maxCLevel()) compressionLevel = ZSTD_maxCLevel();
-                            if (compressionLevel > prefs->maxAdaptLevel) compressionLevel = prefs->maxAdaptLevel;
-                            compressionLevel += (compressionLevel == 0);   /* skip 0 */
-                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
-                        }
-                        if (speedChange == faster) {
-                            DISPLAYLEVEL(6, "faster speed , lighter compression \n")
-                            compressionLevel --;
-                            if (compressionLevel < prefs->minAdaptLevel) compressionLevel = prefs->minAdaptLevel;
-                            compressionLevel -= (compressionLevel == 0);   /* skip 0 */
-                            ZSTD_CCtx_setParameter(ress.cctx, ZSTD_c_compressionLevel, compressionLevel);
-                        }
-                        speedChange = noChange;
-
-                        lastJobID = zfp.currentJobID;
-                    }  /* if (zfp.currentJobID > lastJobID) */
-                }  /* if (g_adaptiveMode) */
-            }  /* if (READY_FOR_UPDATE()) */
+            }  /* if (SHOULD_DISPLAY_PROGRESS() && READY_FOR_UPDATE()) */
         }  /* while ((inBuff.pos != inBuff.size) */
     } while (directive != ZSTD_e_end);
 
-    if (ferror(srcFile)) {
-        EXM_THROW(26, "Read error : I/O error");
-    }
     if (fileSize != UTIL_FILESIZE_UNKNOWN && *readsize != fileSize) {
         EXM_THROW(27, "Read error : Incomplete read : %llu / %llu B",
                 (unsigned long long)*readsize, (unsigned long long)fileSize);
     }
 
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ressPtr->writeCtx);
+
     return compressedfilesize;
 }
 
@@ -1620,20 +1785,18 @@ FIO_compressFilename_internal(FIO_ctx_t* const fCtx,
     /* Status */
     fCtx->totalBytesInput += (size_t)readsize;
     fCtx->totalBytesOutput += (size_t)compressedfilesize;
-    DISPLAYLEVEL(2, "\r%79s\r", "");
-    if (g_display_prefs.displayLevel >= 2 &&
-        !fCtx->hasStdoutOutput &&
-        (g_display_prefs.displayLevel >= 3 || fCtx->nbFilesTotal <= 1)) {
+    DISPLAY_PROGRESS("\r%79s\r", "");
+    if (FIO_shouldDisplayFileSummary(fCtx)) {
         UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) readsize);
         UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) compressedfilesize);
         if (readsize == 0) {
-            DISPLAYLEVEL(2,"%-20s :  (%6.*f%4s => %6.*f%4s, %s) \n",
+            DISPLAY_SUMMARY("%-20s :  (%6.*f%s => %6.*f%s, %s) \n",
                 srcFileName,
                 hr_isize.precision, hr_isize.value, hr_isize.suffix,
                 hr_osize.precision, hr_osize.value, hr_osize.suffix,
                 dstFileName);
         } else {
-            DISPLAYLEVEL(2,"%-20s :%6.2f%%   (%6.*f%4s => %6.*f%4s, %s) \n",
+            DISPLAY_SUMMARY("%-20s :%6.2f%%   (%6.*f%s => %6.*f%s, %s) \n",
                 srcFileName,
                 (double)compressedfilesize / (double)readsize * 100,
                 hr_isize.precision, hr_isize.value, hr_isize.suffix,
@@ -1656,7 +1819,7 @@ FIO_compressFilename_internal(FIO_ctx_t* const fCtx,
 
 
 /*! FIO_compressFilename_dstFile() :
- *  open dstFileName, or pass-through if ress.dstFile != NULL,
+ *  open dstFileName, or pass-through if ress.file != NULL,
  *  then start compression with FIO_compressFilename_internal().
  *  Manages source removal (--rm) and file permissions transfer.
  *  note : ress.srcFile must be != NULL,
@@ -1669,27 +1832,31 @@ static int FIO_compressFilename_dstFile(FIO_ctx_t* const fCtx,
                                         cRess_t ress,
                                         const char* dstFileName,
                                         const char* srcFileName,
+                                        const stat_t* srcFileStat,
                                         int compressionLevel)
 {
     int closeDstFile = 0;
     int result;
-    stat_t statbuf;
-    int transferMTime = 0;
-    assert(ress.srcFile != NULL);
-    if (ress.dstFile == NULL) {
-        int dstFilePermissions = DEFAULT_FILE_PERMISSIONS;
+    int transferStat = 0;
+    int dstFd = -1;
+
+    assert(AIO_ReadPool_getFile(ress.readCtx) != NULL);
+    if (AIO_WritePool_getFile(ress.writeCtx) == NULL) {
+        int dstFileInitialPermissions = DEFAULT_FILE_PERMISSIONS;
         if ( strcmp (srcFileName, stdinmark)
           && strcmp (dstFileName, stdoutmark)
-          && UTIL_stat(srcFileName, &statbuf)
-          && UTIL_isRegularFileStat(&statbuf) ) {
-            dstFilePermissions = statbuf.st_mode;
-            transferMTime = 1;
+          && UTIL_isRegularFileStat(srcFileStat) ) {
+            transferStat = 1;
+            dstFileInitialPermissions = TEMPORARY_FILE_PERMISSIONS;
         }
 
         closeDstFile = 1;
         DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: opening dst: %s \n", dstFileName);
-        ress.dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
-        if (ress.dstFile==NULL) return 1;  /* could not open dstFileName */
+        {   FILE *dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFileInitialPermissions);
+            if (dstFile==NULL) return 1;  /* could not open dstFileName */
+            dstFd = fileno(dstFile);
+            AIO_WritePool_setFile(ress.writeCtx, dstFile);
+        }
         /* Must only be added after FIO_openDstFile() succeeds.
          * Otherwise we may delete the destination file if it already exists,
          * and the user presses Ctrl-C when asked if they wish to overwrite.
@@ -1700,19 +1867,22 @@ static int FIO_compressFilename_dstFile(FIO_ctx_t* const fCtx,
     result = FIO_compressFilename_internal(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
 
     if (closeDstFile) {
-        FILE* const dstFile = ress.dstFile;
-        ress.dstFile = NULL;
-
         clearHandler();
 
+        if (transferStat) {
+            UTIL_setFDStat(dstFd, dstFileName, srcFileStat);
+        }
+
         DISPLAYLEVEL(6, "FIO_compressFilename_dstFile: closing dst: %s \n", dstFileName);
-        if (fclose(dstFile)) { /* error closing dstFile */
+        if (AIO_WritePool_closeFile(ress.writeCtx)) { /* error closing file */
             DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
             result=1;
         }
-        if (transferMTime) {
-            UTIL_utime(dstFileName, &statbuf);
+
+        if (transferStat) {
+            UTIL_utime(dstFileName, srcFileStat);
         }
+
         if ( (result != 0)  /* operation failure */
           && strcmp(dstFileName, stdoutmark)  /* special case : don't remove() stdout */
           ) {
@@ -1736,6 +1906,110 @@ static const char *compressedFileExtensions[] = {
     TXZ_EXTENSION,
     LZ4_EXTENSION,
     TLZ4_EXTENSION,
+    ".7z",
+    ".aa3",
+    ".aac",
+    ".aar",
+    ".ace",
+    ".alac",
+    ".ape",
+    ".apk",
+    ".apng",
+    ".arc",
+    ".archive",
+    ".arj",
+    ".ark",
+    ".asf",
+    ".avi",
+    ".avif",
+    ".ba",
+    ".br",
+    ".bz2",
+    ".cab",
+    ".cdx",
+    ".chm",
+    ".cr2",
+    ".divx",
+    ".dmg",
+    ".dng",
+    ".docm",
+    ".docx",
+    ".dotm",
+    ".dotx",
+    ".dsft",
+    ".ear",
+    ".eftx",
+    ".emz",
+    ".eot",
+    ".epub",
+    ".f4v",
+    ".flac",
+    ".flv",
+    ".gho",
+    ".gif",
+    ".gifv",
+    ".gnp",
+    ".iso",
+    ".jar",
+    ".jpeg",
+    ".jpg",
+    ".jxl",
+    ".lz",
+    ".lzh",
+    ".m4a",
+    ".m4v",
+    ".mkv",
+    ".mov",
+    ".mp2",
+    ".mp3",
+    ".mp4",
+    ".mpa",
+    ".mpc",
+    ".mpe",
+    ".mpeg",
+    ".mpg",
+    ".mpl",
+    ".mpv",
+    ".msi",
+    ".odp",
+    ".ods",
+    ".odt",
+    ".ogg",
+    ".ogv",
+    ".otp",
+    ".ots",
+    ".ott",
+    ".pea",
+    ".png",
+    ".pptx",
+    ".qt",
+    ".rar",
+    ".s7z",
+    ".sfx",
+    ".sit",
+    ".sitx",
+    ".sqx",
+    ".svgz",
+    ".swf",
+    ".tbz2",
+    ".tib",
+    ".tlz",
+    ".vob",
+    ".war",
+    ".webm",
+    ".webp",
+    ".wma",
+    ".wmv",
+    ".woff",
+    ".woff2",
+    ".wvl",
+    ".xlsx",
+    ".xpi",
+    ".xps",
+    ".zip",
+    ".zipx",
+    ".zoo",
+    ".zpaq",
     NULL
 };
 
@@ -1752,18 +2026,27 @@ FIO_compressFilename_srcFile(FIO_ctx_t* const fCtx,
                              int compressionLevel)
 {
     int result;
+    FILE* srcFile;
+    stat_t srcFileStat;
+    U64 fileSize = UTIL_FILESIZE_UNKNOWN;
     DISPLAYLEVEL(6, "FIO_compressFilename_srcFile: %s \n", srcFileName);
 
-    /* ensure src is not a directory */
-    if (UTIL_isDirectory(srcFileName)) {
-        DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
-        return 1;
-    }
+    if (strcmp(srcFileName, stdinmark)) {
+        if (UTIL_stat(srcFileName, &srcFileStat)) {
+            /* failure to stat at all is handled during opening */
 
-    /* ensure src is not the same as dict (if present) */
-    if (ress.dictFileName != NULL && UTIL_isSameFile(srcFileName, ress.dictFileName)) {
-        DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName);
-        return 1;
+            /* ensure src is not a directory */
+            if (UTIL_isDirectoryStat(&srcFileStat)) {
+                DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
+                return 1;
+            }
+
+            /* ensure src is not the same as dict (if present) */
+            if (ress.dictFileName != NULL && UTIL_isSameFileStat(srcFileName, ress.dictFileName, &srcFileStat, &ress.dictFileStat)) {
+                DISPLAYLEVEL(1, "zstd: cannot use %s as an input file and dictionary \n", srcFileName);
+                return 1;
+            }
+        }
     }
 
     /* Check if "srcFile" is compressed. Only done if --exclude-compressed flag is used
@@ -1775,16 +2058,30 @@ FIO_compressFilename_srcFile(FIO_ctx_t* const fCtx,
         return 0;
     }
 
-    ress.srcFile = FIO_openSrcFile(prefs, srcFileName);
-    if (ress.srcFile == NULL) return 1;   /* srcFile could not be opened */
+    srcFile = FIO_openSrcFile(prefs, srcFileName, &srcFileStat);
+    if (srcFile == NULL) return 1;   /* srcFile could not be opened */
+
+    /* Don't use AsyncIO for small files */
+    if (strcmp(srcFileName, stdinmark)) /* Stdin doesn't have stats */
+        fileSize = UTIL_getFileSizeStat(&srcFileStat);
+    if(fileSize != UTIL_FILESIZE_UNKNOWN && fileSize < ZSTD_BLOCKSIZE_MAX * 3) {
+        AIO_ReadPool_setAsync(ress.readCtx, 0);
+        AIO_WritePool_setAsync(ress.writeCtx, 0);
+    } else {
+        AIO_ReadPool_setAsync(ress.readCtx, 1);
+        AIO_WritePool_setAsync(ress.writeCtx, 1);
+    }
 
-    result = FIO_compressFilename_dstFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
+    AIO_ReadPool_setFile(ress.readCtx, srcFile);
+    result = FIO_compressFilename_dstFile(
+            fCtx, prefs, ress,
+            dstFileName, srcFileName,
+            &srcFileStat, compressionLevel);
+    AIO_ReadPool_closeFile(ress.readCtx);
 
-    fclose(ress.srcFile);
-    ress.srcFile = NULL;
-    if ( prefs->removeSrcFile   /* --rm */
-      && result == 0       /* success */
-      && strcmp(srcFileName, stdinmark)   /* exception : don't erase stdin */
+    if ( prefs->removeSrcFile  /* --rm */
+      && result == 0           /* success */
+      && strcmp(srcFileName, stdinmark)  /* exception : don't erase stdin */
       ) {
         /* We must clear the handler, since after this point calling it would
          * delete both the source and destination files.
@@ -1796,16 +2093,18 @@ FIO_compressFilename_srcFile(FIO_ctx_t* const fCtx,
     return result;
 }
 
-static const char* checked_index(const char* options[], size_t length, size_t index) {
+static const char*
+checked_index(const char* options[], size_t length, size_t index) {
     assert(index < length);
-    // Necessary to avoid warnings since -O3 will omit the above `assert`
+    /* Necessary to avoid warnings since -O3 will omit the above `assert` */
     (void) length;
     return options[index];
 }
 
-#define INDEX(options, index) checked_index((options), sizeof(options)  / sizeof(char*), (index))
+#define INDEX(options, index) checked_index((options), sizeof(options)  / sizeof(char*), (size_t)(index))
 
-void FIO_displayCompressionParameters(const FIO_prefs_t* prefs) {
+void FIO_displayCompressionParameters(const FIO_prefs_t* prefs)
+{
     static const char* formatOptions[5] = {ZSTD_EXTENSION, GZ_EXTENSION, XZ_EXTENSION,
                                            LZMA_EXTENSION, LZ4_EXTENSION};
     static const char* sparseOptions[3] = {" --no-sparse", "", " --sparse"};
@@ -1844,7 +2143,7 @@ int FIO_compressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, const
                          const char* srcFileName, const char* dictFileName,
                          int compressionLevel, ZSTD_compressionParameters comprParams)
 {
-    cRess_t const ress = FIO_createCResources(prefs, dictFileName, UTIL_getFileSize(srcFileName), compressionLevel, comprParams);
+    cRess_t ress = FIO_createCResources(prefs, dictFileName, UTIL_getFileSize(srcFileName), compressionLevel, comprParams);
     int const result = FIO_compressFilename_srcFile(fCtx, prefs, ress, dstFileName, srcFileName, compressionLevel);
 
 #define DISPLAY_LEVEL_DEFAULT 2
@@ -1866,6 +2165,11 @@ FIO_determineCompressedName(const char* srcFileName, const char* outDirName, con
     char* outDirFilename = NULL;
     size_t sfnSize = strlen(srcFileName);
     size_t const srcSuffixLen = strlen(suffix);
+
+    if(!strcmp(srcFileName, stdinmark)) {
+        return stdoutmark;
+    }
+
     if (outDirName) {
         outDirFilename = FIO_createFilename_fromOutDir(srcFileName, outDirName, srcSuffixLen);
         sfnSize = strlen(outDirFilename);
@@ -1928,23 +2232,24 @@ int FIO_compressMultipleFilenames(FIO_ctx_t* const fCtx,
     /* init */
     assert(outFileName != NULL || suffix != NULL);
     if (outFileName != NULL) {   /* output into a single destination (stdout typically) */
-        if (FIO_removeMultiFilesWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
+        FILE *dstFile;
+        if (FIO_multiFilesConcatWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
             FIO_freeCResources(&ress);
             return 1;
         }
-        ress.dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
-        if (ress.dstFile == NULL) {  /* could not open outFileName */
+        dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
+        if (dstFile == NULL) {  /* could not open outFileName */
             error = 1;
         } else {
+            AIO_WritePool_setFile(ress.writeCtx, dstFile);
             for (; fCtx->currFileIdx < fCtx->nbFilesTotal; ++fCtx->currFileIdx) {
                 status = FIO_compressFilename_srcFile(fCtx, prefs, ress, outFileName, inFileNamesTable[fCtx->currFileIdx], compressionLevel);
                 if (!status) fCtx->nbFilesProcessed++;
                 error |= status;
             }
-            if (fclose(ress.dstFile))
+            if (AIO_WritePool_closeFile(ress.writeCtx))
                 EXM_THROW(29, "Write error (%s) : cannot properly close %s",
                             strerror(errno), outFileName);
-            ress.dstFile = NULL;
         }
     } else {
         if (outMirroredRootDirName)
@@ -1975,16 +2280,23 @@ int FIO_compressMultipleFilenames(FIO_ctx_t* const fCtx,
             FIO_checkFilenameCollisions(inFileNamesTable , (unsigned)fCtx->nbFilesTotal);
     }
 
-    if (fCtx->nbFilesProcessed >= 1 && fCtx->nbFilesTotal > 1 && fCtx->totalBytesInput != 0) {
+    if (FIO_shouldDisplayMultipleFileSummary(fCtx)) {
         UTIL_HumanReadableSize_t hr_isize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesInput);
         UTIL_HumanReadableSize_t hr_osize = UTIL_makeHumanReadableSize((U64) fCtx->totalBytesOutput);
 
-        DISPLAYLEVEL(2, "\r%79s\r", "");
-        DISPLAYLEVEL(2, "%3d files compressed :%.2f%%   (%6.*f%4s => %6.*f%4s)\n",
-                        fCtx->nbFilesProcessed,
-                        (double)fCtx->totalBytesOutput/((double)fCtx->totalBytesInput)*100,
-                        hr_isize.precision, hr_isize.value, hr_isize.suffix,
-                        hr_osize.precision, hr_osize.value, hr_osize.suffix);
+        DISPLAY_PROGRESS("\r%79s\r", "");
+        if (fCtx->totalBytesInput == 0) {
+            DISPLAY_SUMMARY("%3d files compressed : (%6.*f%4s => %6.*f%4s)\n",
+                            fCtx->nbFilesProcessed,
+                            hr_isize.precision, hr_isize.value, hr_isize.suffix,
+                            hr_osize.precision, hr_osize.value, hr_osize.suffix);
+        } else {
+            DISPLAY_SUMMARY("%3d files compressed : %.2f%% (%6.*f%4s => %6.*f%4s)\n",
+                            fCtx->nbFilesProcessed,
+                            (double)fCtx->totalBytesOutput/((double)fCtx->totalBytesInput)*100,
+                            hr_isize.precision, hr_isize.value, hr_isize.suffix,
+                            hr_osize.precision, hr_osize.value, hr_osize.suffix);
+        }
     }
 
     FIO_freeCResources(&ress);
@@ -2001,22 +2313,28 @@ int FIO_compressMultipleFilenames(FIO_ctx_t* const fCtx,
  *  Decompression
  ***************************************************************************/
 typedef struct {
-    void*  srcBuffer;
-    size_t srcBufferSize;
-    size_t srcBufferLoaded;
-    void*  dstBuffer;
-    size_t dstBufferSize;
+    FIO_Dict_t dict;
     ZSTD_DStream* dctx;
-    FILE*  dstFile;
+    WritePoolCtx_t *writeCtx;
+    ReadPoolCtx_t *readCtx;
 } dRess_t;
 
 static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFileName)
 {
+    int useMMap = prefs->mmapDict == ZSTD_ps_enable;
+    int forceNoUseMMap = prefs->mmapDict == ZSTD_ps_disable;
+    stat_t statbuf;
     dRess_t ress;
+    memset(&statbuf, 0, sizeof(statbuf));
     memset(&ress, 0, sizeof(ress));
 
-    if (prefs->patchFromMode)
-        FIO_adjustMemLimitForPatchFromMode(prefs, UTIL_getFileSize(dictFileName), 0 /* just use the dict size */);
+    FIO_getDictFileStat(dictFileName, &statbuf);
+
+    if (prefs->patchFromMode){
+        U64 const dictSize = UTIL_getFileSizeStat(&statbuf);
+        useMMap |= dictSize > prefs->memLimit;
+        FIO_adjustMemLimitForPatchFromMode(prefs, dictSize, 0 /* just use the dict size */);
+    }
 
     /* Allocation */
     ress.dctx = ZSTD_createDStream();
@@ -2025,159 +2343,54 @@ static dRess_t FIO_createDResources(FIO_prefs_t* const prefs, const char* dictFi
     CHECK( ZSTD_DCtx_setMaxWindowSize(ress.dctx, prefs->memLimit) );
     CHECK( ZSTD_DCtx_setParameter(ress.dctx, ZSTD_d_forceIgnoreChecksum, !prefs->checksumFlag));
 
-    ress.srcBufferSize = ZSTD_DStreamInSize();
-    ress.srcBuffer = malloc(ress.srcBufferSize);
-    ress.dstBufferSize = ZSTD_DStreamOutSize();
-    ress.dstBuffer = malloc(ress.dstBufferSize);
-    if (!ress.srcBuffer || !ress.dstBuffer)
-        EXM_THROW(61, "Allocation error : not enough memory");
-
     /* dictionary */
-    {   void* dictBuffer;
-        size_t const dictBufferSize = FIO_createDictBuffer(&dictBuffer, dictFileName, prefs);
-        CHECK( ZSTD_initDStream_usingDict(ress.dctx, dictBuffer, dictBufferSize) );
-        free(dictBuffer);
-    }
-
-    return ress;
-}
-
-static void FIO_freeDResources(dRess_t ress)
-{
-    CHECK( ZSTD_freeDStream(ress.dctx) );
-    free(ress.srcBuffer);
-    free(ress.dstBuffer);
-}
-
-
-/** FIO_fwriteSparse() :
-*  @return : storedSkips,
-*            argument for next call to FIO_fwriteSparse() or FIO_fwriteSparseEnd() */
-static unsigned
-FIO_fwriteSparse(FILE* file,
-                 const void* buffer, size_t bufferSize,
-                 const FIO_prefs_t* const prefs,
-                 unsigned storedSkips)
-{
-    const size_t* const bufferT = (const size_t*)buffer;   /* Buffer is supposed malloc'ed, hence aligned on size_t */
-    size_t bufferSizeT = bufferSize / sizeof(size_t);
-    const size_t* const bufferTEnd = bufferT + bufferSizeT;
-    const size_t* ptrT = bufferT;
-    static const size_t segmentSizeT = (32 KB) / sizeof(size_t);   /* check every 32 KB */
-
-    if (prefs->testMode) return 0;  /* do not output anything in test mode */
-
-    if (!prefs->sparseFileSupport) {  /* normal write */
-        size_t const sizeCheck = fwrite(buffer, 1, bufferSize, file);
-        if (sizeCheck != bufferSize)
-            EXM_THROW(70, "Write error : cannot write decoded block : %s",
-                            strerror(errno));
-        return 0;
-    }
-
-    /* avoid int overflow */
-    if (storedSkips > 1 GB) {
-        if (LONG_SEEK(file, 1 GB, SEEK_CUR) != 0)
-            EXM_THROW(91, "1 GB skip error (sparse file support)");
-        storedSkips -= 1 GB;
-    }
-
-    while (ptrT < bufferTEnd) {
-        size_t nb0T;
-
-        /* adjust last segment if < 32 KB */
-        size_t seg0SizeT = segmentSizeT;
-        if (seg0SizeT > bufferSizeT) seg0SizeT = bufferSizeT;
-        bufferSizeT -= seg0SizeT;
+    {
+        FIO_dictBufferType_t dictBufferType = (useMMap && !forceNoUseMMap) ? FIO_mmapDict : FIO_mallocDict;
+        FIO_initDict(&ress.dict, dictFileName, prefs, &statbuf, dictBufferType);
 
-        /* count leading zeroes */
-        for (nb0T=0; (nb0T < seg0SizeT) && (ptrT[nb0T] == 0); nb0T++) ;
-        storedSkips += (unsigned)(nb0T * sizeof(size_t));
+        CHECK(ZSTD_DCtx_reset(ress.dctx, ZSTD_reset_session_only) );
 
-        if (nb0T != seg0SizeT) {   /* not all 0s */
-            size_t const nbNon0ST = seg0SizeT - nb0T;
-            /* skip leading zeros */
-            if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
-                EXM_THROW(92, "Sparse skip error ; try --no-sparse");
-            storedSkips = 0;
-            /* write the rest */
-            if (fwrite(ptrT + nb0T, sizeof(size_t), nbNon0ST, file) != nbNon0ST)
-                EXM_THROW(93, "Write error : cannot write decoded block : %s",
-                            strerror(errno));
+        if (prefs->patchFromMode){
+            CHECK(ZSTD_DCtx_refPrefix(ress.dctx, ress.dict.dictBuffer, ress.dict.dictBufferSize));
+        } else {
+            CHECK(ZSTD_DCtx_loadDictionary_byReference(ress.dctx, ress.dict.dictBuffer, ress.dict.dictBufferSize));
         }
-        ptrT += seg0SizeT;
     }
 
-    {   static size_t const maskT = sizeof(size_t)-1;
-        if (bufferSize & maskT) {
-            /* size not multiple of sizeof(size_t) : implies end of block */
-            const char* const restStart = (const char*)bufferTEnd;
-            const char* restPtr = restStart;
-            const char* const restEnd = (const char*)buffer + bufferSize;
-            assert(restEnd > restStart && restEnd < restStart + sizeof(size_t));
-            for ( ; (restPtr < restEnd) && (*restPtr == 0); restPtr++) ;
-            storedSkips += (unsigned) (restPtr - restStart);
-            if (restPtr != restEnd) {
-                /* not all remaining bytes are 0 */
-                size_t const restSize = (size_t)(restEnd - restPtr);
-                if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
-                    EXM_THROW(92, "Sparse skip error ; try --no-sparse");
-                if (fwrite(restPtr, 1, restSize, file) != restSize)
-                    EXM_THROW(95, "Write error : cannot write end of decoded block : %s",
-                        strerror(errno));
-                storedSkips = 0;
-    }   }   }
-
-    return storedSkips;
+    ress.writeCtx = AIO_WritePool_create(prefs, ZSTD_DStreamOutSize());
+    ress.readCtx = AIO_ReadPool_create(prefs, ZSTD_DStreamInSize());
+    return ress;
 }
 
-static void
-FIO_fwriteSparseEnd(const FIO_prefs_t* const prefs, FILE* file, unsigned storedSkips)
+static void FIO_freeDResources(dRess_t ress)
 {
-    if (prefs->testMode) assert(storedSkips == 0);
-    if (storedSkips>0) {
-        assert(prefs->sparseFileSupport > 0);  /* storedSkips>0 implies sparse support is enabled */
-        (void)prefs;   /* assert can be disabled, in which case prefs becomes unused */
-        if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0)
-            EXM_THROW(69, "Final skip error (sparse file support)");
-        /* last zero must be explicitly written,
-         * so that skipped ones get implicitly translated as zero by FS */
-        {   const char lastZeroByte[1] = { 0 };
-            if (fwrite(lastZeroByte, 1, 1, file) != 1)
-                EXM_THROW(69, "Write error : cannot write last zero : %s", strerror(errno));
-    }   }
+    FIO_freeDict(&(ress.dict));
+    CHECK( ZSTD_freeDStream(ress.dctx) );
+    AIO_WritePool_free(ress.writeCtx);
+    AIO_ReadPool_free(ress.readCtx);
 }
 
-
-/** FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode
-    @return : 0 (no error) */
-static int FIO_passThrough(const FIO_prefs_t* const prefs,
-                           FILE* foutput, FILE* finput,
-                           void* buffer, size_t bufferSize,
-                           size_t alreadyLoaded)
+/* FIO_passThrough() : just copy input into output, for compatibility with gzip -df mode
+ * @return : 0 (no error) */
+static int FIO_passThrough(dRess_t *ress)
 {
-    size_t const blockSize = MIN(64 KB, bufferSize);
-    size_t readFromInput;
-    unsigned storedSkips = 0;
-
-    /* assumption : ress->srcBufferLoaded bytes already loaded and stored within buffer */
-    {   size_t const sizeCheck = fwrite(buffer, 1, alreadyLoaded, foutput);
-        if (sizeCheck != alreadyLoaded) {
-            DISPLAYLEVEL(1, "Pass-through write error : %s\n", strerror(errno));
-            return 1;
-    }   }
+    size_t const blockSize = MIN(MIN(64 KB, ZSTD_DStreamInSize()), ZSTD_DStreamOutSize());
+    IOJob_t *writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
+    AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
 
-    do {
-        readFromInput = fread(buffer, 1, blockSize, finput);
-        storedSkips = FIO_fwriteSparse(foutput, buffer, readFromInput, prefs, storedSkips);
-    } while (readFromInput == blockSize);
-    if (ferror(finput)) {
-        DISPLAYLEVEL(1, "Pass-through read error : %s\n", strerror(errno));
-        return 1;
+    while(ress->readCtx->srcBufferLoaded) {
+        size_t writeSize;
+        writeSize = MIN(blockSize, ress->readCtx->srcBufferLoaded);
+        assert(writeSize <= writeJob->bufferSize);
+        memcpy(writeJob->buffer, ress->readCtx->srcBuffer, writeSize);
+        writeJob->usedBufferSize = writeSize;
+        AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
+        AIO_ReadPool_consumeBytes(ress->readCtx, writeSize);
+        AIO_ReadPool_fillBuffer(ress->readCtx, blockSize);
     }
-    assert(feof(finput));
-
-    FIO_fwriteSparseEnd(prefs, foutput, storedSkips);
+    assert(ress->readCtx->reachedEof);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
     return 0;
 }
 
@@ -2186,16 +2399,17 @@ static int FIO_passThrough(const FIO_prefs_t* const prefs,
 static void
 FIO_zstdErrorHelp(const FIO_prefs_t* const prefs,
                   const dRess_t* ress,
-                  size_t err, const char* srcFileName)
+                  size_t err,
+                  const char* srcFileName)
 {
-    ZSTD_frameHeader header;
+    ZSTD_FrameHeader header;
 
     /* Help message only for one specific error */
     if (ZSTD_getErrorCode(err) != ZSTD_error_frameParameter_windowTooLarge)
         return;
 
     /* Try to decode the frame header */
-    err = ZSTD_getFrameHeader(&header, ress->srcBuffer, ress->srcBufferLoaded);
+    err = ZSTD_getFrameHeader(&header, ress->readCtx->srcBuffer, ress->readCtx->srcBufferLoaded);
     if (err == 0) {
         unsigned long long const windowSize = header.windowSize;
         unsigned const windowLog = FIO_highbit64(windowSize) + ((windowSize & (windowSize - 1)) != 0);
@@ -2218,83 +2432,72 @@ FIO_zstdErrorHelp(const FIO_prefs_t* const prefs,
  */
 #define FIO_ERROR_FRAME_DECODING   ((unsigned long long)(-2))
 static unsigned long long
-FIO_decompressZstdFrame(FIO_ctx_t* const fCtx, dRess_t* ress, FILE* finput,
+FIO_decompressZstdFrame(FIO_ctx_t* const fCtx, dRess_t* ress,
                         const FIO_prefs_t* const prefs,
                         const char* srcFileName,
                         U64 alreadyDecoded)  /* for multi-frames streams */
 {
     U64 frameSize = 0;
-    U32 storedSkips = 0;
+    const char* srcFName20 = srcFileName;
+    IOJob_t* writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
+    assert(writeJob);
 
-    /* display last 20 characters only */
+    /* display last 20 characters only when not --verbose */
     {   size_t const srcFileLength = strlen(srcFileName);
-        if (srcFileLength>20) srcFileName += srcFileLength-20;
+        if ((srcFileLength>20) && (g_display_prefs.displayLevel<3))
+            srcFName20 += srcFileLength-20;
     }
 
     ZSTD_DCtx_reset(ress->dctx, ZSTD_reset_session_only);
 
     /* Header loading : ensures ZSTD_getFrameHeader() will succeed */
-    {   size_t const toDecode = ZSTD_FRAMEHEADERSIZE_MAX;
-        if (ress->srcBufferLoaded < toDecode) {
-            size_t const toRead = toDecode - ress->srcBufferLoaded;
-            void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded;
-            ress->srcBufferLoaded += fread(startPosition, 1, toRead, finput);
-    }   }
+    AIO_ReadPool_fillBuffer(ress->readCtx, ZSTD_FRAMEHEADERSIZE_MAX);
 
     /* Main decompression Loop */
     while (1) {
-        ZSTD_inBuffer  inBuff = { ress->srcBuffer, ress->srcBufferLoaded, 0 };
-        ZSTD_outBuffer outBuff= { ress->dstBuffer, ress->dstBufferSize, 0 };
+        ZSTD_inBuffer  inBuff = setInBuffer( ress->readCtx->srcBuffer, ress->readCtx->srcBufferLoaded, 0 );
+        ZSTD_outBuffer outBuff= setOutBuffer( writeJob->buffer, writeJob->bufferSize, 0 );
         size_t const readSizeHint = ZSTD_decompressStream(ress->dctx, &outBuff, &inBuff);
-        const int displayLevel = (g_display_prefs.progressSetting == FIO_ps_always) ? 1 : 2;
         UTIL_HumanReadableSize_t const hrs = UTIL_makeHumanReadableSize(alreadyDecoded+frameSize);
         if (ZSTD_isError(readSizeHint)) {
             DISPLAYLEVEL(1, "%s : Decoding error (36) : %s \n",
                             srcFileName, ZSTD_getErrorName(readSizeHint));
             FIO_zstdErrorHelp(prefs, ress, readSizeHint, srcFileName);
+            AIO_WritePool_releaseIoJob(writeJob);
             return FIO_ERROR_FRAME_DECODING;
         }
 
         /* Write block */
-        storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, outBuff.pos, prefs, storedSkips);
+        writeJob->usedBufferSize = outBuff.pos;
+        AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
         frameSize += outBuff.pos;
         if (fCtx->nbFilesTotal > 1) {
-            size_t srcFileNameSize = strlen(srcFileName);
-            if (srcFileNameSize > 18) {
-                const char* truncatedSrcFileName = srcFileName + srcFileNameSize - 15;
-                DISPLAYUPDATE(displayLevel, "\rDecompress: %2u/%2u files. Current: ...%s : %.*f%s...    ",
-                                fCtx->currFileIdx+1, fCtx->nbFilesTotal, truncatedSrcFileName, hrs.precision, hrs.value, hrs.suffix);
-            } else {
-                DISPLAYUPDATE(displayLevel, "\rDecompress: %2u/%2u files. Current: %s : %.*f%s...    ",
-                            fCtx->currFileIdx+1, fCtx->nbFilesTotal, srcFileName, hrs.precision, hrs.value, hrs.suffix);
-            }
+            DISPLAYUPDATE_PROGRESS(
+                        "\rDecompress: %2u/%2u files. Current: %s : %.*f%s...    ",
+                        fCtx->currFileIdx+1, fCtx->nbFilesTotal, srcFName20, hrs.precision, hrs.value, hrs.suffix);
         } else {
-            DISPLAYUPDATE(displayLevel, "\r%-20.20s : %.*f%s...     ",
-                            srcFileName, hrs.precision, hrs.value, hrs.suffix);
+            DISPLAYUPDATE_PROGRESS("\r%-20.20s : %.*f%s...     ",
+                            srcFName20, hrs.precision, hrs.value, hrs.suffix);
         }
 
-        if (inBuff.pos > 0) {
-            memmove(ress->srcBuffer, (char*)ress->srcBuffer + inBuff.pos, inBuff.size - inBuff.pos);
-            ress->srcBufferLoaded -= inBuff.pos;
-        }
+        AIO_ReadPool_consumeBytes(ress->readCtx, inBuff.pos);
 
         if (readSizeHint == 0) break;   /* end of frame */
 
         /* Fill input buffer */
-        {   size_t const toDecode = MIN(readSizeHint, ress->srcBufferSize);  /* support large skippable frames */
-            if (ress->srcBufferLoaded < toDecode) {
-                size_t const toRead = toDecode - ress->srcBufferLoaded;   /* > 0 */
-                void* const startPosition = (char*)ress->srcBuffer + ress->srcBufferLoaded;
-                size_t const readSize = fread(startPosition, 1, toRead, finput);
+        {   size_t const toDecode = MIN(readSizeHint, ZSTD_DStreamInSize());  /* support large skippable frames */
+            if (ress->readCtx->srcBufferLoaded < toDecode) {
+                size_t const readSize = AIO_ReadPool_fillBuffer(ress->readCtx, toDecode);
                 if (readSize==0) {
                     DISPLAYLEVEL(1, "%s : Read error (39) : premature end \n",
-                                    srcFileName);
+                                 srcFileName);
+                    AIO_WritePool_releaseIoJob(writeJob);
                     return FIO_ERROR_FRAME_DECODING;
                 }
-                ress->srcBufferLoaded += readSize;
-    }   }   }
+            }   }   }
 
-    FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
 
     return frameSize;
 }
@@ -2302,37 +2505,36 @@ FIO_decompressZstdFrame(FIO_ctx_t* const fCtx, dRess_t* ress, FILE* finput,
 
 #ifdef ZSTD_GZDECOMPRESS
 static unsigned long long
-FIO_decompressGzFrame(dRess_t* ress, FILE* srcFile,
-                      const FIO_prefs_t* const prefs,
-                      const char* srcFileName)
+FIO_decompressGzFrame(dRess_t* ress, const char* srcFileName)
 {
     unsigned long long outFileSize = 0;
     z_stream strm;
     int flush = Z_NO_FLUSH;
     int decodingError = 0;
-    unsigned storedSkips = 0;
+    IOJob_t *writeJob = NULL;
 
     strm.zalloc = Z_NULL;
     strm.zfree = Z_NULL;
     strm.opaque = Z_NULL;
     strm.next_in = 0;
     strm.avail_in = 0;
-    /* see http://www.zlib.net/manual.html */
+    /* see https://www.zlib.net/manual.html */
     if (inflateInit2(&strm, 15 /* maxWindowLogSize */ + 16 /* gzip only */) != Z_OK)
         return FIO_ERROR_FRAME_DECODING;
 
-    strm.next_out = (Bytef*)ress->dstBuffer;
-    strm.avail_out = (uInt)ress->dstBufferSize;
-    strm.avail_in = (uInt)ress->srcBufferLoaded;
-    strm.next_in = (z_const unsigned char*)ress->srcBuffer;
+    writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
+    strm.next_out = (Bytef*)writeJob->buffer;
+    strm.avail_out = (uInt)writeJob->bufferSize;
+    strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
+    strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
 
     for ( ; ; ) {
         int ret;
         if (strm.avail_in == 0) {
-            ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile);
-            if (ress->srcBufferLoaded == 0) flush = Z_FINISH;
-            strm.next_in = (z_const unsigned char*)ress->srcBuffer;
-            strm.avail_in = (uInt)ress->srcBufferLoaded;
+            AIO_ReadPool_consumeAndRefill(ress->readCtx);
+            if (ress->readCtx->srcBufferLoaded == 0) flush = Z_FINISH;
+            strm.next_in = (z_const unsigned char*)ress->readCtx->srcBuffer;
+            strm.avail_in = (uInt)ress->readCtx->srcBufferLoaded;
         }
         ret = inflate(&strm, flush);
         if (ret == Z_BUF_ERROR) {
@@ -2343,35 +2545,34 @@ FIO_decompressGzFrame(dRess_t* ress, FILE* srcFile,
             DISPLAYLEVEL(1, "zstd: %s: inflate error %d \n", srcFileName, ret);
             decodingError = 1; break;
         }
-        {   size_t const decompBytes = ress->dstBufferSize - strm.avail_out;
+        {   size_t const decompBytes = writeJob->bufferSize - strm.avail_out;
             if (decompBytes) {
-                storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips);
+                writeJob->usedBufferSize = decompBytes;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 outFileSize += decompBytes;
-                strm.next_out = (Bytef*)ress->dstBuffer;
-                strm.avail_out = (uInt)ress->dstBufferSize;
+                strm.next_out = (Bytef*)writeJob->buffer;
+                strm.avail_out = (uInt)writeJob->bufferSize;
             }
         }
         if (ret == Z_STREAM_END) break;
     }
 
-    if (strm.avail_in > 0)
-        memmove(ress->srcBuffer, strm.next_in, strm.avail_in);
-    ress->srcBufferLoaded = strm.avail_in;
+    AIO_ReadPool_consumeBytes(ress->readCtx, ress->readCtx->srcBufferLoaded - strm.avail_in);
+
     if ( (inflateEnd(&strm) != Z_OK)  /* release resources ; error detected */
       && (decodingError==0) ) {
         DISPLAYLEVEL(1, "zstd: %s: inflateEnd error \n", srcFileName);
         decodingError = 1;
     }
-    FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
     return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
 }
 #endif
 
-
 #ifdef ZSTD_LZMADECOMPRESS
 static unsigned long long
-FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile,
-                        const FIO_prefs_t* const prefs,
+FIO_decompressLzmaFrame(dRess_t* ress,
                         const char* srcFileName, int plain_lzma)
 {
     unsigned long long outFileSize = 0;
@@ -2379,7 +2580,7 @@ FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile,
     lzma_action action = LZMA_RUN;
     lzma_ret initRet;
     int decodingError = 0;
-    unsigned storedSkips = 0;
+    IOJob_t *writeJob = NULL;
 
     strm.next_in = 0;
     strm.avail_in = 0;
@@ -2396,18 +2597,19 @@ FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile,
         return FIO_ERROR_FRAME_DECODING;
     }
 
-    strm.next_out = (BYTE*)ress->dstBuffer;
-    strm.avail_out = ress->dstBufferSize;
-    strm.next_in = (BYTE const*)ress->srcBuffer;
-    strm.avail_in = ress->srcBufferLoaded;
+    writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
+    strm.next_out = (BYTE*)writeJob->buffer;
+    strm.avail_out = writeJob->bufferSize;
+    strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
+    strm.avail_in = ress->readCtx->srcBufferLoaded;
 
     for ( ; ; ) {
         lzma_ret ret;
         if (strm.avail_in == 0) {
-            ress->srcBufferLoaded = fread(ress->srcBuffer, 1, ress->srcBufferSize, srcFile);
-            if (ress->srcBufferLoaded == 0) action = LZMA_FINISH;
-            strm.next_in = (BYTE const*)ress->srcBuffer;
-            strm.avail_in = ress->srcBufferLoaded;
+            AIO_ReadPool_consumeAndRefill(ress->readCtx);
+            if (ress->readCtx->srcBufferLoaded == 0) action = LZMA_FINISH;
+            strm.next_in = (BYTE const*)ress->readCtx->srcBuffer;
+            strm.avail_in = ress->readCtx->srcBufferLoaded;
         }
         ret = lzma_code(&strm, action);
 
@@ -2420,104 +2622,90 @@ FIO_decompressLzmaFrame(dRess_t* ress, FILE* srcFile,
                             srcFileName, ret);
             decodingError = 1; break;
         }
-        {   size_t const decompBytes = ress->dstBufferSize - strm.avail_out;
+        {   size_t const decompBytes = writeJob->bufferSize - strm.avail_out;
             if (decompBytes) {
-                storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decompBytes, prefs, storedSkips);
+                writeJob->usedBufferSize = decompBytes;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 outFileSize += decompBytes;
-                strm.next_out = (BYTE*)ress->dstBuffer;
-                strm.avail_out = ress->dstBufferSize;
+                strm.next_out = (BYTE*)writeJob->buffer;
+                strm.avail_out = writeJob->bufferSize;
         }   }
         if (ret == LZMA_STREAM_END) break;
     }
 
-    if (strm.avail_in > 0)
-        memmove(ress->srcBuffer, strm.next_in, strm.avail_in);
-    ress->srcBufferLoaded = strm.avail_in;
+    AIO_ReadPool_consumeBytes(ress->readCtx, ress->readCtx->srcBufferLoaded - strm.avail_in);
     lzma_end(&strm);
-    FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
     return decodingError ? FIO_ERROR_FRAME_DECODING : outFileSize;
 }
 #endif
 
 #ifdef ZSTD_LZ4DECOMPRESS
 static unsigned long long
-FIO_decompressLz4Frame(dRess_t* ress, FILE* srcFile,
-                       const FIO_prefs_t* const prefs,
-                       const char* srcFileName)
+FIO_decompressLz4Frame(dRess_t* ress, const char* srcFileName)
 {
     unsigned long long filesize = 0;
-    LZ4F_errorCode_t nextToLoad;
+    LZ4F_errorCode_t nextToLoad = 4;
     LZ4F_decompressionContext_t dCtx;
     LZ4F_errorCode_t const errorCode = LZ4F_createDecompressionContext(&dCtx, LZ4F_VERSION);
     int decodingError = 0;
-    unsigned storedSkips = 0;
+    IOJob_t *writeJob = NULL;
 
     if (LZ4F_isError(errorCode)) {
         DISPLAYLEVEL(1, "zstd: failed to create lz4 decompression context \n");
         return FIO_ERROR_FRAME_DECODING;
     }
 
-    /* Init feed with magic number (already consumed from FILE* sFile) */
-    {   size_t inSize = 4;
-        size_t outSize= 0;
-        MEM_writeLE32(ress->srcBuffer, LZ4_MAGICNUMBER);
-        nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &outSize, ress->srcBuffer, &inSize, NULL);
-        if (LZ4F_isError(nextToLoad)) {
-            DISPLAYLEVEL(1, "zstd: %s: lz4 header error : %s \n",
-                            srcFileName, LZ4F_getErrorName(nextToLoad));
-            LZ4F_freeDecompressionContext(dCtx);
-            return FIO_ERROR_FRAME_DECODING;
-    }   }
+    writeJob = AIO_WritePool_acquireJob(ress->writeCtx);
 
     /* Main Loop */
     for (;nextToLoad;) {
-        size_t readSize;
         size_t pos = 0;
-        size_t decodedBytes = ress->dstBufferSize;
+        size_t decodedBytes = writeJob->bufferSize;
+        int fullBufferDecoded = 0;
 
         /* Read input */
-        if (nextToLoad > ress->srcBufferSize) nextToLoad = ress->srcBufferSize;
-        readSize = fread(ress->srcBuffer, 1, nextToLoad, srcFile);
-        if (!readSize) break;   /* reached end of file or stream */
+        AIO_ReadPool_fillBuffer(ress->readCtx, nextToLoad);
+        if(!ress->readCtx->srcBufferLoaded) break; /* reached end of file */
 
-        while ((pos < readSize) || (decodedBytes == ress->dstBufferSize)) {  /* still to read, or still to flush */
+        while ((pos < ress->readCtx->srcBufferLoaded) || fullBufferDecoded) {  /* still to read, or still to flush */
             /* Decode Input (at least partially) */
-            size_t remaining = readSize - pos;
-            decodedBytes = ress->dstBufferSize;
-            nextToLoad = LZ4F_decompress(dCtx, ress->dstBuffer, &decodedBytes, (char*)(ress->srcBuffer)+pos, &remaining, NULL);
+            size_t remaining = ress->readCtx->srcBufferLoaded - pos;
+            decodedBytes = writeJob->bufferSize;
+            nextToLoad = LZ4F_decompress(dCtx, writeJob->buffer, &decodedBytes, (char*)(ress->readCtx->srcBuffer)+pos,
+                                         &remaining, NULL);
             if (LZ4F_isError(nextToLoad)) {
                 DISPLAYLEVEL(1, "zstd: %s: lz4 decompression error : %s \n",
                                 srcFileName, LZ4F_getErrorName(nextToLoad));
                 decodingError = 1; nextToLoad = 0; break;
             }
             pos += remaining;
+            assert(pos <= ress->readCtx->srcBufferLoaded);
+            fullBufferDecoded = decodedBytes == writeJob->bufferSize;
 
             /* Write Block */
             if (decodedBytes) {
                 UTIL_HumanReadableSize_t hrs;
-                storedSkips = FIO_fwriteSparse(ress->dstFile, ress->dstBuffer, decodedBytes, prefs, storedSkips);
+                writeJob->usedBufferSize = decodedBytes;
+                AIO_WritePool_enqueueAndReacquireWriteJob(&writeJob);
                 filesize += decodedBytes;
                 hrs = UTIL_makeHumanReadableSize(filesize);
-                DISPLAYUPDATE(2, "\rDecompressed : %.*f%s  ", hrs.precision, hrs.value, hrs.suffix);
+                DISPLAYUPDATE_PROGRESS("\rDecompressed : %.*f%s  ", hrs.precision, hrs.value, hrs.suffix);
             }
 
             if (!nextToLoad) break;
         }
+        AIO_ReadPool_consumeBytes(ress->readCtx, pos);
     }
-    /* can be out because readSize == 0, which could be an fread() error */
-    if (ferror(srcFile)) {
-        DISPLAYLEVEL(1, "zstd: %s: read error \n", srcFileName);
-        decodingError=1;
-    }
-
     if (nextToLoad!=0) {
         DISPLAYLEVEL(1, "zstd: %s: unfinished lz4 stream \n", srcFileName);
         decodingError=1;
     }
 
     LZ4F_freeDecompressionContext(dCtx);
-    ress->srcBufferLoaded = 0; /* LZ4F will reach exact frame boundary */
-    FIO_fwriteSparseEnd(prefs, ress->dstFile, storedSkips);
+    AIO_WritePool_releaseIoJob(writeJob);
+    AIO_WritePool_sparseWriteEnd(ress->writeCtx);
 
     return decodingError ? FIO_ERROR_FRAME_DECODING : filesize;
 }
@@ -2532,23 +2720,30 @@ FIO_decompressLz4Frame(dRess_t* ress, FILE* srcFile,
  *           1 : error
  */
 static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
-                          dRess_t ress, FILE* srcFile,
-                          const FIO_prefs_t* const prefs,
-                          const char* dstFileName, const char* srcFileName)
+                                dRess_t ress, const FIO_prefs_t* const prefs,
+                                const char* dstFileName, const char* srcFileName)
 {
     unsigned readSomething = 0;
     unsigned long long filesize = 0;
-    assert(srcFile != NULL);
+    int passThrough = prefs->passThrough;
+
+    if (passThrough == -1) {
+        /* If pass-through mode is not explicitly enabled or disabled,
+         * default to the legacy behavior of enabling it if we are writing
+         * to stdout with the overwrite flag enabled.
+         */
+        passThrough = prefs->overwrite && !strcmp(dstFileName, stdoutmark);
+    }
+    assert(passThrough == 0 || passThrough == 1);
 
     /* for each frame */
     for ( ; ; ) {
         /* check magic number -> version */
         size_t const toRead = 4;
-        const BYTE* const buf = (const BYTE*)ress.srcBuffer;
-        if (ress.srcBufferLoaded < toRead)  /* load up to 4 bytes for header */
-            ress.srcBufferLoaded += fread((char*)ress.srcBuffer + ress.srcBufferLoaded,
-                                          (size_t)1, toRead - ress.srcBufferLoaded, srcFile);
-        if (ress.srcBufferLoaded==0) {
+        const BYTE* buf;
+        AIO_ReadPool_fillBuffer(ress.readCtx, toRead);
+        buf = (const BYTE*)ress.readCtx->srcBuffer;
+        if (ress.readCtx->srcBufferLoaded==0) {
             if (readSomething==0) {  /* srcFile is empty (which is invalid) */
                 DISPLAYLEVEL(1, "zstd: %s: unexpected end of file \n", srcFileName);
                 return 1;
@@ -2556,17 +2751,20 @@ static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
             break;   /* no more input */
         }
         readSomething = 1;   /* there is at least 1 byte in srcFile */
-        if (ress.srcBufferLoaded < toRead) {
+        if (ress.readCtx->srcBufferLoaded < toRead) { /* not enough input to check magic number */
+            if (passThrough) {
+                return FIO_passThrough(&ress);
+            }
             DISPLAYLEVEL(1, "zstd: %s: unknown header \n", srcFileName);
             return 1;
         }
-        if (ZSTD_isFrame(buf, ress.srcBufferLoaded)) {
-            unsigned long long const frameSize = FIO_decompressZstdFrame(fCtx, &ress, srcFile, prefs, srcFileName, filesize);
+        if (ZSTD_isFrame(buf, ress.readCtx->srcBufferLoaded)) {
+            unsigned long long const frameSize = FIO_decompressZstdFrame(fCtx, &ress, prefs, srcFileName, filesize);
             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
             filesize += frameSize;
         } else if (buf[0] == 31 && buf[1] == 139) { /* gz magic number */
 #ifdef ZSTD_GZDECOMPRESS
-            unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFile, prefs, srcFileName);
+            unsigned long long const frameSize = FIO_decompressGzFrame(&ress, srcFileName);
             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
             filesize += frameSize;
 #else
@@ -2576,7 +2774,7 @@ static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
         } else if ((buf[0] == 0xFD && buf[1] == 0x37)  /* xz magic number */
                 || (buf[0] == 0x5D && buf[1] == 0x00)) { /* lzma header (no magic number) */
 #ifdef ZSTD_LZMADECOMPRESS
-            unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFile, prefs, srcFileName, buf[0] != 0xFD);
+            unsigned long long const frameSize = FIO_decompressLzmaFrame(&ress, srcFileName, buf[0] != 0xFD);
             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
             filesize += frameSize;
 #else
@@ -2585,18 +2783,15 @@ static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
 #endif
         } else if (MEM_readLE32(buf) == LZ4_MAGICNUMBER) {
 #ifdef ZSTD_LZ4DECOMPRESS
-            unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFile, prefs, srcFileName);
+            unsigned long long const frameSize = FIO_decompressLz4Frame(&ress, srcFileName);
             if (frameSize == FIO_ERROR_FRAME_DECODING) return 1;
             filesize += frameSize;
 #else
             DISPLAYLEVEL(1, "zstd: %s: lz4 file cannot be uncompressed (zstd compiled without HAVE_LZ4) -- ignored \n", srcFileName);
             return 1;
 #endif
-        } else if ((prefs->overwrite) && !strcmp (dstFileName, stdoutmark)) {  /* pass-through mode */
-            return FIO_passThrough(prefs,
-                                   ress.dstFile, srcFile,
-                                   ress.srcBuffer, ress.srcBufferSize,
-                                   ress.srcBufferLoaded);
+        } else if (passThrough) {
+            return FIO_passThrough(&ress);
         } else {
             DISPLAYLEVEL(1, "zstd: %s: unsupported format \n", srcFileName);
             return 1;
@@ -2604,48 +2799,47 @@ static int FIO_decompressFrames(FIO_ctx_t* const fCtx,
 
     /* Final Status */
     fCtx->totalBytesOutput += (size_t)filesize;
-    DISPLAYLEVEL(2, "\r%79s\r", "");
-    /* No status message in pipe mode (stdin - stdout) or multi-files mode */
-    if ((g_display_prefs.displayLevel >= 2 && fCtx->nbFilesTotal <= 1) ||
-        g_display_prefs.displayLevel >= 3 ||
-        g_display_prefs.progressSetting == FIO_ps_always) {
-        DISPLAYLEVEL(1, "\r%-20s: %llu bytes \n", srcFileName, filesize);
-    }
+    DISPLAY_PROGRESS("\r%79s\r", "");
+    if (FIO_shouldDisplayFileSummary(fCtx))
+        DISPLAY_SUMMARY("%-20s: %llu bytes \n", srcFileName, filesize);
 
     return 0;
 }
 
 /** FIO_decompressDstFile() :
-    open `dstFileName`,
-    or path-through if ress.dstFile is already != 0,
+    open `dstFileName`, or pass-through if writeCtx's file is already != 0,
     then start decompression process (FIO_decompressFrames()).
     @return : 0 : OK
               1 : operation aborted
 */
 static int FIO_decompressDstFile(FIO_ctx_t* const fCtx,
                                  FIO_prefs_t* const prefs,
-                                 dRess_t ress, FILE* srcFile,
-                                 const char* dstFileName, const char* srcFileName)
+                                 dRess_t ress,
+                                 const char* dstFileName,
+                                 const char* srcFileName,
+                                 const stat_t* srcFileStat)
 {
     int result;
-    stat_t statbuf;
     int releaseDstFile = 0;
-    int transferMTime = 0;
+    int transferStat = 0;
+    int dstFd = 0;
 
-    if ((ress.dstFile == NULL) && (prefs->testMode==0)) {
+    if ((AIO_WritePool_getFile(ress.writeCtx) == NULL) && (prefs->testMode == 0)) {
+        FILE *dstFile;
         int dstFilePermissions = DEFAULT_FILE_PERMISSIONS;
         if ( strcmp(srcFileName, stdinmark)   /* special case : don't transfer permissions from stdin */
           && strcmp(dstFileName, stdoutmark)
-          && UTIL_stat(srcFileName, &statbuf)
-          && UTIL_isRegularFileStat(&statbuf) ) {
-            dstFilePermissions = statbuf.st_mode;
-            transferMTime = 1;
+          && UTIL_isRegularFileStat(srcFileStat) ) {
+            transferStat = 1;
+            dstFilePermissions = TEMPORARY_FILE_PERMISSIONS;
         }
 
         releaseDstFile = 1;
 
-        ress.dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
-        if (ress.dstFile==NULL) return 1;
+        dstFile = FIO_openDstFile(fCtx, prefs, srcFileName, dstFileName, dstFilePermissions);
+        if (dstFile==NULL) return 1;
+        dstFd = fileno(dstFile);
+        AIO_WritePool_setFile(ress.writeCtx, dstFile);
 
         /* Must only be added after FIO_openDstFile() succeeds.
          * Otherwise we may delete the destination file if it already exists,
@@ -2654,19 +2848,22 @@ static int FIO_decompressDstFile(FIO_ctx_t* const fCtx,
         addHandler(dstFileName);
     }
 
-    result = FIO_decompressFrames(fCtx, ress, srcFile, prefs, dstFileName, srcFileName);
+    result = FIO_decompressFrames(fCtx, ress, prefs, dstFileName, srcFileName);
 
     if (releaseDstFile) {
-        FILE* const dstFile = ress.dstFile;
         clearHandler();
-        ress.dstFile = NULL;
-        if (fclose(dstFile)) {
+
+        if (transferStat) {
+            UTIL_setFDStat(dstFd, dstFileName, srcFileStat);
+        }
+
+        if (AIO_WritePool_closeFile(ress.writeCtx)) {
             DISPLAYLEVEL(1, "zstd: %s: %s \n", dstFileName, strerror(errno));
             result = 1;
         }
 
-        if (transferMTime) {
-            UTIL_utime(dstFileName, &statbuf);
+        if (transferStat) {
+            UTIL_utime(dstFileName, srcFileStat);
         }
 
         if ( (result != 0)  /* operation failure */
@@ -2688,18 +2885,34 @@ static int FIO_decompressDstFile(FIO_ctx_t* const fCtx,
 static int FIO_decompressSrcFile(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs, dRess_t ress, const char* dstFileName, const char* srcFileName)
 {
     FILE* srcFile;
+    stat_t srcFileStat;
     int result;
+    U64 fileSize = UTIL_FILESIZE_UNKNOWN;
 
     if (UTIL_isDirectory(srcFileName)) {
         DISPLAYLEVEL(1, "zstd: %s is a directory -- ignored \n", srcFileName);
         return 1;
     }
 
-    srcFile = FIO_openSrcFile(prefs, srcFileName);
+    srcFile = FIO_openSrcFile(prefs, srcFileName, &srcFileStat);
     if (srcFile==NULL) return 1;
-    ress.srcBufferLoaded = 0;
 
-    result = FIO_decompressDstFile(fCtx, prefs, ress, srcFile, dstFileName, srcFileName);
+    /* Don't use AsyncIO for small files */
+    if (strcmp(srcFileName, stdinmark)) /* Stdin doesn't have stats */
+        fileSize = UTIL_getFileSizeStat(&srcFileStat);
+    if(fileSize != UTIL_FILESIZE_UNKNOWN && fileSize < ZSTD_BLOCKSIZE_MAX * 3) {
+        AIO_ReadPool_setAsync(ress.readCtx, 0);
+        AIO_WritePool_setAsync(ress.writeCtx, 0);
+    } else {
+        AIO_ReadPool_setAsync(ress.readCtx, 1);
+        AIO_WritePool_setAsync(ress.writeCtx, 1);
+    }
+
+    AIO_ReadPool_setFile(ress.readCtx, srcFile);
+
+    result = FIO_decompressDstFile(fCtx, prefs, ress, dstFileName, srcFileName, &srcFileStat);
+
+    AIO_ReadPool_setFile(ress.readCtx, NULL);
 
     /* Close file */
     if (fclose(srcFile)) {
@@ -2731,6 +2944,8 @@ int FIO_decompressFilename(FIO_ctx_t* const fCtx, FIO_prefs_t* const prefs,
 
     int const decodingError = FIO_decompressSrcFile(fCtx, prefs, ress, dstFileName, srcFileName);
 
+
+
     FIO_freeDResources(ress);
     return decodingError;
 }
@@ -2788,6 +3003,11 @@ FIO_determineDstName(const char* srcFileName, const char* outDirName)
 
     size_t srcSuffixLen;
     const char* const srcSuffix = strrchr(srcFileName, '.');
+
+    if(!strcmp(srcFileName, stdinmark)) {
+        return stdoutmark;
+    }
+
     if (srcSuffix == NULL) {
         DISPLAYLEVEL(1,
             "zstd: %s: unknown suffix (%s expected). "
@@ -2869,20 +3089,21 @@ FIO_decompressMultipleFilenames(FIO_ctx_t* const fCtx,
     dRess_t ress = FIO_createDResources(prefs, dictFileName);
 
     if (outFileName) {
-        if (FIO_removeMultiFilesWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
+        if (FIO_multiFilesConcatWarning(fCtx, prefs, outFileName, 1 /* displayLevelCutoff */)) {
             FIO_freeDResources(ress);
             return 1;
         }
         if (!prefs->testMode) {
-            ress.dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
-            if (ress.dstFile == 0) EXM_THROW(19, "cannot open %s", outFileName);
+            FILE* dstFile = FIO_openDstFile(fCtx, prefs, NULL, outFileName, DEFAULT_FILE_PERMISSIONS);
+            if (dstFile == 0) EXM_THROW(19, "cannot open %s", outFileName);
+            AIO_WritePool_setFile(ress.writeCtx, dstFile);
         }
         for (; fCtx->currFileIdx < fCtx->nbFilesTotal; fCtx->currFileIdx++) {
             status = FIO_decompressSrcFile(fCtx, prefs, ress, outFileName, srcNamesTable[fCtx->currFileIdx]);
             if (!status) fCtx->nbFilesProcessed++;
             error |= status;
         }
-        if ((!prefs->testMode) && (fclose(ress.dstFile)))
+        if ((!prefs->testMode) && (AIO_WritePool_closeFile(ress.writeCtx)))
             EXM_THROW(72, "Write error : %s : cannot properly close output file",
                         strerror(errno));
     } else {
@@ -2912,8 +3133,11 @@ FIO_decompressMultipleFilenames(FIO_ctx_t* const fCtx,
             FIO_checkFilenameCollisions(srcNamesTable , (unsigned)fCtx->nbFilesTotal);
     }
 
-    if (fCtx->nbFilesProcessed >= 1  && fCtx->nbFilesTotal > 1 && fCtx->totalBytesOutput != 0)
-        DISPLAYLEVEL(2, "%d files decompressed : %6zu bytes total \n", fCtx->nbFilesProcessed, fCtx->totalBytesOutput);
+    if (FIO_shouldDisplayMultipleFileSummary(fCtx)) {
+        DISPLAY_PROGRESS("\r%79s\r", "");
+        DISPLAY_SUMMARY("%d files decompressed : %6llu bytes total \n",
+            fCtx->nbFilesProcessed, (unsigned long long)fCtx->totalBytesOutput);
+    }
 
     FIO_freeDResources(ress);
     return error;
@@ -2931,7 +3155,9 @@ typedef struct {
     int numSkippableFrames;
     int decompUnavailable;
     int usesCheck;
+    BYTE checksum[4];
     U32 nbFiles;
+    unsigned dictID;
 } fileInfo_t;
 
 typedef enum {
@@ -2939,7 +3165,7 @@ typedef enum {
   info_frame_error=1,
   info_not_zstd=2,
   info_file_error=3,
-  info_truncated_input=4,
+  info_truncated_input=4
 } InfoError;
 
 #define ERROR_IF(c,n,...) {             \
@@ -2976,7 +3202,7 @@ FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
         {   U32 const magicNumber = MEM_readLE32(headerBuffer);
             /* Zstandard frame */
             if (magicNumber == ZSTD_MAGICNUMBER) {
-                ZSTD_frameHeader header;
+                ZSTD_FrameHeader header;
                 U64 const frameContentSize = ZSTD_getFrameContentSize(headerBuffer, numBytesRead);
                 if ( frameContentSize == ZSTD_CONTENTSIZE_ERROR
                   || frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN ) {
@@ -2986,6 +3212,12 @@ FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
                 }
                 ERROR_IF(ZSTD_getFrameHeader(&header, headerBuffer, numBytesRead) != 0,
                         info_frame_error, "Error: could not decode frame header");
+                if (info->dictID != 0 && info->dictID != header.dictID) {
+                    DISPLAY("WARNING: File contains multiple frames with different dictionary IDs. Showing dictID 0 instead");
+                    info->dictID = 0;
+                } else {
+                    info->dictID = header.dictID;
+                }
                 info->windowSize = header.windowSize;
                 /* move to the end of the frame header */
                 {   size_t const headerSize = ZSTD_frameHeaderSize(headerBuffer, numBytesRead);
@@ -3018,8 +3250,8 @@ FIO_analyzeFrames(fileInfo_t* info, FILE* const srcFile)
                     int const contentChecksumFlag = (frameHeaderDescriptor & (1 << 2)) >> 2;
                     if (contentChecksumFlag) {
                         info->usesCheck = 1;
-                        ERROR_IF(fseek(srcFile, 4, SEEK_CUR) != 0,
-                                info_frame_error, "Error: could not skip past checksum");
+                        ERROR_IF(fread(info->checksum, 1, 4, srcFile) != 4,
+                                info_frame_error, "Error: could not read checksum");
                 }   }
                 info->numActualFrames++;
             }
@@ -3045,10 +3277,11 @@ static InfoError
 getFileInfo_fileConfirmed(fileInfo_t* info, const char* inFileName)
 {
     InfoError status;
-    FILE* const srcFile = FIO_openSrcFile(NULL, inFileName);
+    stat_t srcFileStat;
+    FILE* const srcFile = FIO_openSrcFile(NULL, inFileName, &srcFileStat);
     ERROR_IF(srcFile == NULL, info_file_error, "Error: could not open source file %s", inFileName);
 
-    info->compressedSize = UTIL_getFileSize(inFileName);
+    info->compressedSize = UTIL_getFileSizeStat(&srcFileStat);
     status = FIO_analyzeFrames(info, srcFile);
 
     fclose(srcFile);
@@ -3098,6 +3331,7 @@ displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel)
         DISPLAYOUT("# Zstandard Frames: %d\n", info->numActualFrames);
         if (info->numSkippableFrames)
             DISPLAYOUT("# Skippable Frames: %d\n", info->numSkippableFrames);
+        DISPLAYOUT("DictID: %u\n", info->dictID);
         DISPLAYOUT("Window Size: %.*f%s (%llu B)\n",
                    window_hrs.precision, window_hrs.value, window_hrs.suffix,
                    (unsigned long long)info->windowSize);
@@ -3110,7 +3344,16 @@ displayInfo(const char* inFileName, const fileInfo_t* info, int displayLevel)
                     (unsigned long long)info->decompressedSize);
             DISPLAYOUT("Ratio: %.4f\n", ratio);
         }
-        DISPLAYOUT("Check: %s\n", checkString);
+
+        if (info->usesCheck && info->numActualFrames == 1) {
+            DISPLAYOUT("Check: %s %02x%02x%02x%02x\n", checkString,
+                info->checksum[3], info->checksum[2],
+                info->checksum[1], info->checksum[0]
+            );
+        } else {
+            DISPLAYOUT("Check: %s\n", checkString);
+        }
+
         DISPLAYOUT("\n");
     }
 }
@@ -3174,7 +3417,7 @@ int FIO_listMultipleFiles(unsigned numFiles, const char** filenameTable, int dis
     }   }
 
     if (numFiles == 0) {
-        if (!IS_CONSOLE(stdin)) {
+        if (!UTIL_isConsole(stdin)) {
             DISPLAYLEVEL(1, "zstd: --list does not support reading from standard input \n");
         }
         DISPLAYLEVEL(1, "No files given \n");
diff --git a/sys/contrib/zstd/programs/fileio.h b/sys/contrib/zstd/programs/fileio.h
index 61094db83cba..cb53ef537810 100644
--- a/sys/contrib/zstd/programs/fileio.h
+++ b/sys/contrib/zstd/programs/fileio.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -12,14 +12,11 @@
 #ifndef FILEIO_H_23981798732
 #define FILEIO_H_23981798732
 
+#include "fileio_types.h"
+#include "util.h"                  /* FileNamesTable */
 #define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_compressionParameters */
 #include "../lib/zstd.h"           /* ZSTD_* */
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-
 /* *************************************
 *  Special i/o constants
 **************************************/
@@ -53,10 +50,6 @@ extern "C" {
 /*-*************************************
 *  Types
 ***************************************/
-typedef enum { FIO_zstdCompression, FIO_gzipCompression, FIO_xzCompression, FIO_lzmaCompression, FIO_lz4Compression } FIO_compressionType_t;
-
-typedef struct FIO_prefs_s FIO_prefs_t;
-
 FIO_prefs_t* FIO_createPreferences(void);
 void FIO_freePreferences(FIO_prefs_t* const prefs);
 
@@ -66,9 +59,6 @@ typedef struct FIO_ctx_s FIO_ctx_t;
 FIO_ctx_t* FIO_createContext(void);
 void FIO_freeContext(FIO_ctx_t* const fCtx);
 
-typedef struct FIO_display_prefs_s FIO_display_prefs_t;
-
-typedef enum { FIO_ps_auto, FIO_ps_never, FIO_ps_always } FIO_progressSetting_e;
 
 /*-*************************************
 *  Parameters
@@ -76,7 +66,7 @@ typedef enum { FIO_ps_auto, FIO_ps_never, FIO_ps_always } FIO_progressSetting_e;
 /* FIO_prefs_t functions */
 void FIO_setCompressionType(FIO_prefs_t* const prefs, FIO_compressionType_t compressionType);
 void FIO_overwriteMode(FIO_prefs_t* const prefs);
-void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, unsigned adapt);
+void FIO_setAdaptiveMode(FIO_prefs_t* const prefs, int adapt);
 void FIO_setAdaptMin(FIO_prefs_t* const prefs, int minCLevel);
 void FIO_setAdaptMax(FIO_prefs_t* const prefs, int maxCLevel);
 void FIO_setUseRowMatchFinder(FIO_prefs_t* const prefs, int useRowMatchFinder);
@@ -91,8 +81,8 @@ void FIO_setLdmMinMatch(FIO_prefs_t* const prefs, int ldmMinMatch);
 void FIO_setMemLimit(FIO_prefs_t* const prefs, unsigned memLimit);
 void FIO_setNbWorkers(FIO_prefs_t* const prefs, int nbWorkers);
 void FIO_setOverlapLog(FIO_prefs_t* const prefs, int overlapLog);
-void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, unsigned flag);
-void FIO_setSparseWrite(FIO_prefs_t* const prefs, unsigned sparse);  /**< 0: no sparse; 1: disable on stdout; 2: always enabled */
+void FIO_setRemoveSrcFile(FIO_prefs_t* const prefs, int flag);
+void FIO_setSparseWrite(FIO_prefs_t* const prefs, int sparse);  /**< 0: no sparse; 1: disable on stdout; 2: always enabled */
 void FIO_setRsyncable(FIO_prefs_t* const prefs, int rsyncable);
 void FIO_setStreamSrcSize(FIO_prefs_t* const prefs, size_t streamSrcSize);
 void FIO_setTargetCBlockSize(FIO_prefs_t* const prefs, size_t targetCBlockSize);
@@ -100,7 +90,7 @@ void FIO_setSrcSizeHint(FIO_prefs_t* const prefs, size_t srcSizeHint);
 void FIO_setTestMode(FIO_prefs_t* const prefs, int testMode);
 void FIO_setLiteralCompressionMode(
         FIO_prefs_t* const prefs,
-        ZSTD_paramSwitch_e mode);
+        ZSTD_ParamSwitch_e mode);
 
 void FIO_setProgressSetting(FIO_progressSetting_e progressSetting);
 void FIO_setNotificationLevel(int level);
@@ -109,6 +99,9 @@ void FIO_setAllowBlockDevices(FIO_prefs_t* const prefs, int allowBlockDevices);
 void FIO_setPatchFromMode(FIO_prefs_t* const prefs, int value);
 void FIO_setContentSize(FIO_prefs_t* const prefs, int value);
 void FIO_displayCompressionParameters(const FIO_prefs_t* prefs);
+void FIO_setAsyncIOFlag(FIO_prefs_t* const prefs, int value);
+void FIO_setPassThroughFlag(FIO_prefs_t* const prefs, int value);
+void FIO_setMMapDict(FIO_prefs_t* const prefs, ZSTD_ParamSwitch_e value);
 
 /* FIO_ctx_t functions */
 void FIO_setNbFilesTotal(FIO_ctx_t* const fCtx, int value);
@@ -171,10 +164,8 @@ int FIO_checkFilenameCollisions(const char** filenameTable, unsigned nbFiles);
 /* custom crash signal handler */
 void FIO_addAbortHandler(void);
 
-
-
-#if defined (__cplusplus)
-}
-#endif
+char const* FIO_zlibVersion(void);
+char const* FIO_lz4Version(void);
+char const* FIO_lzmaVersion(void);
 
 #endif  /* FILEIO_H_23981798732 */
diff --git a/sys/contrib/zstd/programs/fileio_asyncio.c b/sys/contrib/zstd/programs/fileio_asyncio.c
new file mode 100644
index 000000000000..42a47201656e
--- /dev/null
+++ b/sys/contrib/zstd/programs/fileio_asyncio.c
@@ -0,0 +1,663 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "platform.h"
+#include <stdio.h>      /* fprintf, open, fdopen, fread, _fileno, stdin, stdout */
+#include <stdlib.h>     /* malloc, free */
+#include <assert.h>
+#include <errno.h>      /* errno */
+
+#if defined (_MSC_VER)
+#  include <sys/stat.h>
+#  include <io.h>
+#endif
+
+#include "fileio_asyncio.h"
+#include "fileio_common.h"
+
+/* **********************************************************************
+ *  Sparse write
+ ************************************************************************/
+
+/** AIO_fwriteSparse() :
+*  @return : storedSkips,
+*            argument for next call to AIO_fwriteSparse() or AIO_fwriteSparseEnd() */
+static unsigned
+AIO_fwriteSparse(FILE* file,
+                 const void* buffer, size_t bufferSize,
+                 const FIO_prefs_t* const prefs,
+                 unsigned storedSkips)
+{
+    const size_t* const bufferT = (const size_t*)buffer;   /* Buffer is supposed malloc'ed, hence aligned on size_t */
+    size_t bufferSizeT = bufferSize / sizeof(size_t);
+    const size_t* const bufferTEnd = bufferT + bufferSizeT;
+    const size_t* ptrT = bufferT;
+    static const size_t segmentSizeT = (32 KB) / sizeof(size_t);   /* check every 32 KB */
+
+    if (prefs->testMode) return 0;  /* do not output anything in test mode */
+
+    if (!prefs->sparseFileSupport) {  /* normal write */
+        size_t const sizeCheck = fwrite(buffer, 1, bufferSize, file);
+        if (sizeCheck != bufferSize)
+            EXM_THROW(70, "Write error : cannot write block : %s",
+                      strerror(errno));
+        return 0;
+    }
+
+    /* avoid int overflow */
+    if (storedSkips > 1 GB) {
+        if (LONG_SEEK(file, 1 GB, SEEK_CUR) != 0)
+        EXM_THROW(91, "1 GB skip error (sparse file support)");
+        storedSkips -= 1 GB;
+    }
+
+    while (ptrT < bufferTEnd) {
+        size_t nb0T;
+
+        /* adjust last segment if < 32 KB */
+        size_t seg0SizeT = segmentSizeT;
+        if (seg0SizeT > bufferSizeT) seg0SizeT = bufferSizeT;
+        bufferSizeT -= seg0SizeT;
+
+        /* count leading zeroes */
+        for (nb0T=0; (nb0T < seg0SizeT) && (ptrT[nb0T] == 0); nb0T++) ;
+        storedSkips += (unsigned)(nb0T * sizeof(size_t));
+
+        if (nb0T != seg0SizeT) {   /* not all 0s */
+            size_t const nbNon0ST = seg0SizeT - nb0T;
+            /* skip leading zeros */
+            if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
+                EXM_THROW(92, "Sparse skip error ; try --no-sparse");
+            storedSkips = 0;
+            /* write the rest */
+            if (fwrite(ptrT + nb0T, sizeof(size_t), nbNon0ST, file) != nbNon0ST)
+                EXM_THROW(93, "Write error : cannot write block : %s",
+                          strerror(errno));
+        }
+        ptrT += seg0SizeT;
+    }
+
+    {   static size_t const maskT = sizeof(size_t)-1;
+        if (bufferSize & maskT) {
+            /* size not multiple of sizeof(size_t) : implies end of block */
+            const char* const restStart = (const char*)bufferTEnd;
+            const char* restPtr = restStart;
+            const char* const restEnd = (const char*)buffer + bufferSize;
+            assert(restEnd > restStart && restEnd < restStart + sizeof(size_t));
+            for ( ; (restPtr < restEnd) && (*restPtr == 0); restPtr++) ;
+            storedSkips += (unsigned) (restPtr - restStart);
+            if (restPtr != restEnd) {
+                /* not all remaining bytes are 0 */
+                size_t const restSize = (size_t)(restEnd - restPtr);
+                if (LONG_SEEK(file, storedSkips, SEEK_CUR) != 0)
+                    EXM_THROW(92, "Sparse skip error ; try --no-sparse");
+                if (fwrite(restPtr, 1, restSize, file) != restSize)
+                    EXM_THROW(95, "Write error : cannot write end of decoded block : %s",
+                              strerror(errno));
+                storedSkips = 0;
+            }   }   }
+
+    return storedSkips;
+}
+
+static void
+AIO_fwriteSparseEnd(const FIO_prefs_t* const prefs, FILE* file, unsigned storedSkips)
+{
+    if (prefs->testMode) assert(storedSkips == 0);
+    if (storedSkips>0) {
+        assert(prefs->sparseFileSupport > 0);  /* storedSkips>0 implies sparse support is enabled */
+        (void)prefs;   /* assert can be disabled, in which case prefs becomes unused */
+        if (LONG_SEEK(file, storedSkips-1, SEEK_CUR) != 0)
+            EXM_THROW(69, "Final skip error (sparse file support)");
+        /* last zero must be explicitly written,
+         * so that skipped ones get implicitly translated as zero by FS */
+        {   const char lastZeroByte[1] = { 0 };
+            if (fwrite(lastZeroByte, 1, 1, file) != 1)
+                EXM_THROW(69, "Write error : cannot write last zero : %s", strerror(errno));
+        }   }
+}
+
+
+/* **********************************************************************
+ *  AsyncIO functionality
+ ************************************************************************/
+
+/* AIO_supported:
+ * Returns 1 if AsyncIO is supported on the system, 0 otherwise. */
+int AIO_supported(void) {
+#ifdef ZSTD_MULTITHREAD
+    return 1;
+#else
+    return 0;
+#endif
+}
+
+/* ***********************************
+ *  Generic IoPool implementation
+ *************************************/
+
+static IOJob_t *AIO_IOPool_createIoJob(IOPoolCtx_t *ctx, size_t bufferSize) {
+    IOJob_t* const job  = (IOJob_t*) malloc(sizeof(IOJob_t));
+    void* const buffer = malloc(bufferSize);
+    if(!job || !buffer)
+        EXM_THROW(101, "Allocation error : not enough memory");
+    job->buffer = buffer;
+    job->bufferSize = bufferSize;
+    job->usedBufferSize = 0;
+    job->file = NULL;
+    job->ctx = ctx;
+    job->offset = 0;
+    return job;
+}
+
+
+/* AIO_IOPool_createThreadPool:
+ * Creates a thread pool and a mutex for threaded IO pool.
+ * Displays warning if asyncio is requested but MT isn't available. */
+static void AIO_IOPool_createThreadPool(IOPoolCtx_t* ctx, const FIO_prefs_t* prefs) {
+    ctx->threadPool = NULL;
+    ctx->threadPoolActive = 0;
+    if(prefs->asyncIO) {
+        if (ZSTD_pthread_mutex_init(&ctx->ioJobsMutex, NULL))
+            EXM_THROW(102,"Failed creating ioJobsMutex mutex");
+        /* We want MAX_IO_JOBS-2 queue items because we need to always have 1 free buffer to
+         * decompress into and 1 buffer that's actively written to disk and owned by the writing thread. */
+        assert(MAX_IO_JOBS >= 2);
+        ctx->threadPool = POOL_create(1, MAX_IO_JOBS - 2);
+        ctx->threadPoolActive = 1;
+        if (!ctx->threadPool)
+            EXM_THROW(104, "Failed creating I/O thread pool");
+    }
+}
+
+/* AIO_IOPool_init:
+ * Allocates and sets and a new I/O thread pool including its included availableJobs. */
+static void AIO_IOPool_init(IOPoolCtx_t* ctx, const FIO_prefs_t* prefs, POOL_function poolFunction, size_t bufferSize) {
+    int i;
+    AIO_IOPool_createThreadPool(ctx, prefs);
+    ctx->prefs = prefs;
+    ctx->poolFunction = poolFunction;
+    ctx->totalIoJobs = ctx->threadPool ? MAX_IO_JOBS : 2;
+    ctx->availableJobsCount = ctx->totalIoJobs;
+    for(i=0; i < ctx->availableJobsCount; i++) {
+        ctx->availableJobs[i] = AIO_IOPool_createIoJob(ctx, bufferSize);
+    }
+    ctx->jobBufferSize = bufferSize;
+    ctx->file = NULL;
+}
+
+
+/* AIO_IOPool_threadPoolActive:
+ * Check if current operation uses thread pool.
+ * Note that in some cases we have a thread pool initialized but choose not to use it. */
+static int AIO_IOPool_threadPoolActive(IOPoolCtx_t* ctx) {
+    return ctx->threadPool && ctx->threadPoolActive;
+}
+
+
+/* AIO_IOPool_lockJobsMutex:
+ * Locks the IO jobs mutex if threading is active */
+static void AIO_IOPool_lockJobsMutex(IOPoolCtx_t* ctx) {
+    if(AIO_IOPool_threadPoolActive(ctx))
+        ZSTD_pthread_mutex_lock(&ctx->ioJobsMutex);
+}
+
+/* AIO_IOPool_unlockJobsMutex:
+ * Unlocks the IO jobs mutex if threading is active */
+static void AIO_IOPool_unlockJobsMutex(IOPoolCtx_t* ctx) {
+    if(AIO_IOPool_threadPoolActive(ctx))
+        ZSTD_pthread_mutex_unlock(&ctx->ioJobsMutex);
+}
+
+/* AIO_IOPool_releaseIoJob:
+ * Releases an acquired job back to the pool. Doesn't execute the job. */
+static void AIO_IOPool_releaseIoJob(IOJob_t* job) {
+    IOPoolCtx_t* const ctx = (IOPoolCtx_t *) job->ctx;
+    AIO_IOPool_lockJobsMutex(ctx);
+    assert(ctx->availableJobsCount < ctx->totalIoJobs);
+    ctx->availableJobs[ctx->availableJobsCount++] = job;
+    AIO_IOPool_unlockJobsMutex(ctx);
+}
+
+/* AIO_IOPool_join:
+ * Waits for all tasks in the pool to finish executing. */
+static void AIO_IOPool_join(IOPoolCtx_t* ctx) {
+    if(AIO_IOPool_threadPoolActive(ctx))
+        POOL_joinJobs(ctx->threadPool);
+}
+
+/* AIO_IOPool_setThreaded:
+ * Allows (de)activating threaded mode, to be used when the expected overhead
+ * of threading costs more than the expected gains. */
+static void AIO_IOPool_setThreaded(IOPoolCtx_t* ctx, int threaded) {
+    assert(threaded == 0 || threaded == 1);
+    assert(ctx != NULL);
+    if(ctx->threadPoolActive != threaded) {
+        AIO_IOPool_join(ctx);
+        ctx->threadPoolActive = threaded;
+    }
+}
+
+/* AIO_IOPool_free:
+ * Release a previously allocated IO thread pool. Makes sure all tasks are done and released. */
+static void AIO_IOPool_destroy(IOPoolCtx_t* ctx) {
+    int i;
+    if(ctx->threadPool) {
+        /* Make sure we finish all tasks and then free the resources */
+        AIO_IOPool_join(ctx);
+        /* Make sure we are not leaking availableJobs */
+        assert(ctx->availableJobsCount == ctx->totalIoJobs);
+        POOL_free(ctx->threadPool);
+        ZSTD_pthread_mutex_destroy(&ctx->ioJobsMutex);
+    }
+    assert(ctx->file == NULL);
+    for(i=0; i<ctx->availableJobsCount; i++) {
+        IOJob_t* job = (IOJob_t*) ctx->availableJobs[i];
+        free(job->buffer);
+        free(job);
+    }
+}
+
+/* AIO_IOPool_acquireJob:
+ * Returns an available io job to be used for a future io. */
+static IOJob_t* AIO_IOPool_acquireJob(IOPoolCtx_t* ctx) {
+    IOJob_t* job;
+    assert(ctx->file != NULL || ctx->prefs->testMode);
+    AIO_IOPool_lockJobsMutex(ctx);
+    assert(ctx->availableJobsCount > 0);
+    job = (IOJob_t*) ctx->availableJobs[--ctx->availableJobsCount];
+    AIO_IOPool_unlockJobsMutex(ctx);
+    job->usedBufferSize = 0;
+    job->file = ctx->file;
+    job->offset = 0;
+    return job;
+}
+
+
+/* AIO_IOPool_setFile:
+ * Sets the destination file for future files in the pool.
+ * Requires completion of all queued jobs and release of all otherwise acquired jobs. */
+static void AIO_IOPool_setFile(IOPoolCtx_t* ctx, FILE* file) {
+    assert(ctx!=NULL);
+    AIO_IOPool_join(ctx);
+    assert(ctx->availableJobsCount == ctx->totalIoJobs);
+    ctx->file = file;
+}
+
+static FILE* AIO_IOPool_getFile(const IOPoolCtx_t* ctx) {
+    return ctx->file;
+}
+
+/* AIO_IOPool_enqueueJob:
+ * Enqueues an io job for execution.
+ * The queued job shouldn't be used directly after queueing it. */
+static void AIO_IOPool_enqueueJob(IOJob_t* job) {
+    IOPoolCtx_t* const ctx = (IOPoolCtx_t *)job->ctx;
+    if(AIO_IOPool_threadPoolActive(ctx))
+        POOL_add(ctx->threadPool, ctx->poolFunction, job);
+    else
+        ctx->poolFunction(job);
+}
+
+/* ***********************************
+ *  WritePool implementation
+ *************************************/
+
+/* AIO_WritePool_acquireJob:
+ * Returns an available write job to be used for a future write. */
+IOJob_t* AIO_WritePool_acquireJob(WritePoolCtx_t* ctx) {
+    return AIO_IOPool_acquireJob(&ctx->base);
+}
+
+/* AIO_WritePool_enqueueAndReacquireWriteJob:
+ * Queues a write job for execution and acquires a new one.
+ * After execution `job`'s pointed value would change to the newly acquired job.
+ * Make sure to set `usedBufferSize` to the wanted length before call.
+ * The queued job shouldn't be used directly after queueing it. */
+void AIO_WritePool_enqueueAndReacquireWriteJob(IOJob_t **job) {
+    AIO_IOPool_enqueueJob(*job);
+    *job = AIO_IOPool_acquireJob((IOPoolCtx_t *)(*job)->ctx);
+}
+
+/* AIO_WritePool_sparseWriteEnd:
+ * Ends sparse writes to the current file.
+ * Blocks on completion of all current write jobs before executing. */
+void AIO_WritePool_sparseWriteEnd(WritePoolCtx_t* ctx) {
+    assert(ctx != NULL);
+    AIO_IOPool_join(&ctx->base);
+    AIO_fwriteSparseEnd(ctx->base.prefs, ctx->base.file, ctx->storedSkips);
+    ctx->storedSkips = 0;
+}
+
+/* AIO_WritePool_setFile:
+ * Sets the destination file for future writes in the pool.
+ * Requires completion of all queues write jobs and release of all otherwise acquired jobs.
+ * Also requires ending of sparse write if a previous file was used in sparse mode. */
+void AIO_WritePool_setFile(WritePoolCtx_t* ctx, FILE* file) {
+    AIO_IOPool_setFile(&ctx->base, file);
+    assert(ctx->storedSkips == 0);
+}
+
+/* AIO_WritePool_getFile:
+ * Returns the file the writePool is currently set to write to. */
+FILE* AIO_WritePool_getFile(const WritePoolCtx_t* ctx) {
+    return AIO_IOPool_getFile(&ctx->base);
+}
+
+/* AIO_WritePool_releaseIoJob:
+ * Releases an acquired job back to the pool. Doesn't execute the job. */
+void AIO_WritePool_releaseIoJob(IOJob_t* job) {
+    AIO_IOPool_releaseIoJob(job);
+}
+
+/* AIO_WritePool_closeFile:
+ * Ends sparse write and closes the writePool's current file and sets the file to NULL.
+ * Requires completion of all queues write jobs and release of all otherwise acquired jobs.  */
+int AIO_WritePool_closeFile(WritePoolCtx_t* ctx) {
+    FILE* const dstFile = ctx->base.file;
+    assert(dstFile!=NULL || ctx->base.prefs->testMode!=0);
+    AIO_WritePool_sparseWriteEnd(ctx);
+    AIO_IOPool_setFile(&ctx->base, NULL);
+    return fclose(dstFile);
+}
+
+/* AIO_WritePool_executeWriteJob:
+ * Executes a write job synchronously. Can be used as a function for a thread pool. */
+static void AIO_WritePool_executeWriteJob(void* opaque){
+    IOJob_t* const job = (IOJob_t*) opaque;
+    WritePoolCtx_t* const ctx = (WritePoolCtx_t*) job->ctx;
+    ctx->storedSkips = AIO_fwriteSparse(job->file, job->buffer, job->usedBufferSize, ctx->base.prefs, ctx->storedSkips);
+    AIO_IOPool_releaseIoJob(job);
+}
+
+/* AIO_WritePool_create:
+ * Allocates and sets and a new write pool including its included jobs. */
+WritePoolCtx_t* AIO_WritePool_create(const FIO_prefs_t* prefs, size_t bufferSize) {
+    WritePoolCtx_t* const ctx = (WritePoolCtx_t*) malloc(sizeof(WritePoolCtx_t));
+    if(!ctx) EXM_THROW(100, "Allocation error : not enough memory");
+    AIO_IOPool_init(&ctx->base, prefs, AIO_WritePool_executeWriteJob, bufferSize);
+    ctx->storedSkips = 0;
+    return ctx;
+}
+
+/* AIO_WritePool_free:
+ * Frees and releases a writePool and its resources. Closes destination file if needs to. */
+void AIO_WritePool_free(WritePoolCtx_t* ctx) {
+    /* Make sure we finish all tasks and then free the resources */
+    if(AIO_WritePool_getFile(ctx))
+        AIO_WritePool_closeFile(ctx);
+    AIO_IOPool_destroy(&ctx->base);
+    assert(ctx->storedSkips==0);
+    free(ctx);
+}
+
+/* AIO_WritePool_setAsync:
+ * Allows (de)activating async mode, to be used when the expected overhead
+ * of asyncio costs more than the expected gains. */
+void AIO_WritePool_setAsync(WritePoolCtx_t* ctx, int async) {
+    AIO_IOPool_setThreaded(&ctx->base, async);
+}
+
+
+/* ***********************************
+ *  ReadPool implementation
+ *************************************/
+static void AIO_ReadPool_releaseAllCompletedJobs(ReadPoolCtx_t* ctx) {
+    int i;
+    for(i=0; i<ctx->completedJobsCount; i++) {
+        IOJob_t* job = (IOJob_t*) ctx->completedJobs[i];
+        AIO_IOPool_releaseIoJob(job);
+    }
+    ctx->completedJobsCount = 0;
+}
+
+static void AIO_ReadPool_addJobToCompleted(IOJob_t* job) {
+    ReadPoolCtx_t* const ctx = (ReadPoolCtx_t *)job->ctx;
+    AIO_IOPool_lockJobsMutex(&ctx->base);
+    assert(ctx->completedJobsCount < MAX_IO_JOBS);
+    ctx->completedJobs[ctx->completedJobsCount++] = job;
+    if(AIO_IOPool_threadPoolActive(&ctx->base)) {
+        ZSTD_pthread_cond_signal(&ctx->jobCompletedCond);
+    }
+    AIO_IOPool_unlockJobsMutex(&ctx->base);
+}
+
+/* AIO_ReadPool_findNextWaitingOffsetCompletedJob_locked:
+ * Looks through the completed jobs for a job matching the waitingOnOffset and returns it,
+ * if job wasn't found returns NULL.
+ * IMPORTANT: assumes ioJobsMutex is locked. */
+static IOJob_t* AIO_ReadPool_findNextWaitingOffsetCompletedJob_locked(ReadPoolCtx_t* ctx) {
+    IOJob_t *job = NULL;
+    int i;
+    /* This implementation goes through all completed jobs and looks for the one matching the next offset.
+     * While not strictly needed for a single threaded reader implementation (as in such a case we could expect
+     * reads to be completed in order) this implementation was chosen as it better fits other asyncio
+     * interfaces (such as io_uring) that do not provide promises regarding order of completion. */
+    for (i=0; i<ctx->completedJobsCount; i++) {
+        job = (IOJob_t *) ctx->completedJobs[i];
+        if (job->offset == ctx->waitingOnOffset) {
+            ctx->completedJobs[i] = ctx->completedJobs[--ctx->completedJobsCount];
+            return job;
+        }
+    }
+    return NULL;
+}
+
+/* AIO_ReadPool_numReadsInFlight:
+ * Returns the number of IO read jobs currently in flight. */
+static size_t AIO_ReadPool_numReadsInFlight(ReadPoolCtx_t* ctx) {
+    const int jobsHeld = (ctx->currentJobHeld==NULL ? 0 : 1);
+    return (size_t)(ctx->base.totalIoJobs - (ctx->base.availableJobsCount + ctx->completedJobsCount + jobsHeld));
+}
+
+/* AIO_ReadPool_getNextCompletedJob:
+ * Returns a completed IOJob_t for the next read in line based on waitingOnOffset and advances waitingOnOffset.
+ * Would block. */
+static IOJob_t* AIO_ReadPool_getNextCompletedJob(ReadPoolCtx_t* ctx) {
+    IOJob_t *job = NULL;
+    AIO_IOPool_lockJobsMutex(&ctx->base);
+
+    job = AIO_ReadPool_findNextWaitingOffsetCompletedJob_locked(ctx);
+
+    /* As long as we didn't find the job matching the next read, and we have some reads in flight continue waiting */
+    while (!job && (AIO_ReadPool_numReadsInFlight(ctx) > 0)) {
+        assert(ctx->base.threadPool != NULL); /* we shouldn't be here if we work in sync mode */
+        ZSTD_pthread_cond_wait(&ctx->jobCompletedCond, &ctx->base.ioJobsMutex);
+        job = AIO_ReadPool_findNextWaitingOffsetCompletedJob_locked(ctx);
+    }
+
+    if(job) {
+        assert(job->offset == ctx->waitingOnOffset);
+        ctx->waitingOnOffset += job->usedBufferSize;
+    }
+
+    AIO_IOPool_unlockJobsMutex(&ctx->base);
+    return job;
+}
+
+
+/* AIO_ReadPool_executeReadJob:
+ * Executes a read job synchronously. Can be used as a function for a thread pool. */
+static void AIO_ReadPool_executeReadJob(void* opaque){
+    IOJob_t* const job = (IOJob_t*) opaque;
+    ReadPoolCtx_t* const ctx = (ReadPoolCtx_t *)job->ctx;
+    if(ctx->reachedEof) {
+        job->usedBufferSize = 0;
+        AIO_ReadPool_addJobToCompleted(job);
+        return;
+    }
+    job->usedBufferSize = fread(job->buffer, 1, job->bufferSize, job->file);
+    if(job->usedBufferSize < job->bufferSize) {
+        if(ferror(job->file)) {
+            EXM_THROW(37, "Read error");
+        } else if(feof(job->file)) {
+            ctx->reachedEof = 1;
+        } else {
+            EXM_THROW(37, "Unexpected short read");
+        }
+    }
+    AIO_ReadPool_addJobToCompleted(job);
+}
+
+static void AIO_ReadPool_enqueueRead(ReadPoolCtx_t* ctx) {
+    IOJob_t* const job = AIO_IOPool_acquireJob(&ctx->base);
+    job->offset = ctx->nextReadOffset;
+    ctx->nextReadOffset += job->bufferSize;
+    AIO_IOPool_enqueueJob(job);
+}
+
+static void AIO_ReadPool_startReading(ReadPoolCtx_t* ctx) {
+    while(ctx->base.availableJobsCount) {
+        AIO_ReadPool_enqueueRead(ctx);
+    }
+}
+
+/* AIO_ReadPool_setFile:
+ * Sets the source file for future read in the pool. Initiates reading immediately if file is not NULL.
+ * Waits for all current enqueued tasks to complete if a previous file was set. */
+void AIO_ReadPool_setFile(ReadPoolCtx_t* ctx, FILE* file) {
+    assert(ctx!=NULL);
+    AIO_IOPool_join(&ctx->base);
+    AIO_ReadPool_releaseAllCompletedJobs(ctx);
+    if (ctx->currentJobHeld) {
+        AIO_IOPool_releaseIoJob((IOJob_t *)ctx->currentJobHeld);
+        ctx->currentJobHeld = NULL;
+    }
+    AIO_IOPool_setFile(&ctx->base, file);
+    ctx->nextReadOffset = 0;
+    ctx->waitingOnOffset = 0;
+    ctx->srcBuffer = ctx->coalesceBuffer;
+    ctx->srcBufferLoaded = 0;
+    ctx->reachedEof = 0;
+    if(file != NULL)
+        AIO_ReadPool_startReading(ctx);
+}
+
+/* AIO_ReadPool_create:
+ * Allocates and sets and a new readPool including its included jobs.
+ * bufferSize should be set to the maximal buffer we want to read at a time, will also be used
+ * as our basic read size. */
+ReadPoolCtx_t* AIO_ReadPool_create(const FIO_prefs_t* prefs, size_t bufferSize) {
+    ReadPoolCtx_t* const ctx = (ReadPoolCtx_t*) malloc(sizeof(ReadPoolCtx_t));
+    if(!ctx) EXM_THROW(100, "Allocation error : not enough memory");
+    AIO_IOPool_init(&ctx->base, prefs, AIO_ReadPool_executeReadJob, bufferSize);
+
+    ctx->coalesceBuffer = (U8*) malloc(bufferSize * 2);
+    if(!ctx->coalesceBuffer) EXM_THROW(100, "Allocation error : not enough memory");
+    ctx->srcBuffer = ctx->coalesceBuffer;
+    ctx->srcBufferLoaded = 0;
+    ctx->completedJobsCount = 0;
+    ctx->currentJobHeld = NULL;
+
+    if(ctx->base.threadPool)
+        if (ZSTD_pthread_cond_init(&ctx->jobCompletedCond, NULL))
+            EXM_THROW(103,"Failed creating jobCompletedCond cond");
+
+    return ctx;
+}
+
+/* AIO_ReadPool_free:
+ * Frees and releases a readPool and its resources. Closes source file. */
+void AIO_ReadPool_free(ReadPoolCtx_t* ctx) {
+    if(AIO_ReadPool_getFile(ctx))
+        AIO_ReadPool_closeFile(ctx);
+    if(ctx->base.threadPool)
+        ZSTD_pthread_cond_destroy(&ctx->jobCompletedCond);
+    AIO_IOPool_destroy(&ctx->base);
+    free(ctx->coalesceBuffer);
+    free(ctx);
+}
+
+/* AIO_ReadPool_consumeBytes:
+ * Consumes byes from srcBuffer's beginning and updates srcBufferLoaded accordingly. */
+void AIO_ReadPool_consumeBytes(ReadPoolCtx_t* ctx, size_t n) {
+    assert(n <= ctx->srcBufferLoaded);
+    ctx->srcBufferLoaded -= n;
+    ctx->srcBuffer += n;
+}
+
+/* AIO_ReadPool_releaseCurrentlyHeldAndGetNext:
+ * Release the current held job and get the next one, returns NULL if no next job available. */
+static IOJob_t* AIO_ReadPool_releaseCurrentHeldAndGetNext(ReadPoolCtx_t* ctx) {
+    if (ctx->currentJobHeld) {
+        AIO_IOPool_releaseIoJob((IOJob_t *)ctx->currentJobHeld);
+        ctx->currentJobHeld = NULL;
+        AIO_ReadPool_enqueueRead(ctx);
+    }
+    ctx->currentJobHeld = AIO_ReadPool_getNextCompletedJob(ctx);
+    return (IOJob_t*) ctx->currentJobHeld;
+}
+
+/* AIO_ReadPool_fillBuffer:
+ * Tries to fill the buffer with at least n or jobBufferSize bytes (whichever is smaller).
+ * Returns if srcBuffer has at least the expected number of bytes loaded or if we've reached the end of the file.
+ * Return value is the number of bytes added to the buffer.
+ * Note that srcBuffer might have up to 2 times jobBufferSize bytes. */
+size_t AIO_ReadPool_fillBuffer(ReadPoolCtx_t* ctx, size_t n) {
+    IOJob_t *job;
+    int useCoalesce = 0;
+    if(n > ctx->base.jobBufferSize)
+        n = ctx->base.jobBufferSize;
+
+    /* We are good, don't read anything */
+    if (ctx->srcBufferLoaded >= n)
+        return 0;
+
+    /* We still have bytes loaded, but not enough to satisfy caller. We need to get the next job
+     * and coalesce the remaining bytes with the next job's buffer */
+    if (ctx->srcBufferLoaded > 0) {
+        useCoalesce = 1;
+        memcpy(ctx->coalesceBuffer, ctx->srcBuffer, ctx->srcBufferLoaded);
+        ctx->srcBuffer = ctx->coalesceBuffer;
+    }
+
+    /* Read the next chunk */
+    job = AIO_ReadPool_releaseCurrentHeldAndGetNext(ctx);
+    if(!job)
+        return 0;
+    if(useCoalesce) {
+        assert(ctx->srcBufferLoaded + job->usedBufferSize <= 2*ctx->base.jobBufferSize);
+        memcpy(ctx->coalesceBuffer + ctx->srcBufferLoaded, job->buffer, job->usedBufferSize);
+        ctx->srcBufferLoaded += job->usedBufferSize;
+    }
+    else {
+        ctx->srcBuffer = (U8 *) job->buffer;
+        ctx->srcBufferLoaded = job->usedBufferSize;
+    }
+    return job->usedBufferSize;
+}
+
+/* AIO_ReadPool_consumeAndRefill:
+ * Consumes the current buffer and refills it with bufferSize bytes. */
+size_t AIO_ReadPool_consumeAndRefill(ReadPoolCtx_t* ctx) {
+    AIO_ReadPool_consumeBytes(ctx, ctx->srcBufferLoaded);
+    return AIO_ReadPool_fillBuffer(ctx, ctx->base.jobBufferSize);
+}
+
+/* AIO_ReadPool_getFile:
+ * Returns the current file set for the read pool. */
+FILE* AIO_ReadPool_getFile(const ReadPoolCtx_t* ctx) {
+    return AIO_IOPool_getFile(&ctx->base);
+}
+
+/* AIO_ReadPool_closeFile:
+ * Closes the current set file. Waits for all current enqueued tasks to complete and resets state. */
+int AIO_ReadPool_closeFile(ReadPoolCtx_t* ctx) {
+    FILE* const file = AIO_ReadPool_getFile(ctx);
+    AIO_ReadPool_setFile(ctx, NULL);
+    return fclose(file);
+}
+
+/* AIO_ReadPool_setAsync:
+ * Allows (de)activating async mode, to be used when the expected overhead
+ * of asyncio costs more than the expected gains. */
+void AIO_ReadPool_setAsync(ReadPoolCtx_t* ctx, int async) {
+    AIO_IOPool_setThreaded(&ctx->base, async);
+}
diff --git a/sys/contrib/zstd/programs/fileio_asyncio.h b/sys/contrib/zstd/programs/fileio_asyncio.h
new file mode 100644
index 000000000000..d4980ef5b01e
--- /dev/null
+++ b/sys/contrib/zstd/programs/fileio_asyncio.h
@@ -0,0 +1,195 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+ /*
+  * FileIO AsyncIO exposes read/write IO pools that allow doing IO asynchronously.
+  * Current implementation relies on having one thread that reads and one that
+  * writes.
+  * Each IO pool supports up to `MAX_IO_JOBS` that can be enqueued for work, but
+  * are performed serially by the appropriate worker thread.
+  * Most systems exposes better primitives to perform asynchronous IO, such as
+  * io_uring on newer linux systems. The API is built in such a way that in the
+  * future we could replace the threads with better solutions when available.
+  */
+
+#ifndef ZSTD_FILEIO_ASYNCIO_H
+#define ZSTD_FILEIO_ASYNCIO_H
+
+#include "../lib/common/mem.h"     /* U32, U64 */
+#include "fileio_types.h"
+#include "platform.h"
+#include "util.h"
+#include "../lib/common/pool.h"
+#include "../lib/common/threading.h"
+
+#define MAX_IO_JOBS          (10)
+
+typedef struct {
+    /* These struct fields should be set only on creation and not changed afterwards */
+    POOL_ctx* threadPool;
+    int threadPoolActive;
+    int totalIoJobs;
+    const FIO_prefs_t* prefs;
+    POOL_function poolFunction;
+
+    /* Controls the file we currently write to, make changes only by using provided utility functions */
+    FILE* file;
+
+    /* The jobs and availableJobsCount fields are accessed by both the main and worker threads and should
+     * only be mutated after locking the mutex */
+    ZSTD_pthread_mutex_t ioJobsMutex;
+    void* availableJobs[MAX_IO_JOBS];
+    int availableJobsCount;
+    size_t jobBufferSize;
+} IOPoolCtx_t;
+
+typedef struct {
+    IOPoolCtx_t base;
+
+    /* State regarding the currently read file */
+    int reachedEof;
+    U64 nextReadOffset;
+    U64 waitingOnOffset;
+
+    /* We may hold an IOJob object as needed if we actively expose its buffer. */
+    void *currentJobHeld;
+
+    /* Coalesce buffer is used to join two buffers in case where we need to read more bytes than left in
+     * the first of them. Shouldn't be accessed from outside ot utility functions. */
+    U8 *coalesceBuffer;
+
+    /* Read buffer can be used by consumer code, take care when copying this pointer aside as it might
+     * change when consuming / refilling buffer. */
+    U8 *srcBuffer;
+    size_t srcBufferLoaded;
+
+    /* We need to know what tasks completed so we can use their buffers when their time comes.
+     * Should only be accessed after locking base.ioJobsMutex . */
+    void* completedJobs[MAX_IO_JOBS];
+    int completedJobsCount;
+    ZSTD_pthread_cond_t jobCompletedCond;
+} ReadPoolCtx_t;
+
+typedef struct {
+    IOPoolCtx_t base;
+    unsigned storedSkips;
+} WritePoolCtx_t;
+
+typedef struct {
+    /* These fields are automatically set and shouldn't be changed by non WritePool code. */
+    void *ctx;
+    FILE* file;
+    void *buffer;
+    size_t bufferSize;
+
+    /* This field should be changed before a job is queued for execution and should contain the number
+     * of bytes to write from the buffer. */
+    size_t usedBufferSize;
+    U64 offset;
+} IOJob_t;
+
+/* AIO_supported:
+ * Returns 1 if AsyncIO is supported on the system, 0 otherwise. */
+int AIO_supported(void);
+
+
+/* AIO_WritePool_releaseIoJob:
+ * Releases an acquired job back to the pool. Doesn't execute the job. */
+void AIO_WritePool_releaseIoJob(IOJob_t *job);
+
+/* AIO_WritePool_acquireJob:
+ * Returns an available write job to be used for a future write. */
+IOJob_t* AIO_WritePool_acquireJob(WritePoolCtx_t *ctx);
+
+/* AIO_WritePool_enqueueAndReacquireWriteJob:
+ * Enqueues a write job for execution and acquires a new one.
+ * After execution `job`'s pointed value would change to the newly acquired job.
+ * Make sure to set `usedBufferSize` to the wanted length before call.
+ * The queued job shouldn't be used directly after queueing it. */
+void AIO_WritePool_enqueueAndReacquireWriteJob(IOJob_t **job);
+
+/* AIO_WritePool_sparseWriteEnd:
+ * Ends sparse writes to the current file.
+ * Blocks on completion of all current write jobs before executing. */
+void AIO_WritePool_sparseWriteEnd(WritePoolCtx_t *ctx);
+
+/* AIO_WritePool_setFile:
+ * Sets the destination file for future writes in the pool.
+ * Requires completion of all queues write jobs and release of all otherwise acquired jobs.
+ * Also requires ending of sparse write if a previous file was used in sparse mode. */
+void AIO_WritePool_setFile(WritePoolCtx_t *ctx, FILE* file);
+
+/* AIO_WritePool_getFile:
+ * Returns the file the writePool is currently set to write to. */
+FILE* AIO_WritePool_getFile(const WritePoolCtx_t* ctx);
+
+/* AIO_WritePool_closeFile:
+ * Ends sparse write and closes the writePool's current file and sets the file to NULL.
+ * Requires completion of all queues write jobs and release of all otherwise acquired jobs.  */
+int AIO_WritePool_closeFile(WritePoolCtx_t *ctx);
+
+/* AIO_WritePool_create:
+ * Allocates and sets and a new write pool including its included jobs.
+ * bufferSize should be set to the maximal buffer we want to write to at a time. */
+WritePoolCtx_t* AIO_WritePool_create(const FIO_prefs_t* prefs, size_t bufferSize);
+
+/* AIO_WritePool_free:
+ * Frees and releases a writePool and its resources. Closes destination file. */
+void AIO_WritePool_free(WritePoolCtx_t* ctx);
+
+/* AIO_WritePool_setAsync:
+ * Allows (de)activating async mode, to be used when the expected overhead
+ * of asyncio costs more than the expected gains. */
+void AIO_WritePool_setAsync(WritePoolCtx_t* ctx, int async);
+
+/* AIO_ReadPool_create:
+ * Allocates and sets and a new readPool including its included jobs.
+ * bufferSize should be set to the maximal buffer we want to read at a time, will also be used
+ * as our basic read size. */
+ReadPoolCtx_t* AIO_ReadPool_create(const FIO_prefs_t* prefs, size_t bufferSize);
+
+/* AIO_ReadPool_free:
+ * Frees and releases a readPool and its resources. Closes source file. */
+void AIO_ReadPool_free(ReadPoolCtx_t* ctx);
+
+/* AIO_ReadPool_setAsync:
+ * Allows (de)activating async mode, to be used when the expected overhead
+ * of asyncio costs more than the expected gains. */
+void AIO_ReadPool_setAsync(ReadPoolCtx_t* ctx, int async);
+
+/* AIO_ReadPool_consumeBytes:
+ * Consumes byes from srcBuffer's beginning and updates srcBufferLoaded accordingly. */
+void AIO_ReadPool_consumeBytes(ReadPoolCtx_t *ctx, size_t n);
+
+/* AIO_ReadPool_fillBuffer:
+ * Makes sure buffer has at least n bytes loaded (as long as n is not bigger than the initialized bufferSize).
+ * Returns if srcBuffer has at least n bytes loaded or if we've reached the end of the file.
+ * Return value is the number of bytes added to the buffer.
+ * Note that srcBuffer might have up to 2 times bufferSize bytes. */
+size_t AIO_ReadPool_fillBuffer(ReadPoolCtx_t *ctx, size_t n);
+
+/* AIO_ReadPool_consumeAndRefill:
+ * Consumes the current buffer and refills it with bufferSize bytes. */
+size_t AIO_ReadPool_consumeAndRefill(ReadPoolCtx_t *ctx);
+
+/* AIO_ReadPool_setFile:
+ * Sets the source file for future read in the pool. Initiates reading immediately if file is not NULL.
+ * Waits for all current enqueued tasks to complete if a previous file was set. */
+void AIO_ReadPool_setFile(ReadPoolCtx_t *ctx, FILE* file);
+
+/* AIO_ReadPool_getFile:
+ * Returns the current file set for the read pool. */
+FILE* AIO_ReadPool_getFile(const ReadPoolCtx_t *ctx);
+
+/* AIO_ReadPool_closeFile:
+ * Closes the current set file. Waits for all current enqueued tasks to complete and resets state. */
+int AIO_ReadPool_closeFile(ReadPoolCtx_t *ctx);
+
+#endif /* ZSTD_FILEIO_ASYNCIO_H */
diff --git a/sys/contrib/zstd/programs/fileio_common.h b/sys/contrib/zstd/programs/fileio_common.h
new file mode 100644
index 000000000000..8aa70edea813
--- /dev/null
+++ b/sys/contrib/zstd/programs/fileio_common.h
@@ -0,0 +1,121 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_FILEIO_COMMON_H
+#define ZSTD_FILEIO_COMMON_H
+
+#include "../lib/common/mem.h"     /* U32, U64 */
+#include "fileio_types.h"
+#include "platform.h"
+#include "timefn.h"     /* UTIL_getTime, UTIL_clockSpanMicro */
+
+/*-*************************************
+*  Macros
+***************************************/
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+#undef MAX
+#define MAX(a,b) ((a)>(b) ? (a) : (b))
+#undef MIN  /* in case it would be already defined */
+#define MIN(a,b)    ((a) < (b) ? (a) : (b))
+
+extern FIO_display_prefs_t g_display_prefs;
+
+#define DISPLAY_F(f, ...)    fprintf((f), __VA_ARGS__)
+#define DISPLAYOUT(...)      DISPLAY_F(stdout, __VA_ARGS__)
+#define DISPLAY(...)         DISPLAY_F(stderr, __VA_ARGS__)
+#define DISPLAYLEVEL(l, ...) { if (g_display_prefs.displayLevel>=l) { DISPLAY(__VA_ARGS__); } }
+
+extern UTIL_time_t g_displayClock;
+
+#define REFRESH_RATE  ((U64)(SEC_TO_MICRO / 6))
+#define READY_FOR_UPDATE() (UTIL_clockSpanMicro(g_displayClock) > REFRESH_RATE || g_display_prefs.displayLevel >= 4)
+#define DELAY_NEXT_UPDATE() { g_displayClock = UTIL_getTime(); }
+#define DISPLAYUPDATE(l, ...) {                              \
+        if (g_display_prefs.displayLevel>=l && (g_display_prefs.progressSetting != FIO_ps_never)) { \
+            if (READY_FOR_UPDATE()) { \
+                DELAY_NEXT_UPDATE();                         \
+                DISPLAY(__VA_ARGS__);                        \
+                if (g_display_prefs.displayLevel>=4) fflush(stderr);       \
+    }   }   }
+
+#define SHOULD_DISPLAY_SUMMARY() \
+    (g_display_prefs.displayLevel >= 2 || g_display_prefs.progressSetting == FIO_ps_always)
+#define SHOULD_DISPLAY_PROGRESS()                       \
+    (g_display_prefs.progressSetting != FIO_ps_never && SHOULD_DISPLAY_SUMMARY())
+#define DISPLAY_PROGRESS(...) { if (SHOULD_DISPLAY_PROGRESS()) { DISPLAYLEVEL(1, __VA_ARGS__); }}
+#define DISPLAYUPDATE_PROGRESS(...) { if (SHOULD_DISPLAY_PROGRESS()) { DISPLAYUPDATE(1, __VA_ARGS__); }}
+#define DISPLAY_SUMMARY(...) { if (SHOULD_DISPLAY_SUMMARY()) { DISPLAYLEVEL(1, __VA_ARGS__); } }
+
+#define EXM_THROW(error, ...)                                             \
+{                                                                         \
+    DISPLAYLEVEL(1, "zstd: ");                                            \
+    DISPLAYLEVEL(5, "Error defined at %s, line %i : \n", __FILE__, __LINE__); \
+    DISPLAYLEVEL(1, "error %i : ", error);                                \
+    DISPLAYLEVEL(1, __VA_ARGS__);                                         \
+    DISPLAYLEVEL(1, " \n");                                               \
+    exit(error);                                                          \
+}
+
+#define CHECK_V(v, f)                                \
+    v = f;                                           \
+    if (ZSTD_isError(v)) {                           \
+        DISPLAYLEVEL(5, "%s \n", #f);                \
+        EXM_THROW(11, "%s", ZSTD_getErrorName(v));   \
+    }
+#define CHECK(f) { size_t err; CHECK_V(err, f); }
+
+
+/* Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW */
+#if defined(LIBC_NO_FSEEKO)
+/* Some older libc implementations don't include these functions (e.g. Bionic < 24) */
+#   define LONG_SEEK fseek
+#   define LONG_TELL ftell
+#elif defined(_MSC_VER) && _MSC_VER >= 1400
+#   define LONG_SEEK _fseeki64
+#   define LONG_TELL _ftelli64
+#elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */
+#   define LONG_SEEK fseeko
+#   define LONG_TELL ftello
+#elif defined(__MINGW32__) && !defined(__STRICT_ANSI__) && !defined(__NO_MINGW_LFS) && defined(__MSVCRT__)
+#   define LONG_SEEK fseeko64
+#   define LONG_TELL ftello64
+#elif defined(_WIN32) && !defined(__DJGPP__)
+#   include <windows.h>
+    static int LONG_SEEK(FILE* file, __int64 offset, int origin) {
+        LARGE_INTEGER off;
+        DWORD method;
+        off.QuadPart = offset;
+        if (origin == SEEK_END)
+            method = FILE_END;
+        else if (origin == SEEK_CUR)
+            method = FILE_CURRENT;
+        else
+            method = FILE_BEGIN;
+
+        if (SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, NULL, method))
+            return 0;
+        else
+            return -1;
+    }
+    static __int64 LONG_TELL(FILE* file) {
+        LARGE_INTEGER off, newOff;
+        off.QuadPart = 0;
+        newOff.QuadPart = 0;
+        SetFilePointerEx((HANDLE) _get_osfhandle(_fileno(file)), off, &newOff, FILE_CURRENT);
+        return newOff.QuadPart;
+    }
+#else
+#   define LONG_SEEK fseek
+#   define LONG_TELL ftell
+#endif
+
+#endif /* ZSTD_FILEIO_COMMON_H */
diff --git a/sys/contrib/zstd/programs/fileio_types.h b/sys/contrib/zstd/programs/fileio_types.h
new file mode 100644
index 000000000000..23bda4168d83
--- /dev/null
+++ b/sys/contrib/zstd/programs/fileio_types.h
@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef FILEIO_TYPES_HEADER
+#define FILEIO_TYPES_HEADER
+
+#define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_compressionParameters */
+#include "../lib/zstd.h"           /* ZSTD_* */
+
+/*-*************************************
+*  Parameters: FIO_prefs_t
+***************************************/
+
+typedef struct FIO_display_prefs_s FIO_display_prefs_t;
+
+typedef enum { FIO_ps_auto, FIO_ps_never, FIO_ps_always } FIO_progressSetting_e;
+
+struct FIO_display_prefs_s {
+    int displayLevel;   /* 0 : no display;  1: errors;  2: + result + interaction + warnings;  3: + progression;  4: + information */
+    FIO_progressSetting_e progressSetting;
+};
+
+
+typedef enum { FIO_zstdCompression, FIO_gzipCompression, FIO_xzCompression, FIO_lzmaCompression, FIO_lz4Compression } FIO_compressionType_t;
+
+typedef struct FIO_prefs_s {
+
+    /* Algorithm preferences */
+    FIO_compressionType_t compressionType;
+    int sparseFileSupport;   /* 0: no sparse allowed; 1: auto (file yes, stdout no); 2: force sparse */
+    int dictIDFlag;
+    int checksumFlag;
+    int blockSize;
+    int overlapLog;
+    int adaptiveMode;
+    int useRowMatchFinder;
+    int rsyncable;
+    int minAdaptLevel;
+    int maxAdaptLevel;
+    int ldmFlag;
+    int ldmHashLog;
+    int ldmMinMatch;
+    int ldmBucketSizeLog;
+    int ldmHashRateLog;
+    size_t streamSrcSize;
+    size_t targetCBlockSize;
+    int srcSizeHint;
+    int testMode;
+    ZSTD_ParamSwitch_e literalCompressionMode;
+
+    /* IO preferences */
+    int removeSrcFile;
+    int overwrite;
+    int asyncIO;
+
+    /* Computation resources preferences */
+    unsigned memLimit;
+    int nbWorkers;
+
+    int excludeCompressedFiles;
+    int patchFromMode;
+    int contentSize;
+    int allowBlockDevices;
+    int passThrough;
+    ZSTD_ParamSwitch_e mmapDict;
+} FIO_prefs_t;
+
+typedef enum {FIO_mallocDict, FIO_mmapDict} FIO_dictBufferType_t;
+
+typedef struct {
+    void* dictBuffer;
+    size_t dictBufferSize;
+    FIO_dictBufferType_t dictBufferType;
+#if defined(_MSC_VER) || defined(_WIN32)
+    HANDLE dictHandle;
+#endif
+} FIO_Dict_t;
+
+#endif /* FILEIO_TYPES_HEADER */
diff --git a/sys/contrib/zstd/programs/lorem.c b/sys/contrib/zstd/programs/lorem.c
new file mode 100644
index 000000000000..79030c92f327
--- /dev/null
+++ b/sys/contrib/zstd/programs/lorem.c
@@ -0,0 +1,285 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* Implementation notes:
+ *
+ * This is a very simple lorem ipsum generator
+ * which features a static list of words
+ * and print them one after another randomly
+ * with a fake sentence / paragraph structure.
+ *
+ * The goal is to generate a printable text
+ * that can be used to fake a text compression scenario.
+ * The resulting compression / ratio curve of the lorem ipsum generator
+ * is more satisfying than the previous statistical generator,
+ * which was initially designed for entropy compression,
+ * and lacks a regularity more representative of text.
+ *
+ * The compression ratio achievable on the generated lorem ipsum
+ * is still a bit too good, presumably because the dictionary is a bit too
+ * small. It would be possible to create some more complex scheme, notably by
+ * enlarging the dictionary with a word generator, and adding grammatical rules
+ * (composition) and syntax rules. But that's probably overkill for the intended
+ * goal.
+ */
+
+#include "lorem.h"
+#include <assert.h>
+#include <limits.h> /* INT_MAX */
+#include <string.h> /* memcpy */
+
+#define WORD_MAX_SIZE 20
+
+/* Define the word pool */
+static const char* kWords[] = {
+    "lorem",        "ipsum",      "dolor",       "sit",          "amet",
+    "consectetur",  "adipiscing", "elit",        "sed",          "do",
+    "eiusmod",      "tempor",     "incididunt",  "ut",           "labore",
+    "et",           "dolore",     "magna",       "aliqua",       "dis",
+    "lectus",       "vestibulum", "mattis",      "ullamcorper",  "velit",
+    "commodo",      "a",          "lacus",       "arcu",         "magnis",
+    "parturient",   "montes",     "nascetur",    "ridiculus",    "mus",
+    "mauris",       "nulla",      "malesuada",   "pellentesque", "eget",
+    "gravida",      "in",         "dictum",      "non",          "erat",
+    "nam",          "voluptat",   "maecenas",    "blandit",      "aliquam",
+    "etiam",        "enim",       "lobortis",    "scelerisque",  "fermentum",
+    "dui",          "faucibus",   "ornare",      "at",           "elementum",
+    "eu",           "facilisis",  "odio",        "morbi",        "quis",
+    "eros",         "donec",      "ac",          "orci",         "purus",
+    "turpis",       "cursus",     "leo",         "vel",          "porta",
+    "consequat",    "interdum",   "varius",      "vulputate",    "aliquet",
+    "pharetra",     "nunc",       "auctor",      "urna",         "id",
+    "metus",        "viverra",    "nibh",        "cras",         "mi",
+    "unde",         "omnis",      "iste",        "natus",        "error",
+    "perspiciatis", "voluptatem", "accusantium", "doloremque",   "laudantium",
+    "totam",        "rem",        "aperiam",     "eaque",        "ipsa",
+    "quae",         "ab",         "illo",        "inventore",    "veritatis",
+    "quasi",        "architecto", "beatae",      "vitae",        "dicta",
+    "sunt",         "explicabo",  "nemo",        "ipsam",        "quia",
+    "voluptas",     "aspernatur", "aut",         "odit",         "fugit",
+    "consequuntur", "magni",      "dolores",     "eos",          "qui",
+    "ratione",      "sequi",      "nesciunt",    "neque",        "porro",
+    "quisquam",     "est",        "dolorem",     "adipisci",     "numquam",
+    "eius",         "modi",       "tempora",     "incidunt",     "magnam",
+    "quaerat",      "ad",         "minima",      "veniam",       "nostrum",
+    "ullam",        "corporis",   "suscipit",    "laboriosam",   "nisi",
+    "aliquid",      "ex",         "ea",          "commodi",      "consequatur",
+    "autem",        "eum",        "iure",        "voluptate",    "esse",
+    "quam",         "nihil",      "molestiae",   "illum",        "fugiat",
+    "quo",          "pariatur",   "vero",        "accusamus",    "iusto",
+    "dignissimos",  "ducimus",    "blanditiis",  "praesentium",  "voluptatum",
+    "deleniti",     "atque",      "corrupti",    "quos",         "quas",
+    "molestias",    "excepturi",  "sint",        "occaecati",    "cupiditate",
+    "provident",    "similique",  "culpa",       "officia",      "deserunt",
+    "mollitia",     "animi",      "laborum",     "dolorum",      "fuga",
+    "harum",        "quidem",     "rerum",       "facilis",      "expedita",
+    "distinctio",   "libero",     "tempore",     "cum",          "soluta",
+    "nobis",        "eligendi",   "optio",       "cumque",       "impedit",
+    "minus",        "quod",       "maxime",      "placeat",      "facere",
+    "possimus",     "assumenda",  "repellendus", "temporibus",   "quibusdam",
+    "officiis",     "debitis",    "saepe",       "eveniet",      "voluptates",
+    "repudiandae",  "recusandae", "itaque",      "earum",        "hic",
+    "tenetur",      "sapiente",   "delectus",    "reiciendis",   "cillum",
+    "maiores",      "alias",      "perferendis", "doloribus",    "asperiores",
+    "repellat",     "minim",      "nostrud",     "exercitation", "ullamco",
+    "laboris",      "aliquip",    "duis",        "aute",         "irure",
+};
+static const unsigned kNbWords = sizeof(kWords) / sizeof(kWords[0]);
+
+/* simple 1-dimension distribution, based on word's length, favors small words
+ */
+static const int kWeights[]    = { 0, 8, 6, 4, 3, 2 };
+static const size_t kNbWeights = sizeof(kWeights) / sizeof(kWeights[0]);
+
+#define DISTRIB_SIZE_MAX 650
+static int g_distrib[DISTRIB_SIZE_MAX] = { 0 };
+static unsigned g_distribCount         = 0;
+
+static void countFreqs(
+        const char* words[],
+        size_t nbWords,
+        const int* weights,
+        size_t nbWeights)
+{
+    unsigned total = 0;
+    size_t w;
+    for (w = 0; w < nbWords; w++) {
+        size_t len = strlen(words[w]);
+        int lmax;
+        if (len >= nbWeights)
+            len = nbWeights - 1;
+        lmax = weights[len];
+        total += (unsigned)lmax;
+    }
+    g_distribCount = total;
+    assert(g_distribCount <= DISTRIB_SIZE_MAX);
+}
+
+static void init_word_distrib(
+        const char* words[],
+        size_t nbWords,
+        const int* weights,
+        size_t nbWeights)
+{
+    size_t w, d = 0;
+    countFreqs(words, nbWords, weights, nbWeights);
+    for (w = 0; w < nbWords; w++) {
+        size_t len = strlen(words[w]);
+        int l, lmax;
+        if (len >= nbWeights)
+            len = nbWeights - 1;
+        lmax = weights[len];
+        for (l = 0; l < lmax; l++) {
+            g_distrib[d++] = (int)w;
+        }
+    }
+}
+
+/* Note: this unit only works when invoked sequentially.
+ * No concurrent access is allowed */
+static char* g_ptr         = NULL;
+static size_t g_nbChars    = 0;
+static size_t g_maxChars   = 10000000;
+static unsigned g_randRoot = 0;
+
+#define RDG_rotl32(x, r) ((x << r) | (x >> (32 - r)))
+static unsigned LOREM_rand(unsigned range)
+{
+    static const unsigned prime1 = 2654435761U;
+    static const unsigned prime2 = 2246822519U;
+    unsigned rand32              = g_randRoot;
+    rand32 *= prime1;
+    rand32 ^= prime2;
+    rand32     = RDG_rotl32(rand32, 13);
+    g_randRoot = rand32;
+    return (unsigned)(((unsigned long long)rand32 * range) >> 32);
+}
+
+static void writeLastCharacters(void)
+{
+    size_t lastChars = g_maxChars - g_nbChars;
+    assert(g_maxChars >= g_nbChars);
+    if (lastChars == 0)
+        return;
+    g_ptr[g_nbChars++] = '.';
+    if (lastChars > 2) {
+        memset(g_ptr + g_nbChars, ' ', lastChars - 2);
+    }
+    if (lastChars > 1) {
+        g_ptr[g_maxChars - 1] = '\n';
+    }
+    g_nbChars = g_maxChars;
+}
+
+static void generateWord(const char* word, const char* separator, int upCase)
+{
+    size_t const len = strlen(word) + strlen(separator);
+    if (g_nbChars + len > g_maxChars) {
+        writeLastCharacters();
+        return;
+    }
+    memcpy(g_ptr + g_nbChars, word, strlen(word));
+    if (upCase) {
+        static const char toUp = 'A' - 'a';
+        g_ptr[g_nbChars]       = (char)(g_ptr[g_nbChars] + toUp);
+    }
+    g_nbChars += strlen(word);
+    memcpy(g_ptr + g_nbChars, separator, strlen(separator));
+    g_nbChars += strlen(separator);
+}
+
+static int about(unsigned target)
+{
+    return (int)(LOREM_rand(target) + LOREM_rand(target) + 1);
+}
+
+/* Function to generate a random sentence */
+static void generateSentence(int nbWords)
+{
+    int commaPos       = about(9);
+    int comma2         = commaPos + about(7);
+    int qmark          = (LOREM_rand(11) == 7);
+    const char* endSep = qmark ? "? " : ". ";
+    int i;
+    for (i = 0; i < nbWords; i++) {
+        int const wordID       = g_distrib[LOREM_rand(g_distribCount)];
+        const char* const word = kWords[wordID];
+        const char* sep        = " ";
+        if (i == commaPos)
+            sep = ", ";
+        if (i == comma2)
+            sep = ", ";
+        if (i == nbWords - 1)
+            sep = endSep;
+        generateWord(word, sep, i == 0);
+    }
+}
+
+static void generateParagraph(int nbSentences)
+{
+    int i;
+    for (i = 0; i < nbSentences; i++) {
+        int wordsPerSentence = about(11);
+        generateSentence(wordsPerSentence);
+    }
+    if (g_nbChars < g_maxChars) {
+        g_ptr[g_nbChars++] = '\n';
+    }
+    if (g_nbChars < g_maxChars) {
+        g_ptr[g_nbChars++] = '\n';
+    }
+}
+
+/* It's "common" for lorem ipsum generators to start with the same first
+ * pre-defined sentence */
+static void generateFirstSentence(void)
+{
+    int i;
+    for (i = 0; i < 18; i++) {
+        const char* word      = kWords[i];
+        const char* separator = " ";
+        if (i == 4)
+            separator = ", ";
+        if (i == 7)
+            separator = ", ";
+        generateWord(word, separator, i == 0);
+    }
+    generateWord(kWords[18], ". ", 0);
+}
+
+size_t
+LOREM_genBlock(void* buffer, size_t size, unsigned seed, int first, int fill)
+{
+    g_ptr = (char*)buffer;
+    assert(size < INT_MAX);
+    g_maxChars = size;
+    g_nbChars  = 0;
+    g_randRoot = seed;
+    if (g_distribCount == 0) {
+        init_word_distrib(kWords, kNbWords, kWeights, kNbWeights);
+    }
+
+    if (first) {
+        generateFirstSentence();
+    }
+    while (g_nbChars < g_maxChars) {
+        int sentencePerParagraph = about(7);
+        generateParagraph(sentencePerParagraph);
+        if (!fill)
+            break; /* only generate one paragraph in not-fill mode */
+    }
+    g_ptr = NULL;
+    return g_nbChars;
+}
+
+void LOREM_genBuffer(void* buffer, size_t size, unsigned seed)
+{
+    LOREM_genBlock(buffer, size, seed, 1, 1);
+}
diff --git a/sys/contrib/zstd/programs/lorem.h b/sys/contrib/zstd/programs/lorem.h
new file mode 100644
index 000000000000..4a87f8748a5f
--- /dev/null
+++ b/sys/contrib/zstd/programs/lorem.h
@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+/* lorem ipsum generator */
+
+#include <stddef.h>   /* size_t */
+
+/*
+ * LOREM_genBuffer():
+ * Generate @size bytes of compressible data using lorem ipsum generator
+ * into provided @buffer.
+ */
+void LOREM_genBuffer(void* buffer, size_t size, unsigned seed);
+
+/*
+ * LOREM_genBlock():
+ * Similar to LOREM_genBuffer, with additional controls :
+ * - @first : generate the first sentence
+ * - @fill : fill the entire @buffer,
+ *           if ==0: generate one paragraph at most.
+ * @return : nb of bytes generated into @buffer.
+ */
+size_t LOREM_genBlock(void* buffer, size_t size,
+                      unsigned seed,
+                      int first, int fill);
diff --git a/sys/contrib/zstd/programs/platform.h b/sys/contrib/zstd/programs/platform.h
index b858e3b484c2..e2cc1c3e65f6 100644
--- a/sys/contrib/zstd/programs/platform.h
+++ b/sys/contrib/zstd/programs/platform.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -11,12 +11,6 @@
 #ifndef PLATFORM_H_MODULE
 #define PLATFORM_H_MODULE
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-
-
 /* **************************************
 *  Compiler Options
 ****************************************/
@@ -33,12 +27,12 @@ extern "C" {
 
 /* **************************************
 *  Detect 64-bit OS
-*  http://nadeausoftware.com/articles/2012/02/c_c_tip_how_detect_processor_type_using_compiler_predefined_macros
+*  https://nadeausoftware.com/articles/2012/02/c_c_tip_how_detect_processor_type_using_compiler_predefined_macros
 ****************************************/
 #if defined __ia64 || defined _M_IA64                                                                               /* Intel Itanium */ \
   || defined __powerpc64__ || defined __ppc64__ || defined __PPC64__                                                /* POWER 64-bit */  \
   || (defined __sparc && (defined __sparcv9 || defined __sparc_v9__ || defined __arch64__)) || defined __sparc64__  /* SPARC 64-bit */  \
-  || defined __x86_64__s || defined _M_X64                                                                          /* x86 64-bit */    \
+  || defined __x86_64__ || defined _M_X64                                                                           /* x86 64-bit */    \
   || defined __arm64__ || defined __aarch64__ || defined __ARM64_ARCH_8__                                           /* ARM 64-bit */    \
   || (defined __mips  && (__mips == 64 || __mips == 4 || __mips == 3))                                              /* MIPS 64-bit */   \
   || defined _LP64 || defined __LP64__ /* NetBSD, OpenBSD */ || defined __64BIT__ /* AIX */ || defined _ADDR64 /* Cray */               \
@@ -74,13 +68,12 @@ extern "C" {
 ***************************************************************/
 #ifndef PLATFORM_POSIX_VERSION
 
-#  if (defined(__APPLE__) && defined(__MACH__)) || defined(__SVR4) || defined(_AIX) || defined(__hpux) /* POSIX.1-2001 (SUSv3) conformant */ \
-     || defined(__DragonFly__) || defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__)  /* BSD distros */
+#  if (defined(__APPLE__) && defined(__MACH__)) || defined(__SVR4) || defined(_AIX) || defined(__hpux) /* POSIX.1-2001 (SUSv3) conformant */
      /* exception rule : force posix version to 200112L,
       * note: it's better to use unistd.h's _POSIX_VERSION whenever possible */
 #    define PLATFORM_POSIX_VERSION 200112L
 
-/* try to determine posix version through official unistd.h's _POSIX_VERSION (http://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html).
+/* try to determine posix version through official unistd.h's _POSIX_VERSION (https://pubs.opengroup.org/onlinepubs/7908799/xsh/unistd.h.html).
  * note : there is no simple way to know in advance if <unistd.h> is present or not on target system,
  * Posix specification mandates its presence and its content, but target system must respect this spec.
  * It's necessary to _not_ #include <unistd.h> whenever target OS is not unix-like
@@ -89,7 +82,7 @@ extern "C" {
  */
 #  elif !defined(_WIN32) \
      && ( defined(__unix__) || defined(__unix) \
-       || defined(__midipix__) || defined(__VMS) || defined(__HAIKU__) )
+       || defined(_QNX_SOURCE) || defined(__midipix__) || defined(__VMS) || defined(__HAIKU__) )
 
 #    if defined(__linux__) || defined(__linux) || defined(__CYGWIN__)
 #      ifndef _POSIX_C_SOURCE
@@ -127,6 +120,10 @@ extern "C" {
 
 /*-*********************************************
 *  Detect if isatty() and fileno() are available
+*
+*  Note: Use UTIL_isConsole() for the zstd CLI
+*  instead, as it allows faking is console for
+*  testing.
 ************************************************/
 #if (defined(__linux__) && (PLATFORM_POSIX_VERSION > 1)) \
  || (PLATFORM_POSIX_VERSION >= 200112L) \
@@ -137,14 +134,24 @@ extern "C" {
 #elif defined(MSDOS) || defined(OS2)
 #  include <io.h>       /* _isatty */
 #  define IS_CONSOLE(stdStream) _isatty(_fileno(stdStream))
-#elif defined(WIN32) || defined(_WIN32)
+#elif defined(_WIN32)
 #  include <io.h>      /* _isatty */
 #  include <windows.h> /* DeviceIoControl, HANDLE, FSCTL_SET_SPARSE */
 #  include <stdio.h>   /* FILE */
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
 static __inline int IS_CONSOLE(FILE* stdStream) {
     DWORD dummy;
     return _isatty(_fileno(stdStream)) && GetConsoleMode((HANDLE)_get_osfhandle(_fileno(stdStream)), &dummy);
 }
+
+#if defined (__cplusplus)
+}
+#endif
+
 #else
 #  define IS_CONSOLE(stdStream) 0
 #endif
@@ -153,7 +160,7 @@ static __inline int IS_CONSOLE(FILE* stdStream) {
 /******************************
 *  OS-specific IO behaviors
 ******************************/
-#if defined(MSDOS) || defined(OS2) || defined(WIN32) || defined(_WIN32)
+#if defined(MSDOS) || defined(OS2) || defined(_WIN32)
 #  include <fcntl.h>   /* _O_BINARY */
 #  include <io.h>      /* _setmode, _fileno, _get_osfhandle */
 #  if !defined(__DJGPP__)
@@ -192,13 +199,13 @@ static __inline int IS_CONSOLE(FILE* stdStream) {
 
 
 #ifndef ZSTD_SETPRIORITY_SUPPORT
-   /* mandates presence of <sys/resource.h> and support for setpriority() : http://man7.org/linux/man-pages/man2/setpriority.2.html */
+   /* mandates presence of <sys/resource.h> and support for setpriority() : https://man7.org/linux/man-pages/man2/setpriority.2.html */
 #  define ZSTD_SETPRIORITY_SUPPORT (PLATFORM_POSIX_VERSION >= 200112L)
 #endif
 
 
 #ifndef ZSTD_NANOSLEEP_SUPPORT
-   /* mandates support of nanosleep() within <time.h> : http://man7.org/linux/man-pages/man2/nanosleep.2.html */
+   /* mandates support of nanosleep() within <time.h> : https://man7.org/linux/man-pages/man2/nanosleep.2.html */
 #  if (defined(__linux__) && (PLATFORM_POSIX_VERSION >= 199309L)) \
    || (PLATFORM_POSIX_VERSION >= 200112L)
 #     define ZSTD_NANOSLEEP_SUPPORT 1
@@ -207,9 +214,4 @@ static __inline int IS_CONSOLE(FILE* stdStream) {
 #  endif
 #endif
 
-
-#if defined (__cplusplus)
-}
-#endif
-
 #endif /* PLATFORM_H_MODULE */
diff --git a/sys/contrib/zstd/programs/timefn.c b/sys/contrib/zstd/programs/timefn.c
index 64577b0e932a..4f045226b8bf 100644
--- a/sys/contrib/zstd/programs/timefn.c
+++ b/sys/contrib/zstd/programs/timefn.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -12,7 +12,8 @@
 /* ===  Dependencies  === */
 
 #include "timefn.h"
-
+#include "platform.h" /* set _POSIX_C_SOURCE */
+#include <time.h>     /* CLOCK_MONOTONIC, TIME_UTC */
 
 /*-****************************************
 *  Time functions
@@ -20,12 +21,11 @@
 
 #if defined(_WIN32)   /* Windows */
 
+#include <windows.h>  /* LARGE_INTEGER */
 #include <stdlib.h>   /* abort */
 #include <stdio.h>    /* perror */
 
-UTIL_time_t UTIL_getTime(void) { UTIL_time_t x; QueryPerformanceCounter(&x); return x; }
-
-PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+UTIL_time_t UTIL_getTime(void)
 {
     static LARGE_INTEGER ticksPerSecond;
     static int init = 0;
@@ -36,30 +36,20 @@ PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
         }
         init = 1;
     }
-    return 1000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
-}
-
-PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
-{
-    static LARGE_INTEGER ticksPerSecond;
-    static int init = 0;
-    if (!init) {
-        if (!QueryPerformanceFrequency(&ticksPerSecond)) {
-            perror("timefn::QueryPerformanceFrequency");
-            abort();
-        }
-        init = 1;
+    {   UTIL_time_t r;
+        LARGE_INTEGER x;
+        QueryPerformanceCounter(&x);
+        r.t = (PTime)(x.QuadPart * 1000000000ULL / ticksPerSecond.QuadPart);
+        return r;
     }
-    return 1000000000ULL*(clockEnd.QuadPart - clockStart.QuadPart)/ticksPerSecond.QuadPart;
 }
 
 
-
 #elif defined(__APPLE__) && defined(__MACH__)
 
-UTIL_time_t UTIL_getTime(void) { return mach_absolute_time(); }
+#include <mach/mach_time.h> /* mach_timebase_info_data_t, mach_timebase_info, mach_absolute_time */
 
-PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+UTIL_time_t UTIL_getTime(void)
 {
     static mach_timebase_info_data_t rate;
     static int init = 0;
@@ -67,23 +57,39 @@ PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd)
         mach_timebase_info(&rate);
         init = 1;
     }
-    return (((clockEnd - clockStart) * (PTime)rate.numer) / ((PTime)rate.denom))/1000ULL;
+    {   UTIL_time_t r;
+        r.t = mach_absolute_time() * (PTime)rate.numer / (PTime)rate.denom;
+        return r;
+    }
 }
 
-PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+/* POSIX.1-2001 (optional) */
+#elif defined(CLOCK_MONOTONIC)
+
+#include <stdlib.h>   /* abort */
+#include <stdio.h>    /* perror */
+
+UTIL_time_t UTIL_getTime(void)
 {
-    static mach_timebase_info_data_t rate;
-    static int init = 0;
-    if (!init) {
-        mach_timebase_info(&rate);
-        init = 1;
+    /* time must be initialized, othersize it may fail msan test.
+     * No good reason, likely a limitation of timespec_get() for some target */
+    struct timespec time = { 0, 0 };
+    if (clock_gettime(CLOCK_MONOTONIC, &time) != 0) {
+        perror("timefn::clock_gettime(CLOCK_MONOTONIC)");
+        abort();
+    }
+    {   UTIL_time_t r;
+        r.t = (PTime)time.tv_sec * 1000000000ULL + (PTime)time.tv_nsec;
+        return r;
     }
-    return ((clockEnd - clockStart) * (PTime)rate.numer) / ((PTime)rate.denom);
 }
 
 
-/* C11 requires timespec_get, but FreeBSD 11 lacks it, while still claiming C11 compliance.
-   Android also lacks it but does define TIME_UTC. */
+/* C11 requires support of timespec_get().
+ * However, FreeBSD 11 claims C11 compliance while lacking timespec_get().
+ * Double confirm timespec_get() support by checking the definition of TIME_UTC.
+ * However, some versions of Android manage to simultaneously define TIME_UTC
+ * and lack timespec_get() support... */
 #elif (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) /* C11 */) \
     && defined(TIME_UTC) && !defined(__ANDROID__)
 
@@ -94,65 +100,49 @@ UTIL_time_t UTIL_getTime(void)
 {
     /* time must be initialized, othersize it may fail msan test.
      * No good reason, likely a limitation of timespec_get() for some target */
-    UTIL_time_t time = UTIL_TIME_INITIALIZER;
+    struct timespec time = { 0, 0 };
     if (timespec_get(&time, TIME_UTC) != TIME_UTC) {
-        perror("timefn::timespec_get");
+        perror("timefn::timespec_get(TIME_UTC)");
         abort();
     }
-    return time;
-}
-
-static UTIL_time_t UTIL_getSpanTime(UTIL_time_t begin, UTIL_time_t end)
-{
-    UTIL_time_t diff;
-    if (end.tv_nsec < begin.tv_nsec) {
-        diff.tv_sec = (end.tv_sec - 1) - begin.tv_sec;
-        diff.tv_nsec = (end.tv_nsec + 1000000000ULL) - begin.tv_nsec;
-    } else {
-        diff.tv_sec = end.tv_sec - begin.tv_sec;
-        diff.tv_nsec = end.tv_nsec - begin.tv_nsec;
+    {   UTIL_time_t r;
+        r.t = (PTime)time.tv_sec * 1000000000ULL + (PTime)time.tv_nsec;
+        return r;
     }
-    return diff;
 }
 
-PTime UTIL_getSpanTimeMicro(UTIL_time_t begin, UTIL_time_t end)
-{
-    UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
-    PTime micro = 0;
-    micro += 1000000ULL * diff.tv_sec;
-    micro += diff.tv_nsec / 1000ULL;
-    return micro;
-}
 
-PTime UTIL_getSpanTimeNano(UTIL_time_t begin, UTIL_time_t end)
+#else   /* relies on standard C90 (note : clock_t produces wrong measurements for multi-threaded workloads) */
+
+UTIL_time_t UTIL_getTime(void)
 {
-    UTIL_time_t const diff = UTIL_getSpanTime(begin, end);
-    PTime nano = 0;
-    nano += 1000000000ULL * diff.tv_sec;
-    nano += diff.tv_nsec;
-    return nano;
+    UTIL_time_t r;
+    r.t = (PTime)clock() * 1000000000ULL / CLOCKS_PER_SEC;
+    return r;
 }
 
-
-
-#else   /* relies on standard C90 (note : clock_t measurements can be wrong when using multi-threading) */
-
-UTIL_time_t UTIL_getTime(void) { return clock(); }
-PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
-PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd) { return 1000000000ULL * (clockEnd - clockStart) / CLOCKS_PER_SEC; }
+#define TIME_MT_MEASUREMENTS_NOT_SUPPORTED
 
 #endif
 
+/* ==== Common functions, valid for all time API ==== */
 
+PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
+{
+    return clockEnd.t - clockStart.t;
+}
+
+PTime UTIL_getSpanTimeMicro(UTIL_time_t begin, UTIL_time_t end)
+{
+    return UTIL_getSpanTimeNano(begin, end) / 1000ULL;
+}
 
-/* returns time span in microseconds */
 PTime UTIL_clockSpanMicro(UTIL_time_t clockStart )
 {
     UTIL_time_t const clockEnd = UTIL_getTime();
     return UTIL_getSpanTimeMicro(clockStart, clockEnd);
 }
 
-/* returns time span in microseconds */
 PTime UTIL_clockSpanNano(UTIL_time_t clockStart )
 {
     UTIL_time_t const clockEnd = UTIL_getTime();
@@ -167,3 +157,12 @@ void UTIL_waitForNextTick(void)
         clockEnd = UTIL_getTime();
     } while (UTIL_getSpanTimeNano(clockStart, clockEnd) == 0);
 }
+
+int UTIL_support_MT_measurements(void)
+{
+# if defined(TIME_MT_MEASUREMENTS_NOT_SUPPORTED)
+    return 0;
+# else
+    return 1;
+# endif
+}
diff --git a/sys/contrib/zstd/programs/timefn.h b/sys/contrib/zstd/programs/timefn.h
index 3fcd78a28ec1..80f72e228a35 100644
--- a/sys/contrib/zstd/programs/timefn.h
+++ b/sys/contrib/zstd/programs/timefn.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -11,79 +11,49 @@
 #ifndef TIME_FN_H_MODULE_287987
 #define TIME_FN_H_MODULE_287987
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-
-/*-****************************************
-*  Dependencies
-******************************************/
-#include <time.h>         /* clock_t, clock, CLOCKS_PER_SEC */
-
-
-
 /*-****************************************
-*  Local Types
+*  Types
 ******************************************/
 
 #if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
 # if defined(_AIX)
 #  include <inttypes.h>
 # else
-#  include <stdint.h> /* intptr_t */
+#  include <stdint.h> /* uint64_t */
 # endif
   typedef uint64_t           PTime;  /* Precise Time */
 #else
   typedef unsigned long long PTime;  /* does not support compilers without long long support */
 #endif
 
+/* UTIL_time_t contains a nanosecond time counter.
+ * The absolute value is not meaningful.
+ * It's only valid to compute the difference between 2 measurements. */
+typedef struct { PTime t; } UTIL_time_t;
+#define UTIL_TIME_INITIALIZER { 0 }
 
 
 /*-****************************************
 *  Time functions
 ******************************************/
-#if defined(_WIN32)   /* Windows */
-
-    #include <windows.h>   /* LARGE_INTEGER */
-    typedef LARGE_INTEGER UTIL_time_t;
-    #define UTIL_TIME_INITIALIZER { { 0, 0 } }
-
-#elif defined(__APPLE__) && defined(__MACH__)
-
-    #include <mach/mach_time.h>
-    typedef PTime UTIL_time_t;
-    #define UTIL_TIME_INITIALIZER 0
-
-/* C11 requires timespec_get, but FreeBSD 11 lacks it, while still claiming C11 compliance.
-   Android also lacks it but does define TIME_UTC. */
-#elif (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) /* C11 */) \
-    && defined(TIME_UTC) && !defined(__ANDROID__)
-
-    typedef struct timespec UTIL_time_t;
-    #define UTIL_TIME_INITIALIZER { 0, 0 }
-
-#else   /* relies on standard C90 (note : clock_t measurements can be wrong when using multi-threading) */
 
-    typedef clock_t UTIL_time_t;
-    #define UTIL_TIME_INITIALIZER 0
-
-#endif
+UTIL_time_t UTIL_getTime(void);
 
+/* Timer resolution can be low on some platforms.
+ * To improve accuracy, it's recommended to wait for a new tick
+ * before starting benchmark measurements */
+void UTIL_waitForNextTick(void);
+/* tells if timefn will return correct time measurements
+ * in presence of multi-threaded workload.
+ * note : this is not the case if only C90 clock_t measurements are available */
+int UTIL_support_MT_measurements(void);
 
-UTIL_time_t UTIL_getTime(void);
-PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd);
 PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd);
-
-#define SEC_TO_MICRO ((PTime)1000000)
-PTime UTIL_clockSpanMicro(UTIL_time_t clockStart);
 PTime UTIL_clockSpanNano(UTIL_time_t clockStart);
 
-void UTIL_waitForNextTick(void);
-
+PTime UTIL_getSpanTimeMicro(UTIL_time_t clockStart, UTIL_time_t clockEnd);
+PTime UTIL_clockSpanMicro(UTIL_time_t clockStart);
 
-#if defined (__cplusplus)
-}
-#endif
+#define SEC_TO_MICRO ((PTime)1000000)  /* nb of microseconds in a second */
 
 #endif /* TIME_FN_H_MODULE_287987 */
diff --git a/sys/contrib/zstd/programs/util.c b/sys/contrib/zstd/programs/util.c
index d69b72a37ca6..065a35855f27 100644
--- a/sys/contrib/zstd/programs/util.c
+++ b/sys/contrib/zstd/programs/util.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -8,11 +8,6 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-
 /*-****************************************
 *  Dependencies
 ******************************************/
@@ -23,16 +18,27 @@ extern "C" {
 #include <errno.h>
 #include <assert.h>
 
+#if defined(__FreeBSD__)
+#include <sys/param.h> /* __FreeBSD_version */
+#endif /* #ifdef __FreeBSD__ */
+
 #if defined(_WIN32)
 #  include <sys/utime.h>  /* utime */
 #  include <io.h>         /* _chmod */
+#  define ZSTD_USE_UTIMENSAT 0
 #else
 #  include <unistd.h>     /* chown, stat */
-#  if PLATFORM_POSIX_VERSION < 200809L || !defined(st_mtime)
-#    include <utime.h>    /* utime */
+#  include <sys/stat.h>   /* utimensat, st_mtime */
+#  if (PLATFORM_POSIX_VERSION >= 200809L && defined(st_mtime)) \
+      || (defined(__FreeBSD__) && __FreeBSD_version >= 1100056)
+#    define ZSTD_USE_UTIMENSAT 1
 #  else
+#    define ZSTD_USE_UTIMENSAT 0
+#  endif
+#  if ZSTD_USE_UTIMENSAT
 #    include <fcntl.h>    /* AT_FDCWD */
-#    include <sys/stat.h> /* utimensat */
+#  else
+#    include <utime.h>    /* utime */
 #  endif
 #endif
 
@@ -66,6 +72,27 @@ extern "C" {
 #define UTIL_DISPLAY(...)         fprintf(stderr, __VA_ARGS__)
 #define UTIL_DISPLAYLEVEL(l, ...) { if (g_utilDisplayLevel>=l) { UTIL_DISPLAY(__VA_ARGS__); } }
 
+static int g_traceDepth = 0;
+int g_traceFileStat = 0;
+
+#define UTIL_TRACE_CALL(...)                                         \
+    {                                                                \
+        if (g_traceFileStat) {                                       \
+            UTIL_DISPLAY("Trace:FileStat: %*s> ", g_traceDepth, ""); \
+            UTIL_DISPLAY(__VA_ARGS__);                               \
+            UTIL_DISPLAY("\n");                                      \
+            ++g_traceDepth;                                          \
+        }                                                            \
+    }
+
+#define UTIL_TRACE_RET(ret)                                                     \
+    {                                                                           \
+        if (g_traceFileStat) {                                                  \
+            --g_traceDepth;                                                     \
+            UTIL_DISPLAY("Trace:FileStat: %*s< %d\n", g_traceDepth, "", (ret)); \
+        }                                                                      \
+    }
+
 /* A modified version of realloc().
  * If UTIL_realloc() fails the original block is freed.
  */
@@ -81,6 +108,17 @@ UTIL_STATIC void* UTIL_realloc(void *ptr, size_t size)
     #define chmod _chmod
 #endif
 
+#ifndef ZSTD_HAVE_FCHMOD
+#if PLATFORM_POSIX_VERSION >= 199309L
+#define ZSTD_HAVE_FCHMOD
+#endif
+#endif
+
+#ifndef ZSTD_HAVE_FCHOWN
+#if PLATFORM_POSIX_VERSION >= 200809L
+#define ZSTD_HAVE_FCHOWN
+#endif
+#endif
 
 /*-****************************************
 *  Console log
@@ -100,7 +138,7 @@ int UTIL_requireUserConfirmation(const char* prompt, const char* abortMsg,
     ch = getchar();
     result = 0;
     if (strchr(acceptableLetters, ch) == NULL) {
-        UTIL_DISPLAY("%s", abortMsg);
+        UTIL_DISPLAY("%s \n", abortMsg);
         result = 1;
     }
     /* flush the rest */
@@ -121,21 +159,51 @@ int UTIL_requireUserConfirmation(const char* prompt, const char* abortMsg,
 *  Functions
 ***************************************/
 
-int UTIL_stat(const char* filename, stat_t* statbuf)
+void UTIL_traceFileStat(void)
 {
+    g_traceFileStat = 1;
+}
+
+int UTIL_fstat(const int fd, const char* filename, stat_t* statbuf)
+{
+    int ret;
+    UTIL_TRACE_CALL("UTIL_stat(%d, %s)", fd, filename);
 #if defined(_MSC_VER)
-    return !_stat64(filename, statbuf);
+    if (fd >= 0) {
+        ret = !_fstat64(fd, statbuf);
+    } else {
+        ret = !_stat64(filename, statbuf);
+    }
 #elif defined(__MINGW32__) && defined (__MSVCRT__)
-    return !_stati64(filename, statbuf);
+    if (fd >= 0) {
+        ret = !_fstati64(fd, statbuf);
+    } else {
+        ret = !_stati64(filename, statbuf);
+    }
 #else
-    return !stat(filename, statbuf);
+    if (fd >= 0) {
+        ret = !fstat(fd, statbuf);
+    } else {
+        ret = !stat(filename, statbuf);
+    }
 #endif
+    UTIL_TRACE_RET(ret);
+    return ret;
+}
+
+int UTIL_stat(const char* filename, stat_t* statbuf)
+{
+    return UTIL_fstat(-1, filename, statbuf);
 }
 
 int UTIL_isRegularFile(const char* infilename)
 {
     stat_t statbuf;
-    return UTIL_stat(infilename, &statbuf) && UTIL_isRegularFileStat(&statbuf);
+    int ret;
+    UTIL_TRACE_CALL("UTIL_isRegularFile(%s)", infilename);
+    ret = UTIL_stat(infilename, &statbuf) && UTIL_isRegularFileStat(&statbuf);
+    UTIL_TRACE_RET(ret);
+    return ret;
 }
 
 int UTIL_isRegularFileStat(const stat_t* statbuf)
@@ -150,72 +218,152 @@ int UTIL_isRegularFileStat(const stat_t* statbuf)
 /* like chmod, but avoid changing permission of /dev/null */
 int UTIL_chmod(char const* filename, const stat_t* statbuf, mode_t permissions)
 {
+    return UTIL_fchmod(-1, filename, statbuf, permissions);
+}
+
+int UTIL_fchmod(const int fd, char const* filename, const stat_t* statbuf, mode_t permissions)
+{
     stat_t localStatBuf;
+    UTIL_TRACE_CALL("UTIL_chmod(%s, %#4o)", filename, (unsigned)permissions);
     if (statbuf == NULL) {
-        if (!UTIL_stat(filename, &localStatBuf)) return 0;
+        if (!UTIL_fstat(fd, filename, &localStatBuf)) {
+            UTIL_TRACE_RET(0);
+            return 0;
+        }
         statbuf = &localStatBuf;
     }
-    if (!UTIL_isRegularFileStat(statbuf)) return 0; /* pretend success, but don't change anything */
-    return chmod(filename, permissions);
+    if (!UTIL_isRegularFileStat(statbuf)) {
+        UTIL_TRACE_RET(0);
+        return 0; /* pretend success, but don't change anything */
+    }
+#ifdef ZSTD_HAVE_FCHMOD
+    if (fd >= 0) {
+        int ret;
+        UTIL_TRACE_CALL("fchmod");
+        ret = fchmod(fd, permissions);
+        UTIL_TRACE_RET(ret);
+        UTIL_TRACE_RET(ret);
+        return ret;
+    } else
+#endif
+    {
+        int ret;
+        UTIL_TRACE_CALL("chmod");
+        ret = chmod(filename, permissions);
+        UTIL_TRACE_RET(ret);
+        UTIL_TRACE_RET(ret);
+        return ret;
+    }
 }
 
 /* set access and modification times */
 int UTIL_utime(const char* filename, const stat_t *statbuf)
 {
     int ret;
+    UTIL_TRACE_CALL("UTIL_utime(%s)", filename);
     /* We check that st_mtime is a macro here in order to give us confidence
      * that struct stat has a struct timespec st_mtim member. We need this
      * check because there are some platforms that claim to be POSIX 2008
      * compliant but which do not have st_mtim... */
-#if (PLATFORM_POSIX_VERSION >= 200809L) && defined(st_mtime)
-    /* (atime, mtime) */
-    struct timespec timebuf[2] = { {0, UTIME_NOW} };
-    timebuf[1] = statbuf->st_mtim;
-    ret = utimensat(AT_FDCWD, filename, timebuf, 0);
+    /* FreeBSD has implemented POSIX 2008 for a long time but still only
+     * advertises support for POSIX 2001. They have a version macro that
+     * lets us safely gate them in.
+     * See https://docs.freebsd.org/en/books/porters-handbook/versions/.
+     */
+#if ZSTD_USE_UTIMENSAT
+    {
+        /* (atime, mtime) */
+        struct timespec timebuf[2] = { {0, UTIME_NOW} };
+        timebuf[1] = statbuf->st_mtim;
+        ret = utimensat(AT_FDCWD, filename, timebuf, 0);
+    }
 #else
-    struct utimbuf timebuf;
-    timebuf.actime = time(NULL);
-    timebuf.modtime = statbuf->st_mtime;
-    ret = utime(filename, &timebuf);
+    {
+        struct utimbuf timebuf;
+        timebuf.actime = time(NULL);
+        timebuf.modtime = statbuf->st_mtime;
+        ret = utime(filename, &timebuf);
+    }
 #endif
     errno = 0;
+    UTIL_TRACE_RET(ret);
     return ret;
 }
 
 int UTIL_setFileStat(const char *filename, const stat_t *statbuf)
 {
-    int res = 0;
+    return UTIL_setFDStat(-1, filename, statbuf);
+}
 
+int UTIL_setFDStat(const int fd, const char *filename, const stat_t *statbuf)
+{
+    int res = 0;
     stat_t curStatBuf;
-    if (!UTIL_stat(filename, &curStatBuf) || !UTIL_isRegularFileStat(&curStatBuf))
+    UTIL_TRACE_CALL("UTIL_setFileStat(%d, %s)", fd, filename);
+
+    if (!UTIL_fstat(fd, filename, &curStatBuf) || !UTIL_isRegularFileStat(&curStatBuf)) {
+        UTIL_TRACE_RET(-1);
         return -1;
+    }
 
-    /* set access and modification times */
-    res += UTIL_utime(filename, statbuf);
+    /* Mimic gzip's behavior:
+     *
+     * "Change the group first, then the permissions, then the owner.
+     * That way, the permissions will be correct on systems that allow
+     * users to give away files, without introducing a security hole.
+     * Security depends on permissions not containing the setuid or
+     * setgid bits." */
 
 #if !defined(_WIN32)
-    res += chown(filename, statbuf->st_uid, statbuf->st_gid);  /* Copy ownership */
+#ifdef ZSTD_HAVE_FCHOWN
+    if (fd >= 0) {
+        res += fchown(fd, -1, statbuf->st_gid);  /* Apply group ownership */
+    } else
+#endif
+    {
+        res += chown(filename, -1, statbuf->st_gid);  /* Apply group ownership */
+    }
 #endif
 
-    res += UTIL_chmod(filename, &curStatBuf, statbuf->st_mode & 07777);  /* Copy file permissions */
+    res += UTIL_fchmod(fd, filename, &curStatBuf, statbuf->st_mode & 0777);  /* Copy file permissions */
+
+#if !defined(_WIN32)
+#ifdef ZSTD_HAVE_FCHOWN
+    if (fd >= 0) {
+        res += fchown(fd, statbuf->st_uid, -1);  /* Apply user ownership */
+    } else
+#endif
+    {
+        res += chown(filename, statbuf->st_uid, -1);  /* Apply user ownership */
+    }
+#endif
 
     errno = 0;
+    UTIL_TRACE_RET(-res);
     return -res; /* number of errors is returned */
 }
 
 int UTIL_isDirectory(const char* infilename)
 {
     stat_t statbuf;
-    return UTIL_stat(infilename, &statbuf) && UTIL_isDirectoryStat(&statbuf);
+    int ret;
+    UTIL_TRACE_CALL("UTIL_isDirectory(%s)", infilename);
+    ret = UTIL_stat(infilename, &statbuf) && UTIL_isDirectoryStat(&statbuf);
+    UTIL_TRACE_RET(ret);
+    return ret;
 }
 
 int UTIL_isDirectoryStat(const stat_t* statbuf)
 {
+    int ret;
+    UTIL_TRACE_CALL("UTIL_isDirectoryStat()");
 #if defined(_MSC_VER)
-    return (statbuf->st_mode & _S_IFDIR) != 0;
+    ret = (statbuf->st_mode & _S_IFDIR) != 0;
 #else
-    return S_ISDIR(statbuf->st_mode) != 0;
+    ret = S_ISDIR(statbuf->st_mode) != 0;
 #endif
+    UTIL_TRACE_RET(ret);
+    return ret;
 }
 
 int UTIL_compareStr(const void *p1, const void *p2) {
@@ -224,33 +372,68 @@ int UTIL_compareStr(const void *p1, const void *p2) {
 
 int UTIL_isSameFile(const char* fName1, const char* fName2)
 {
+    int ret;
     assert(fName1 != NULL); assert(fName2 != NULL);
+    UTIL_TRACE_CALL("UTIL_isSameFile(%s, %s)", fName1, fName2);
 #if defined(_MSC_VER) || defined(_WIN32)
     /* note : Visual does not support file identification by inode.
      *        inode does not work on Windows, even with a posix layer, like msys2.
      *        The following work-around is limited to detecting exact name repetition only,
      *        aka `filename` is considered different from `subdir/../filename` */
-    return !strcmp(fName1, fName2);
+    ret = !strcmp(fName1, fName2);
 #else
     {   stat_t file1Stat;
         stat_t file2Stat;
-        return UTIL_stat(fName1, &file1Stat)
+        ret =  UTIL_stat(fName1, &file1Stat)
             && UTIL_stat(fName2, &file2Stat)
-            && (file1Stat.st_dev == file2Stat.st_dev)
-            && (file1Stat.st_ino == file2Stat.st_ino);
+            && UTIL_isSameFileStat(fName1, fName2, &file1Stat, &file2Stat);
     }
 #endif
+    UTIL_TRACE_RET(ret);
+    return ret;
+}
+
+int UTIL_isSameFileStat(
+        const char* fName1, const char* fName2,
+        const stat_t* file1Stat, const stat_t* file2Stat)
+{
+    int ret;
+    assert(fName1 != NULL); assert(fName2 != NULL);
+    UTIL_TRACE_CALL("UTIL_isSameFileStat(%s, %s)", fName1, fName2);
+#if defined(_MSC_VER) || defined(_WIN32)
+    /* note : Visual does not support file identification by inode.
+     *        inode does not work on Windows, even with a posix layer, like msys2.
+     *        The following work-around is limited to detecting exact name repetition only,
+     *        aka `filename` is considered different from `subdir/../filename` */
+    (void)file1Stat;
+    (void)file2Stat;
+    ret = !strcmp(fName1, fName2);
+#else
+    {
+        ret =  (file1Stat->st_dev == file2Stat->st_dev)
+            && (file1Stat->st_ino == file2Stat->st_ino);
+    }
+#endif
+    UTIL_TRACE_RET(ret);
+    return ret;
 }
 
 /* UTIL_isFIFO : distinguish named pipes */
 int UTIL_isFIFO(const char* infilename)
 {
+    UTIL_TRACE_CALL("UTIL_isFIFO(%s)", infilename);
 /* macro guards, as defined in : https://linux.die.net/man/2/lstat */
 #if PLATFORM_POSIX_VERSION >= 200112L
-    stat_t statbuf;
-    if (UTIL_stat(infilename, &statbuf) && UTIL_isFIFOStat(&statbuf)) return 1;
+    {
+        stat_t statbuf;
+        if (UTIL_stat(infilename, &statbuf) && UTIL_isFIFOStat(&statbuf)) {
+            UTIL_TRACE_RET(1);
+            return 1;
+        }
+    }
 #endif
     (void)infilename;
+    UTIL_TRACE_RET(0);
     return 0;
 }
 
@@ -278,21 +461,69 @@ int UTIL_isBlockDevStat(const stat_t* statbuf)
 
 int UTIL_isLink(const char* infilename)
 {
+    UTIL_TRACE_CALL("UTIL_isLink(%s)", infilename);
 /* macro guards, as defined in : https://linux.die.net/man/2/lstat */
 #if PLATFORM_POSIX_VERSION >= 200112L
-    stat_t statbuf;
-    int const r = lstat(infilename, &statbuf);
-    if (!r && S_ISLNK(statbuf.st_mode)) return 1;
+    {
+        stat_t statbuf;
+        int const r = lstat(infilename, &statbuf);
+        if (!r && S_ISLNK(statbuf.st_mode)) {
+            UTIL_TRACE_RET(1);
+            return 1;
+        }
+    }
 #endif
     (void)infilename;
+    UTIL_TRACE_RET(0);
     return 0;
 }
 
+static int g_fakeStdinIsConsole = 0;
+static int g_fakeStderrIsConsole = 0;
+static int g_fakeStdoutIsConsole = 0;
+
+int UTIL_isConsole(FILE* file)
+{
+    int ret;
+    UTIL_TRACE_CALL("UTIL_isConsole(%d)", fileno(file));
+    if (file == stdin && g_fakeStdinIsConsole)
+        ret = 1;
+    else if (file == stderr && g_fakeStderrIsConsole)
+        ret = 1;
+    else if (file == stdout && g_fakeStdoutIsConsole)
+        ret = 1;
+    else
+        ret = IS_CONSOLE(file);
+    UTIL_TRACE_RET(ret);
+    return ret;
+}
+
+void UTIL_fakeStdinIsConsole(void)
+{
+    g_fakeStdinIsConsole = 1;
+}
+void UTIL_fakeStdoutIsConsole(void)
+{
+    g_fakeStdoutIsConsole = 1;
+}
+void UTIL_fakeStderrIsConsole(void)
+{
+    g_fakeStderrIsConsole = 1;
+}
+
 U64 UTIL_getFileSize(const char* infilename)
 {
     stat_t statbuf;
-    if (!UTIL_stat(infilename, &statbuf)) return UTIL_FILESIZE_UNKNOWN;
-    return UTIL_getFileSizeStat(&statbuf);
+    UTIL_TRACE_CALL("UTIL_getFileSize(%s)", infilename);
+    if (!UTIL_stat(infilename, &statbuf)) {
+        UTIL_TRACE_RET(-1);
+        return UTIL_FILESIZE_UNKNOWN;
+    }
+    {
+        U64 const size = UTIL_getFileSizeStat(&statbuf);
+        UTIL_TRACE_RET((int)size);
+        return size;
+    }
 }
 
 U64 UTIL_getFileSizeStat(const stat_t* statbuf)
@@ -369,11 +600,16 @@ U64 UTIL_getTotalFileSize(const char* const * fileNamesTable, unsigned nbFiles)
 {
     U64 total = 0;
     unsigned n;
+    UTIL_TRACE_CALL("UTIL_getTotalFileSize(%u)", nbFiles);
     for (n=0; n<nbFiles; n++) {
         U64 const size = UTIL_getFileSize(fileNamesTable[n]);
-        if (size == UTIL_FILESIZE_UNKNOWN) return UTIL_FILESIZE_UNKNOWN;
+        if (size == UTIL_FILESIZE_UNKNOWN) {
+            UTIL_TRACE_RET(-1);
+            return UTIL_FILESIZE_UNKNOWN;
+        }
         total += size;
     }
+    UTIL_TRACE_RET((int)total);
     return total;
 }
 
@@ -418,7 +654,7 @@ readLinesFromFile(void* dst, size_t dstCapacity,
     while ( !feof(inputFile) ) {
         size_t const lineLength = readLineFromFile(buf+pos, dstCapacity-pos, inputFile);
         if (lineLength == 0) break;
-        assert(pos + lineLength < dstCapacity);
+        assert(pos + lineLength <= dstCapacity); /* '=' for inputFile not terminated with '\n' */
         pos += lineLength;
         ++nbFiles;
     }
@@ -435,7 +671,6 @@ UTIL_createFileNamesTable_fromFileName(const char* inputFileName)
     size_t nbFiles = 0;
     char* buf;
     size_t bufSize;
-    size_t pos = 0;
     stat_t statbuf;
 
     if (!UTIL_stat(inputFileName, &statbuf) || !UTIL_isRegularFileStat(&statbuf))
@@ -462,12 +697,13 @@ UTIL_createFileNamesTable_fromFileName(const char* inputFileName)
     {   const char** filenamesTable = (const char**) malloc(nbFiles * sizeof(*filenamesTable));
         CONTROL(filenamesTable != NULL);
 
-        {   size_t fnb;
-            for (fnb = 0, pos = 0; fnb < nbFiles; fnb++) {
+        {   size_t fnb, pos = 0;
+            for (fnb = 0; fnb < nbFiles; fnb++) {
                 filenamesTable[fnb] = buf+pos;
                 pos += strlen(buf+pos)+1;  /* +1 for the finishing `\0` */
-        }   }
+            }
         assert(pos <= bufSize);
+        }
 
         return UTIL_assembleFileNamesTable(filenamesTable, nbFiles, buf);
     }
@@ -509,6 +745,16 @@ FileNamesTable* UTIL_allocateFileNamesTable(size_t tableSize)
     return fnt;
 }
 
+int UTIL_searchFileNamesTable(FileNamesTable* table, char const* name) {
+    size_t i;
+    for(i=0 ;i < table->tableSize; i++) {
+        if(!strcmp(table->fileNames[i], name)) {
+            return (int)i;
+        }
+    }
+    return -1;
+}
+
 void UTIL_refFilename(FileNamesTable* fnt, const char* filename)
 {
     assert(fnt->tableSize < fnt->tableCapacity);
@@ -518,7 +764,7 @@ void UTIL_refFilename(FileNamesTable* fnt, const char* filename)
 
 static size_t getTotalTableSize(FileNamesTable* table)
 {
-    size_t fnb = 0, totalSize = 0;
+    size_t fnb, totalSize = 0;
     for(fnb = 0 ; fnb < table->tableSize && table->fileNames[fnb] ; ++fnb) {
         totalSize += strlen(table->fileNames[fnb]) + 1; /* +1 to add '\0' at the end of each fileName */
     }
@@ -559,7 +805,7 @@ UTIL_mergeFileNamesTable(FileNamesTable* table1, FileNamesTable* table2)
         for( idx2=0 ; (idx2 < table2->tableSize) && table2->fileNames[idx2] && (pos < newTotalTableSize) ; ++idx2, ++newTableIdx) {
             size_t const curLen = strlen(table2->fileNames[idx2]);
             memcpy(buf+pos, table2->fileNames[idx2], curLen);
-            assert(newTableIdx <= newTable->tableSize);
+            assert(newTableIdx < newTable->tableSize);
             newTable->fileNames[newTableIdx] = buf+pos;
             pos += curLen+1;
     }   }
@@ -683,8 +929,11 @@ static int UTIL_prepareFileList(const char *dirName,
                 ptrdiff_t newListSize = (*bufEnd - *bufStart) + LIST_SIZE_INCREASE;
                 assert(newListSize >= 0);
                 *bufStart = (char*)UTIL_realloc(*bufStart, (size_t)newListSize);
-                *bufEnd = *bufStart + newListSize;
-                if (*bufStart == NULL) { free(path); closedir(dir); return 0; }
+                if (*bufStart != NULL) {
+                    *bufEnd = *bufStart + newListSize;
+                } else {
+                    free(path); closedir(dir); return 0;
+                }
             }
             if (*bufStart + *pos + pathLength < *bufEnd) {
                 memcpy(*bufStart + *pos, path, pathLength + 1);  /* with final \0 */
@@ -870,30 +1119,27 @@ static const char * trimPath(const char *pathname)
 
 static char* mallocAndJoin2Dir(const char *dir1, const char *dir2)
 {
-    const size_t dir1Size = strlen(dir1);
-    const size_t dir2Size = strlen(dir2);
-    char *outDirBuffer, *buffer, trailingChar;
-
     assert(dir1 != NULL && dir2 != NULL);
-    outDirBuffer = (char *) malloc(dir1Size + dir2Size + 2);
-    CONTROL(outDirBuffer != NULL);
+    {   const size_t dir1Size = strlen(dir1);
+        const size_t dir2Size = strlen(dir2);
+        char *outDirBuffer, *buffer;
 
-    memcpy(outDirBuffer, dir1, dir1Size);
-    outDirBuffer[dir1Size] = '\0';
+        outDirBuffer = (char *) malloc(dir1Size + dir2Size + 2);
+        CONTROL(outDirBuffer != NULL);
 
-    if (dir2[0] == '.')
-        return outDirBuffer;
+        memcpy(outDirBuffer, dir1, dir1Size);
+        outDirBuffer[dir1Size] = '\0';
 
-    buffer = outDirBuffer + dir1Size;
-    trailingChar = *(buffer - 1);
-    if (trailingChar != PATH_SEP) {
-        *buffer = PATH_SEP;
-        buffer++;
-    }
-    memcpy(buffer, dir2, dir2Size);
-    buffer[dir2Size] = '\0';
+        buffer = outDirBuffer + dir1Size;
+        if (dir1Size > 0 && *(buffer - 1) != PATH_SEP) {
+            *buffer = PATH_SEP;
+            buffer++;
+        }
+        memcpy(buffer, dir2, dir2Size);
+        buffer[dir2Size] = '\0';
 
-    return outDirBuffer;
+        return outDirBuffer;
+    }
 }
 
 /* this function will return NULL if input srcFileName is not valid name for mirrored output path */
@@ -999,7 +1245,7 @@ makeUniqueMirroredDestDirs(char** srcDirNames, unsigned nbFile, const char* outD
                                             trimPath(currDirName)))
             uniqueDirNr++;
 
-        /* we need maintain original src dir name instead of trimmed
+        /* we need to maintain original src dir name instead of trimmed
          * dir, so we can retrieve the original src dir's mode_t */
         uniqueDirNames[uniqueDirNr - 1] = currDirName;
     }
@@ -1308,7 +1554,6 @@ failed:
 
 #elif defined(__FreeBSD__)
 
-#include <sys/param.h>
 #include <sys/sysctl.h>
 
 /* Use physical core sysctl when available
@@ -1378,6 +1623,9 @@ int UTIL_countCores(int logical)
 
 int UTIL_countCores(int logical)
 {
+    /* suppress unused parameter warning */
+    (void)logical;
+
     /* assume 1 */
     return 1;
 }
@@ -1393,7 +1641,3 @@ int UTIL_countLogicalCores(void)
 {
     return UTIL_countCores(1);
 }
-
-#if defined (__cplusplus)
-}
-#endif
diff --git a/sys/contrib/zstd/programs/util.h b/sys/contrib/zstd/programs/util.h
index add165d57ce9..d768e7660910 100644
--- a/sys/contrib/zstd/programs/util.h
+++ b/sys/contrib/zstd/programs/util.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -11,25 +11,26 @@
 #ifndef UTIL_H_MODULE
 #define UTIL_H_MODULE
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-
 /*-****************************************
 *  Dependencies
 ******************************************/
 #include "platform.h"     /* PLATFORM_POSIX_VERSION, ZSTD_NANOSLEEP_SUPPORT, ZSTD_SETPRIORITY_SUPPORT */
 #include <stddef.h>       /* size_t, ptrdiff_t */
+#include <stdio.h>        /* FILE */
 #include <sys/types.h>    /* stat, utime */
 #include <sys/stat.h>     /* stat, chmod */
 #include "../lib/common/mem.h"          /* U64 */
-
+#if !(defined(_MSC_VER) || defined(__MINGW32__) || defined (__MSVCRT__))
+#include <libgen.h>
+#endif
 
 /*-************************************************************
-* Avoid fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW
+*  Fix fseek()'s 2GiB barrier with MSVC, macOS, *BSD, MinGW
 ***************************************************************/
-#if defined(_MSC_VER) && (_MSC_VER >= 1400)
+#if defined(LIBC_NO_FSEEKO)
+/* Some older libc implementations don't include these functions (e.g. Bionic < 24) */
+#  define UTIL_fseek fseek
+#elif defined(_MSC_VER) && (_MSC_VER >= 1400)
 #  define UTIL_fseek _fseeki64
 #elif !defined(__64BIT__) && (PLATFORM_POSIX_VERSION >= 200112L) /* No point defining Large file for 64 bit */
 #  define UTIL_fseek fseeko
@@ -39,7 +40,6 @@ extern "C" {
 #  define UTIL_fseek fseek
 #endif
 
-
 /*-*************************************************
 *  Sleep & priority functions: Windows - Posix - others
 ***************************************************/
@@ -88,6 +88,10 @@ extern "C" {
 #endif
 
 
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
 /*-****************************************
 *  Console log
 ******************************************/
@@ -118,7 +122,6 @@ int UTIL_requireUserConfirmation(const char* prompt, const char* abortMsg, const
 #define STRDUP(s) _strdup(s)
 #else
 #define PATH_SEP '/'
-#include <libgen.h>
 #define STRDUP(s) strdup(s)
 #endif
 
@@ -126,15 +129,25 @@ int UTIL_requireUserConfirmation(const char* prompt, const char* abortMsg, const
 /**
  * Calls platform's equivalent of stat() on filename and writes info to statbuf.
  * Returns success (1) or failure (0).
+ *
+ * UTIL_fstat() is like UTIL_stat() but takes an optional fd that refers to the
+ * file in question. It turns out that this can be meaningfully faster. If fd is
+ * -1, behaves just like UTIL_stat() (i.e., falls back to using the filename).
  */
 int UTIL_stat(const char* filename, stat_t* statbuf);
+int UTIL_fstat(const int fd, const char* filename, stat_t* statbuf);
 
 /**
  * Instead of getting a file's stats, this updates them with the info in the
  * provided stat_t. Currently sets owner, group, atime, and mtime. Will only
  * update this info for regular files.
+ *
+ * UTIL_setFDStat() also takes an fd, and will preferentially use that to
+ * indicate which file to modify, If fd is -1, it will fall back to using the
+ * filename.
  */
 int UTIL_setFileStat(const char* filename, const stat_t* statbuf);
+int UTIL_setFDStat(const int fd, const char* filename, const stat_t* statbuf);
 
 /**
  * Set atime to now and mtime to the st_mtim in statbuf.
@@ -159,8 +172,11 @@ U64 UTIL_getFileSizeStat(const stat_t* statbuf);
  * Like chmod(), but only modifies regular files. Provided statbuf may be NULL,
  * in which case this function will stat() the file internally, in order to
  * check whether it should be modified.
+ *
+ * If fd is -1, fd is ignored and the filename is used.
  */
 int UTIL_chmod(char const* filename, const stat_t* statbuf, mode_t permissions);
+int UTIL_fchmod(const int fd, char const* filename, const stat_t* statbuf, mode_t permissions);
 
 /*
  * In the absence of a pre-existing stat result on the file in question, these
@@ -171,10 +187,30 @@ int UTIL_chmod(char const* filename, const stat_t* statbuf, mode_t permissions);
 int UTIL_isRegularFile(const char* infilename);
 int UTIL_isDirectory(const char* infilename);
 int UTIL_isSameFile(const char* file1, const char* file2);
+int UTIL_isSameFileStat(const char* file1, const char* file2, const stat_t* file1Stat, const stat_t* file2Stat);
 int UTIL_isCompressedFile(const char* infilename, const char *extensionList[]);
 int UTIL_isLink(const char* infilename);
 int UTIL_isFIFO(const char* infilename);
 
+/**
+ * Returns with the given file descriptor is a console.
+ * Allows faking whether stdin/stdout/stderr is a console
+ * using UTIL_fake*IsConsole().
+ */
+int UTIL_isConsole(FILE* file);
+
+/**
+ * Pretends that stdin/stdout/stderr is a console for testing.
+ */
+void UTIL_fakeStdinIsConsole(void);
+void UTIL_fakeStdoutIsConsole(void);
+void UTIL_fakeStderrIsConsole(void);
+
+/**
+ * Emit traces for functions that read, or modify file metadata.
+ */
+void UTIL_traceFileStat(void);
+
 #define UTIL_FILESIZE_UNKNOWN  ((U64)(-1))
 U64 UTIL_getFileSize(const char* infilename);
 U64 UTIL_getTotalFileSize(const char* const * fileNamesTable, unsigned nbFiles);
@@ -248,7 +284,6 @@ UTIL_mergeFileNamesTable(FileNamesTable* table1, FileNamesTable* table2);
 /*! UTIL_expandFNT() :
  *  read names from @fnt, and expand those corresponding to directories
  *  update @fnt, now containing only file names,
- * @return : 0 in case of success, 1 if error
  *  note : in case of error, @fnt[0] is NULL
  */
 void UTIL_expandFNT(FileNamesTable** fnt, int followLinks);
@@ -269,6 +304,11 @@ UTIL_createFNT_fromROTable(const char** filenames, size_t nbFilenames);
  */
 FileNamesTable* UTIL_allocateFileNamesTable(size_t tableSize);
 
+/*! UTIL_searchFileNamesTable() :
+ *  Searched through entries in FileNamesTable for a specific name.
+ * @return : index of entry if found or -1 if not found
+ */
+int UTIL_searchFileNamesTable(FileNamesTable* table, char const* name);
 
 /*! UTIL_refFilename() :
  *  Add a reference to read-only name into @fnt table.
@@ -301,7 +341,7 @@ void UTIL_refFilename(FileNamesTable* fnt, const char* filename);
 FileNamesTable*
 UTIL_createExpandedFNT(const char* const* filenames, size_t nbFilenames, int followLinks);
 
-#if defined(_WIN32) || defined(WIN32)
+#if defined(_WIN32)
 DWORD CountSetBits(ULONG_PTR bitMask);
 #endif
 
diff --git a/sys/contrib/zstd/programs/zstd.1 b/sys/contrib/zstd/programs/zstd.1
index c7a19dbacace..5f1519f3323b 100644
--- a/sys/contrib/zstd/programs/zstd.1
+++ b/sys/contrib/zstd/programs/zstd.1
@@ -1,501 +1,392 @@
-.
-.TH "ZSTD" "1" "January 2022" "zstd 1.5.2" "User Commands"
-.
+.TH "ZSTD" "1" "October 2024" "zstd 1.5.6" "User Commands"
 .SH "NAME"
 \fBzstd\fR \- zstd, zstdmt, unzstd, zstdcat \- Compress or decompress \.zst files
-.
 .SH "SYNOPSIS"
-\fBzstd\fR [\fIOPTIONS\fR] [\-|\fIINPUT\-FILE\fR] [\-o \fIOUTPUT\-FILE\fR]
-.
+.TS
+allbox;
+\fBzstd\fR [\fIOPTIONS\fR] [\-	\fIINPUT\-FILE\fR] [\-o \fIOUTPUT\-FILE\fR]
+.TE
 .P
 \fBzstdmt\fR is equivalent to \fBzstd \-T0\fR
-.
 .P
 \fBunzstd\fR is equivalent to \fBzstd \-d\fR
-.
 .P
 \fBzstdcat\fR is equivalent to \fBzstd \-dcf\fR
-.
 .SH "DESCRIPTION"
-\fBzstd\fR is a fast lossless compression algorithm and data compression tool, with command line syntax similar to \fBgzip (1)\fR and \fBxz (1)\fR\. It is based on the \fBLZ77\fR family, with further FSE & huff0 entropy stages\. \fBzstd\fR offers highly configurable compression speed, with fast modes at > 200 MB/s per core, and strong modes nearing lzma compression ratios\. It also features a very fast decoder, with speeds > 500 MB/s per core\.
-.
+\fBzstd\fR is a fast lossless compression algorithm and data compression tool, with command line syntax similar to \fBgzip\fR(1) and \fBxz\fR(1)\. It is based on the \fBLZ77\fR family, with further FSE & huff0 entropy stages\. \fBzstd\fR offers highly configurable compression speed, from fast modes at > 200 MB/s per core, to strong modes with excellent compression ratios\. It also features a very fast decoder, with speeds > 500 MB/s per core, which remains roughly stable at all compression settings\.
 .P
-\fBzstd\fR command line syntax is generally similar to gzip, but features the following differences :
-.
+\fBzstd\fR command line syntax is generally similar to gzip, but features the following few differences:
 .IP "\(bu" 4
-Source files are preserved by default\. It\'s possible to remove them automatically by using the \fB\-\-rm\fR command\.
-.
+Source files are preserved by default\. It's possible to remove them automatically by using the \fB\-\-rm\fR command\.
 .IP "\(bu" 4
 When compressing a single file, \fBzstd\fR displays progress notifications and result summary by default\. Use \fB\-q\fR to turn them off\.
-.
-.IP "\(bu" 4
-\fBzstd\fR does not accept input from console, but it properly accepts \fBstdin\fR when it\'s not the console\.
-.
 .IP "\(bu" 4
 \fBzstd\fR displays a short help page when command line is an error\. Use \fB\-q\fR to turn it off\.
-.
+.IP "\(bu" 4
+\fBzstd\fR does not accept input from console, though it does accept \fBstdin\fR when it's not the console\.
+.IP "\(bu" 4
+\fBzstd\fR does not store the input's filename or attributes, only its contents\.
 .IP "" 0
-.
 .P
-\fBzstd\fR compresses or decompresses each \fIfile\fR according to the selected operation mode\. If no \fIfiles\fR are given or \fIfile\fR is \fB\-\fR, \fBzstd\fR reads from standard input and writes the processed data to standard output\. \fBzstd\fR will refuse to write compressed data to standard output if it is a terminal : it will display an error message and skip the \fIfile\fR\. Similarly, \fBzstd\fR will refuse to read compressed data from standard input if it is a terminal\.
-.
+\fBzstd\fR processes each \fIfile\fR according to the selected operation mode\. If no \fIfiles\fR are given or \fIfile\fR is \fB\-\fR, \fBzstd\fR reads from standard input and writes the processed data to standard output\. \fBzstd\fR will refuse to write compressed data to standard output if it is a terminal: it will display an error message and skip the file\. Similarly, \fBzstd\fR will refuse to read compressed data from standard input if it is a terminal\.
 .P
 Unless \fB\-\-stdout\fR or \fB\-o\fR is specified, \fIfiles\fR are written to a new file whose name is derived from the source \fIfile\fR name:
-.
 .IP "\(bu" 4
 When compressing, the suffix \fB\.zst\fR is appended to the source filename to get the target filename\.
-.
 .IP "\(bu" 4
 When decompressing, the \fB\.zst\fR suffix is removed from the source filename to get the target filename
-.
 .IP "" 0
-.
-.SS "Concatenation with \.zst files"
-It is possible to concatenate \fB\.zst\fR files as is\. \fBzstd\fR will decompress such files as if they were a single \fB\.zst\fR file\.
-.
+.SS "Concatenation with \.zst Files"
+It is possible to concatenate multiple \fB\.zst\fR files\. \fBzstd\fR will decompress such agglomerated file as if it was a single \fB\.zst\fR file\.
 .SH "OPTIONS"
-.
-.SS "Integer suffixes and special values"
+.SS "Integer Suffixes and Special Values"
 In most places where an integer argument is expected, an optional suffix is supported to easily indicate large integers\. There must be no space between the integer and the suffix\.
-.
 .TP
 \fBKiB\fR
-Multiply the integer by 1,024 (2^10)\. \fBKi\fR, \fBK\fR, and \fBKB\fR are accepted as synonyms for \fBKiB\fR\.
-.
+Multiply the integer by 1,024 (2\e^10)\. \fBKi\fR, \fBK\fR, and \fBKB\fR are accepted as synonyms for \fBKiB\fR\.
 .TP
 \fBMiB\fR
-Multiply the integer by 1,048,576 (2^20)\. \fBMi\fR, \fBM\fR, and \fBMB\fR are accepted as synonyms for \fBMiB\fR\.
-.
-.SS "Operation mode"
+Multiply the integer by 1,048,576 (2\e^20)\. \fBMi\fR, \fBM\fR, and \fBMB\fR are accepted as synonyms for \fBMiB\fR\.
+.SS "Operation Mode"
 If multiple operation mode options are given, the last one takes effect\.
-.
 .TP
 \fB\-z\fR, \fB\-\-compress\fR
 Compress\. This is the default operation mode when no operation mode option is specified and no other operation mode is implied from the command name (for example, \fBunzstd\fR implies \fB\-\-decompress\fR)\.
-.
 .TP
 \fB\-d\fR, \fB\-\-decompress\fR, \fB\-\-uncompress\fR
 Decompress\.
-.
 .TP
 \fB\-t\fR, \fB\-\-test\fR
-Test the integrity of compressed \fIfiles\fR\. This option is equivalent to \fB\-\-decompress \-\-stdout\fR except that the decompressed data is discarded instead of being written to standard output\. No files are created or removed\.
-.
+Test the integrity of compressed \fIfiles\fR\. This option is equivalent to \fB\-\-decompress \-\-stdout > /dev/null\fR, decompressed data is discarded and checksummed for errors\. No files are created or removed\.
 .TP
 \fB\-b#\fR
-Benchmark file(s) using compression level #
-.
+Benchmark file(s) using compression level \fI#\fR\. See \fIBENCHMARK\fR below for a description of this operation\.
 .TP
-\fB\-\-train FILEs\fR
-Use FILEs as a training set to create a dictionary\. The training set should contain a lot of small files (> 100)\.
-.
+\fB\-\-train FILES\fR
+Use \fIFILES\fR as a training set to create a dictionary\. The training set should contain a lot of small files (> 100)\. See \fIDICTIONARY BUILDER\fR below for a description of this operation\.
 .TP
 \fB\-l\fR, \fB\-\-list\fR
-Display information related to a zstd compressed file, such as size, ratio, and checksum\. Some of these fields may not be available\. This command can be augmented with the \fB\-v\fR modifier\.
-.
-.SS "Operation modifiers"
-.
+Display information related to a zstd compressed file, such as size, ratio, and checksum\. Some of these fields may not be available\. This command's output can be augmented with the \fB\-v\fR modifier\.
+.SS "Operation Modifiers"
 .IP "\(bu" 4
-\fB\-#\fR: \fB#\fR compression level [1\-19] (default: 3)
-.
+\fB\-#\fR: selects \fB#\fR compression level [1\-19] (default: 3)\. Higher compression levels \fIgenerally\fR produce higher compression ratio at the expense of speed and memory\. A rough rule of thumb is that compression speed is expected to be divided by 2 every 2 levels\. Technically, each level is mapped to a set of advanced parameters (that can also be modified individually, see below)\. Because the compressor's behavior highly depends on the content to compress, there's no guarantee of a smooth progression from one level to another\.
 .IP "\(bu" 4
 \fB\-\-ultra\fR: unlocks high compression levels 20+ (maximum 22), using a lot more memory\. Note that decompression will also require more memory when using these levels\.
-.
 .IP "\(bu" 4
 \fB\-\-fast[=#]\fR: switch to ultra\-fast compression levels\. If \fB=#\fR is not present, it defaults to \fB1\fR\. The higher the value, the faster the compression speed, at the cost of some compression ratio\. This setting overwrites compression level if one was set previously\. Similarly, if a compression level is set after \fB\-\-fast\fR, it overrides it\.
-.
 .IP "\(bu" 4
 \fB\-T#\fR, \fB\-\-threads=#\fR: Compress using \fB#\fR working threads (default: 1)\. If \fB#\fR is 0, attempt to detect and use the number of physical CPU cores\. In all cases, the nb of threads is capped to \fBZSTDMT_NBWORKERS_MAX\fR, which is either 64 in 32\-bit mode, or 256 for 64\-bit environments\. This modifier does nothing if \fBzstd\fR is compiled without multithread support\.
-.
 .IP "\(bu" 4
-\fB\-\-single\-thread\fR: Does not spawn a thread for compression, use a single thread for both I/O and compression\. In this mode, compression is serialized with I/O, which is slightly slower\. (This is different from \fB\-T1\fR, which spawns 1 compression thread in parallel of I/O)\. This mode is the only one available when multithread support is disabled\. Single\-thread mode features lower memory usage\. Final compressed result is slightly different from \fB\-T1\fR\.
-.
+\fB\-\-single\-thread\fR: Use a single thread for both I/O and compression\. As compression is serialized with I/O, this can be slightly slower\. Single\-thread mode features significantly lower memory usage, which can be useful for systems with limited amount of memory, such as 32\-bit systems\.
+.IP
+Note 1: this mode is the only available one when multithread support is disabled\.
+.IP
+Note 2: this mode is different from \fB\-T1\fR, which spawns 1 compression thread in parallel with I/O\. Final compressed result is also slightly different from \fB\-T1\fR\.
 .IP "\(bu" 4
 \fB\-\-auto\-threads={physical,logical} (default: physical)\fR: When using a default amount of threads via \fB\-T0\fR, choose the default based on the number of detected physical or logical cores\.
-.
 .IP "\(bu" 4
-\fB\-\-adapt[=min=#,max=#]\fR : \fBzstd\fR will dynamically adapt compression level to perceived I/O conditions\. Compression level adaptation can be observed live by using command \fB\-v\fR\. Adaptation can be constrained between supplied \fBmin\fR and \fBmax\fR levels\. The feature works when combined with multi\-threading and \fB\-\-long\fR mode\. It does not work with \fB\-\-single\-thread\fR\. It sets window size to 8 MB by default (can be changed manually, see \fBwlog\fR)\. Due to the chaotic nature of dynamic adaptation, compressed result is not reproducible\. \fInote\fR : at the time of this writing, \fB\-\-adapt\fR can remain stuck at low speed when combined with multiple worker threads (>=2)\.
-.
+\fB\-\-adapt[=min=#,max=#]\fR: \fBzstd\fR will dynamically adapt compression level to perceived I/O conditions\. Compression level adaptation can be observed live by using command \fB\-v\fR\. Adaptation can be constrained between supplied \fBmin\fR and \fBmax\fR levels\. The feature works when combined with multi\-threading and \fB\-\-long\fR mode\. It does not work with \fB\-\-single\-thread\fR\. It sets window size to 8 MiB by default (can be changed manually, see \fBwlog\fR)\. Due to the chaotic nature of dynamic adaptation, compressed result is not reproducible\.
+.IP
+\fINote\fR: at the time of this writing, \fB\-\-adapt\fR can remain stuck at low speed when combined with multiple worker threads (>=2)\.
 .IP "\(bu" 4
-\fB\-\-long[=#]\fR: enables long distance matching with \fB#\fR \fBwindowLog\fR, if not \fB#\fR is not present it defaults to \fB27\fR\. This increases the window size (\fBwindowLog\fR) and memory usage for both the compressor and decompressor\. This setting is designed to improve the compression ratio for files with long matches at a large distance\.
-.
+\fB\-\-long[=#]\fR: enables long distance matching with \fB#\fR \fBwindowLog\fR, if \fB#\fR is not present it defaults to \fB27\fR\. This increases the window size (\fBwindowLog\fR) and memory usage for both the compressor and decompressor\. This setting is designed to improve the compression ratio for files with long matches at a large distance\.
 .IP
 Note: If \fBwindowLog\fR is set to larger than 27, \fB\-\-long=windowLog\fR or \fB\-\-memory=windowSize\fR needs to be passed to the decompressor\.
-.
 .IP "\(bu" 4
 \fB\-D DICT\fR: use \fBDICT\fR as Dictionary to compress or decompress FILE(s)
-.
 .IP "\(bu" 4
-\fB\-\-patch\-from FILE\fR: Specify the file to be used as a reference point for zstd\'s diff engine\. This is effectively dictionary compression with some convenient parameter selection, namely that windowSize > srcSize\.
-.
+\fB\-\-patch\-from FILE\fR: Specify the file to be used as a reference point for zstd's diff engine\. This is effectively dictionary compression with some convenient parameter selection, namely that \fIwindowSize\fR > \fIsrcSize\fR\.
+.IP
+Note: cannot use both this and \fB\-D\fR together\.
+.IP
+Note: \fB\-\-long\fR mode will be automatically activated if \fIchainLog\fR < \fIfileLog\fR (\fIfileLog\fR being the \fIwindowLog\fR required to cover the whole file)\. You can also manually force it\.
+.IP
+Note: up to level 15, you can use \fB\-\-patch\-from\fR in \fB\-\-single\-thread\fR mode to improve compression ratio marginally at the cost of speed\. Using '\-\-single\-thread' above level 15 will lead to lower compression ratios\.
 .IP
-Note: cannot use both this and \-D together Note: \fB\-\-long\fR mode will be automatically activated if chainLog < fileLog (fileLog being the windowLog required to cover the whole file)\. You can also manually force it\. Node: for all levels, you can use \-\-patch\-from in \-\-single\-thread mode to improve compression ratio at the cost of speed Note: for level 19, you can get increased compression ratio at the cost of speed by specifying \fB\-\-zstd=targetLength=\fR to be something large (i\.e 4096), and by setting a large \fB\-\-zstd=chainLog=\fR
-.
+Note: for level 19, you can get increased compression ratio at the cost of speed by specifying \fB\-\-zstd=targetLength=\fR to be something large (i\.e\. 4096), and by setting a large \fB\-\-zstd=chainLog=\fR\.
 .IP "\(bu" 4
-\fB\-\-rsyncable\fR : \fBzstd\fR will periodically synchronize the compression state to make the compressed file more rsync\-friendly\. There is a negligible impact to compression ratio, and the faster compression levels will see a small compression speed hit\. This feature does not work with \fB\-\-single\-thread\fR\. You probably don\'t want to use it with long range mode, since it will decrease the effectiveness of the synchronization points, but your mileage may vary\.
-.
+\fB\-\-rsyncable\fR: \fBzstd\fR will periodically synchronize the compression state to make the compressed file more rsync\-friendly\. There is a negligible impact to compression ratio, and a potential impact to compression speed, perceptible at higher speeds, for example when combining \fB\-\-rsyncable\fR with many parallel worker threads\. This feature does not work with \fB\-\-single\-thread\fR\. You probably don't want to use it with long range mode, since it will decrease the effectiveness of the synchronization points, but your mileage may vary\.
 .IP "\(bu" 4
 \fB\-C\fR, \fB\-\-[no\-]check\fR: add integrity check computed from uncompressed data (default: enabled)
-.
 .IP "\(bu" 4
-\fB\-\-[no\-]content\-size\fR: enable / disable whether or not the original size of the file is placed in the header of the compressed file\. The default option is \-\-content\-size (meaning that the original size will be placed in the header)\.
-.
+\fB\-\-[no\-]content\-size\fR: enable / disable whether or not the original size of the file is placed in the header of the compressed file\. The default option is \fB\-\-content\-size\fR (meaning that the original size will be placed in the header)\.
 .IP "\(bu" 4
-\fB\-\-no\-dictID\fR: do not store dictionary ID within frame header (dictionary compression)\. The decoder will have to rely on implicit knowledge about which dictionary to use, it won\'t be able to check if it\'s correct\.
-.
+\fB\-\-no\-dictID\fR: do not store dictionary ID within frame header (dictionary compression)\. The decoder will have to rely on implicit knowledge about which dictionary to use, it won't be able to check if it's correct\.
 .IP "\(bu" 4
-\fB\-M#\fR, \fB\-\-memory=#\fR: Set a memory usage limit\. By default, Zstandard uses 128 MB for decompression as the maximum amount of memory the decompressor is allowed to use, but you can override this manually if need be in either direction (ie\. you can increase or decrease it)\.
-.
+\fB\-M#\fR, \fB\-\-memory=#\fR: Set a memory usage limit\. By default, \fBzstd\fR uses 128 MiB for decompression as the maximum amount of memory the decompressor is allowed to use, but you can override this manually if need be in either direction (i\.e\. you can increase or decrease it)\.
 .IP
-This is also used during compression when using with \-\-patch\-from=\. In this case, this parameter overrides that maximum size allowed for a dictionary\. (128 MB)\.
-.
+This is also used during compression when using with \fB\-\-patch\-from=\fR\. In this case, this parameter overrides that maximum size allowed for a dictionary\. (128 MiB)\.
 .IP
-Additionally, this can be used to limit memory for dictionary training\. This parameter overrides the default limit of 2 GB\. zstd will load training samples up to the memory limit and ignore the rest\.
-.
+Additionally, this can be used to limit memory for dictionary training\. This parameter overrides the default limit of 2 GiB\. zstd will load training samples up to the memory limit and ignore the rest\.
 .IP "\(bu" 4
-\fB\-\-stream\-size=#\fR : Sets the pledged source size of input coming from a stream\. This value must be exact, as it will be included in the produced frame header\. Incorrect stream sizes will cause an error\. This information will be used to better optimize compression parameters, resulting in better and potentially faster compression, especially for smaller source sizes\.
-.
+\fB\-\-stream\-size=#\fR: Sets the pledged source size of input coming from a stream\. This value must be exact, as it will be included in the produced frame header\. Incorrect stream sizes will cause an error\. This information will be used to better optimize compression parameters, resulting in better and potentially faster compression, especially for smaller source sizes\.
 .IP "\(bu" 4
 \fB\-\-size\-hint=#\fR: When handling input from a stream, \fBzstd\fR must guess how large the source size will be when optimizing compression parameters\. If the stream size is relatively small, this guess may be a poor one, resulting in a higher compression ratio than expected\. This feature allows for controlling the guess when needed\. Exact guesses result in better compression ratios\. Overestimates result in slightly degraded compression ratios, while underestimates may result in significant degradation\.
-.
 .IP "\(bu" 4
-\fB\-o FILE\fR: save result into \fBFILE\fR
-.
+\fB\-\-target\-compressed\-block\-size=#\fR: Attempt to produce compressed blocks of approximately this size\. This will split larger blocks in order to approach this target\. This feature is notably useful for improved latency, when the receiver can leverage receiving early incomplete data\. This parameter defines a loose target: compressed blocks will target this size "on average", but individual blocks can still be larger or smaller\. Enabling this feature can decrease compression speed by up to ~10% at level 1\. Higher levels will see smaller relative speed regression, becoming invisible at higher settings\.
 .IP "\(bu" 4
-\fB\-f\fR, \fB\-\-force\fR: disable input and output checks\. Allows overwriting existing files, input from console, output to stdout, operating on links, block devices, etc\.
-.
+\fB\-f\fR, \fB\-\-force\fR: disable input and output checks\. Allows overwriting existing files, input from console, output to stdout, operating on links, block devices, etc\. During decompression and when the output destination is stdout, pass\-through unrecognized formats as\-is\.
 .IP "\(bu" 4
-\fB\-c\fR, \fB\-\-stdout\fR: write to standard output (even if it is the console)
-.
+\fB\-c\fR, \fB\-\-stdout\fR: write to standard output (even if it is the console); keep original files (disable \fB\-\-rm\fR)\.
+.IP "\(bu" 4
+\fB\-o FILE\fR: save result into \fBFILE\fR\. Note that this operation is in conflict with \fB\-c\fR\. If both operations are present on the command line, the last expressed one wins\.
 .IP "\(bu" 4
 \fB\-\-[no\-]sparse\fR: enable / disable sparse FS support, to make files with many zeroes smaller on disk\. Creating sparse files may save disk space and speed up decompression by reducing the amount of disk I/O\. default: enabled when output is into a file, and disabled when output is stdout\. This setting overrides default and can force sparse mode over stdout\.
-.
 .IP "\(bu" 4
-\fB\-\-rm\fR: remove source file(s) after successful compression or decompression\. If used in combination with \-o, will trigger a confirmation prompt (which can be silenced with \-f), as this is a destructive operation\.
-.
+\fB\-\-[no\-]pass\-through\fR enable / disable passing through uncompressed files as\-is\. During decompression when pass\-through is enabled, unrecognized formats will be copied as\-is from the input to the output\. By default, pass\-through will occur when the output destination is stdout and the force (\fB\-f\fR) option is set\.
+.IP "\(bu" 4
+\fB\-\-rm\fR: remove source file(s) after successful compression or decompression\. This command is silently ignored if output is \fBstdout\fR\. If used in combination with \fB\-o\fR, triggers a confirmation prompt (which can be silenced with \fB\-f\fR), as this is a destructive operation\.
 .IP "\(bu" 4
 \fB\-k\fR, \fB\-\-keep\fR: keep source file(s) after successful compression or decompression\. This is the default behavior\.
-.
 .IP "\(bu" 4
 \fB\-r\fR: operate recursively on directories\. It selects all files in the named directory and all its subdirectories\. This can be useful both to reduce command line typing, and to circumvent shell expansion limitations, when there are a lot of files and naming breaks the maximum size of a command line\.
-.
 .IP "\(bu" 4
 \fB\-\-filelist FILE\fR read a list of files to process as content from \fBFILE\fR\. Format is compatible with \fBls\fR output, with one file per line\.
-.
 .IP "\(bu" 4
 \fB\-\-output\-dir\-flat DIR\fR: resulting files are stored into target \fBDIR\fR directory, instead of same directory as origin file\. Be aware that this command can introduce name collision issues, if multiple files, from different directories, end up having the same name\. Collision resolution ensures first file with a given name will be present in \fBDIR\fR, while in combination with \fB\-f\fR, the last file will be present instead\.
-.
 .IP "\(bu" 4
 \fB\-\-output\-dir\-mirror DIR\fR: similar to \fB\-\-output\-dir\-flat\fR, the output files are stored underneath target \fBDIR\fR directory, but this option will replicate input directory hierarchy into output \fBDIR\fR\.
-.
 .IP
 If input directory contains "\.\.", the files in this directory will be ignored\. If input directory is an absolute directory (i\.e\. "/var/tmp/abc"), it will be stored into the "output\-dir/var/tmp/abc"\. If there are multiple input files or directories, name collision resolution will follow the same rules as \fB\-\-output\-dir\-flat\fR\.
-.
 .IP "\(bu" 4
 \fB\-\-format=FORMAT\fR: compress and decompress in other formats\. If compiled with support, zstd can compress to or decompress from other compression algorithm formats\. Possibly available options are \fBzstd\fR, \fBgzip\fR, \fBxz\fR, \fBlzma\fR, and \fBlz4\fR\. If no such format is provided, \fBzstd\fR is the default\.
-.
 .IP "\(bu" 4
 \fB\-h\fR/\fB\-H\fR, \fB\-\-help\fR: display help/long help and exit
-.
 .IP "\(bu" 4
-\fB\-V\fR, \fB\-\-version\fR: display version number and exit\. Advanced : \fB\-vV\fR also displays supported formats\. \fB\-vvV\fR also displays POSIX support\. \fB\-q\fR will only display the version number, suitable for machine reading\.
-.
+\fB\-V\fR, \fB\-\-version\fR: display version number and immediately exit\. note that, since it exits, flags specified after \fB\-V\fR are effectively ignored\. Advanced: \fB\-vV\fR also displays supported formats\. \fB\-vvV\fR also displays POSIX support\. \fB\-qV\fR will only display the version number, suitable for machine reading\.
 .IP "\(bu" 4
 \fB\-v\fR, \fB\-\-verbose\fR: verbose mode, display more information
-.
 .IP "\(bu" 4
 \fB\-q\fR, \fB\-\-quiet\fR: suppress warnings, interactivity, and notifications\. specify twice to suppress errors too\.
-.
 .IP "\(bu" 4
 \fB\-\-no\-progress\fR: do not display the progress bar, but keep all other messages\.
-.
 .IP "\(bu" 4
-\fB\-\-show\-default\-cparams\fR: Shows the default compression parameters that will be used for a particular src file\. If the provided src file is not a regular file (eg\. named pipe), the cli will just output the default parameters\. That is, the parameters that are used when the src size is unknown\.
-.
+\fB\-\-show\-default\-cparams\fR: shows the default compression parameters that will be used for a particular input file, based on the provided compression level and the input size\. If the provided file is not a regular file (e\.g\. a pipe), this flag will output the parameters used for inputs of unknown size\.
+.IP "\(bu" 4
+\fB\-\-exclude\-compressed\fR: only compress files that are not already compressed\.
 .IP "\(bu" 4
 \fB\-\-\fR: All arguments after \fB\-\-\fR are treated as files
-.
 .IP "" 0
-.
-.SS "Restricted usage of Environment Variables"
-Using environment variables to set parameters has security implications\. Therefore, this avenue is intentionally restricted\. Only \fBZSTD_CLEVEL\fR and \fBZSTD_NBTHREADS\fR are currently supported\. They set the compression level and number of threads to use during compression, respectively\.
-.
+.SS "gzip Operation Modifiers"
+When invoked via a \fBgzip\fR symlink, \fBzstd\fR will support further options that intend to mimic the \fBgzip\fR behavior:
+.TP
+\fB\-n\fR, \fB\-\-no\-name\fR
+do not store the original filename and timestamps when compressing a file\. This is the default behavior and hence a no\-op\.
+.TP
+\fB\-\-best\fR
+alias to the option \fB\-9\fR\.
+.SS "Environment Variables"
+Employing environment variables to set parameters has security implications\. Therefore, this avenue is intentionally limited\. Only \fBZSTD_CLEVEL\fR and \fBZSTD_NBTHREADS\fR are currently supported\. They set the default compression level and number of threads to use during compression, respectively\.
 .P
 \fBZSTD_CLEVEL\fR can be used to set the level between 1 and 19 (the "normal" range)\. If the value of \fBZSTD_CLEVEL\fR is not a valid integer, it will be ignored with a warning message\. \fBZSTD_CLEVEL\fR just replaces the default compression level (\fB3\fR)\.
-.
 .P
-\fBZSTD_NBTHREADS\fR can be used to set the number of threads \fBzstd\fR will attempt to use during compression\. If the value of \fBZSTD_NBTHREADS\fR is not a valid unsigned integer, it will be ignored with a warning message\. \fBZSTD_NBTHREADS\fR has a default value of (\fB1\fR), and is capped at ZSTDMT_NBWORKERS_MAX==200\. \fBzstd\fR must be compiled with multithread support for this to have any effect\.
-.
+\fBZSTD_NBTHREADS\fR can be used to set the number of threads \fBzstd\fR will attempt to use during compression\. If the value of \fBZSTD_NBTHREADS\fR is not a valid unsigned integer, it will be ignored with a warning message\. \fBZSTD_NBTHREADS\fR has a default value of (\fB1\fR), and is capped at ZSTDMT_NBWORKERS_MAX==200\. \fBzstd\fR must be compiled with multithread support for this variable to have any effect\.
 .P
 They can both be overridden by corresponding command line arguments: \fB\-#\fR for compression level and \fB\-T#\fR for number of compression threads\.
-.
-.SH "DICTIONARY BUILDER"
-\fBzstd\fR offers \fIdictionary\fR compression, which greatly improves efficiency on small files and messages\. It\'s possible to train \fBzstd\fR with a set of samples, the result of which is saved into a file called a \fBdictionary\fR\. Then during compression and decompression, reference the same dictionary, using command \fB\-D dictionaryFileName\fR\. Compression of small files similar to the sample set will be greatly improved\.
-.
-.TP
-\fB\-\-train FILEs\fR
-Use FILEs as training set to create a dictionary\. The training set should contain a lot of small files (> 100), and weight typically 100x the target dictionary size (for example, 10 MB for a 100 KB dictionary)\. \fB\-\-train\fR can be combined with \fB\-r\fR to indicate a directory rather than listing all the files, which can be useful to circumvent shell expansion limits\.
-.
-.IP
-\fB\-\-train\fR supports multithreading if \fBzstd\fR is compiled with threading support (default)\. Additional parameters can be specified with \fB\-\-train\-fastcover\fR\. The legacy dictionary builder can be accessed with \fB\-\-train\-legacy\fR\. The slower cover dictionary builder can be accessed with \fB\-\-train\-cover\fR\. Default is equivalent to \fB\-\-train\-fastcover=d=8,steps=4\fR\.
-.
-.TP
-\fB\-o file\fR
-Dictionary saved into \fBfile\fR (default name: dictionary)\.
-.
-.TP
-\fB\-\-maxdict=#\fR
-Limit dictionary to specified size (default: 112640)\.
-.
-.TP
-\fB\-#\fR
-Use \fB#\fR compression level during training (optional)\. Will generate statistics more tuned for selected compression level, resulting in a \fIsmall\fR compression ratio improvement for this level\.
-.
-.TP
-\fB\-B#\fR
-Split input files into blocks of size # (default: no split)
-.
-.TP
-\fB\-M#\fR, \fB\-\-memory=#\fR
-Limit the amount of sample data loaded for training (default: 2 GB)\. See above for details\.
-.
-.TP
-\fB\-\-dictID=#\fR
-A dictionary ID is a locally unique ID that a decoder can use to verify it is using the right dictionary\. By default, zstd will create a 4\-bytes random number ID\. It\'s possible to give a precise number instead\. Short numbers have an advantage : an ID < 256 will only need 1 byte in the compressed frame header, and an ID < 65536 will only need 2 bytes\. This compares favorably to 4 bytes default\. However, it\'s up to the dictionary manager to not assign twice the same ID to 2 different dictionaries\.
-.
-.TP
-\fB\-\-train\-cover[=k#,d=#,steps=#,split=#,shrink[=#]]\fR
-Select parameters for the default dictionary builder algorithm named cover\. If \fId\fR is not specified, then it tries \fId\fR = 6 and \fId\fR = 8\. If \fIk\fR is not specified, then it tries \fIsteps\fR values in the range [50, 2000]\. If \fIsteps\fR is not specified, then the default value of 40 is used\. If \fIsplit\fR is not specified or split <= 0, then the default value of 100 is used\. Requires that \fId\fR <= \fIk\fR\. If \fIshrink\fR flag is not used, then the default value for \fIshrinkDict\fR of 0 is used\. If \fIshrink\fR is not specified, then the default value for \fIshrinkDictMaxRegression\fR of 1 is used\.
-.
-.IP
-Selects segments of size \fIk\fR with highest score to put in the dictionary\. The score of a segment is computed by the sum of the frequencies of all the subsegments of size \fId\fR\. Generally \fId\fR should be in the range [6, 8], occasionally up to 16, but the algorithm will run faster with d <= \fI8\fR\. Good values for \fIk\fR vary widely based on the input data, but a safe range is [2 * \fId\fR, 2000]\. If \fIsplit\fR is 100, all input samples are used for both training and testing to find optimal \fId\fR and \fIk\fR to build dictionary\. Supports multithreading if \fBzstd\fR is compiled with threading support\. Having \fIshrink\fR enabled takes a truncated dictionary of minimum size and doubles in size until compression ratio of the truncated dictionary is at most \fIshrinkDictMaxRegression%\fR worse than the compression ratio of the largest dictionary\.
-.
-.IP
-Examples:
-.
-.IP
-\fBzstd \-\-train\-cover FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=k=50,d=8 FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=d=8,steps=500 FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=k=50 FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=k=50,split=60 FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=shrink FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-cover=shrink=2 FILEs\fR
-.
-.TP
-\fB\-\-train\-fastcover[=k#,d=#,f=#,steps=#,split=#,accel=#]\fR
-Same as cover but with extra parameters \fIf\fR and \fIaccel\fR and different default value of split If \fIsplit\fR is not specified, then it tries \fIsplit\fR = 75\. If \fIf\fR is not specified, then it tries \fIf\fR = 20\. Requires that 0 < \fIf\fR < 32\. If \fIaccel\fR is not specified, then it tries \fIaccel\fR = 1\. Requires that 0 < \fIaccel\fR <= 10\. Requires that \fId\fR = 6 or \fId\fR = 8\.
-.
-.IP
-\fIf\fR is log of size of array that keeps track of frequency of subsegments of size \fId\fR\. The subsegment is hashed to an index in the range [0,2^\fIf\fR \- 1]\. It is possible that 2 different subsegments are hashed to the same index, and they are considered as the same subsegment when computing frequency\. Using a higher \fIf\fR reduces collision but takes longer\.
-.
-.IP
-Examples:
-.
-.IP
-\fBzstd \-\-train\-fastcover FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-fastcover=d=8,f=15,accel=2 FILEs\fR
-.
-.TP
-\fB\-\-train\-legacy[=selectivity=#]\fR
-Use legacy dictionary builder algorithm with the given dictionary \fIselectivity\fR (default: 9)\. The smaller the \fIselectivity\fR value, the denser the dictionary, improving its efficiency but reducing its possible maximum size\. \fB\-\-train\-legacy=s=#\fR is also accepted\.
-.
-.IP
-Examples:
-.
-.IP
-\fBzstd \-\-train\-legacy FILEs\fR
-.
-.IP
-\fBzstd \-\-train\-legacy=selectivity=8 FILEs\fR
-.
-.SH "BENCHMARK"
-.
-.TP
-\fB\-b#\fR
-benchmark file(s) using compression level #
-.
-.TP
-\fB\-e#\fR
-benchmark file(s) using multiple compression levels, from \fB\-b#\fR to \fB\-e#\fR (inclusive)
-.
-.TP
-\fB\-i#\fR
-minimum evaluation time, in seconds (default: 3s), benchmark mode only
-.
-.TP
-\fB\-B#\fR, \fB\-\-block\-size=#\fR
-cut file(s) into independent blocks of size # (default: no block)
-.
-.TP
-\fB\-\-priority=rt\fR
-set process priority to real\-time
-.
-.P
-\fBOutput Format:\fR CompressionLevel#Filename : IntputSize \-> OutputSize (CompressionRatio), CompressionSpeed, DecompressionSpeed
-.
-.P
-\fBMethodology:\fR For both compression and decompression speed, the entire input is compressed/decompressed in\-memory to measure speed\. A run lasts at least 1 sec, so when files are small, they are compressed/decompressed several times per run, in order to improve measurement accuracy\.
-.
 .SH "ADVANCED COMPRESSION OPTIONS"
-.
-.SS "\-B#:"
-Select the size of each compression job\. This parameter is only available when multi\-threading is enabled\. Each compression job is run in parallel, so this value indirectly impacts the nb of active threads\. Default job size varies depending on compression level (generally \fB4 * windowSize\fR)\. \fB\-B#\fR makes it possible to manually select a custom size\. Note that job size must respect a minimum value which is enforced transparently\. This minimum is either 512 KB, or \fBoverlapSize\fR, whichever is largest\. Different job sizes will lead to (slightly) different compressed frames\.
-.
+\fBzstd\fR provides 22 predefined regular compression levels plus the fast levels\. A compression level is translated internally into multiple advanced parameters that control the behavior of the compressor (one can observe the result of this translation with \fB\-\-show\-default\-cparams\fR)\. These advanced parameters can be overridden using advanced compression options\.
 .SS "\-\-zstd[=options]:"
-\fBzstd\fR provides 22 predefined compression levels\. The selected or default predefined compression level can be changed with advanced compression options\. The \fIoptions\fR are provided as a comma\-separated list\. You may specify only the options you want to change and the rest will be taken from the selected or default compression level\. The list of available \fIoptions\fR:
-.
+The \fIoptions\fR are provided as a comma\-separated list\. You may specify only the options you want to change and the rest will be taken from the selected or default compression level\. The list of available \fIoptions\fR:
 .TP
 \fBstrategy\fR=\fIstrat\fR, \fBstrat\fR=\fIstrat\fR
 Specify a strategy used by a match finder\.
-.
 .IP
-There are 9 strategies numbered from 1 to 9, from faster to stronger: 1=ZSTD_fast, 2=ZSTD_dfast, 3=ZSTD_greedy, 4=ZSTD_lazy, 5=ZSTD_lazy2, 6=ZSTD_btlazy2, 7=ZSTD_btopt, 8=ZSTD_btultra, 9=ZSTD_btultra2\.
-.
+There are 9 strategies numbered from 1 to 9, from fastest to strongest: 1=\fBZSTD_fast\fR, 2=\fBZSTD_dfast\fR, 3=\fBZSTD_greedy\fR, 4=\fBZSTD_lazy\fR, 5=\fBZSTD_lazy2\fR, 6=\fBZSTD_btlazy2\fR, 7=\fBZSTD_btopt\fR, 8=\fBZSTD_btultra\fR, 9=\fBZSTD_btultra2\fR\.
 .TP
 \fBwindowLog\fR=\fIwlog\fR, \fBwlog\fR=\fIwlog\fR
 Specify the maximum number of bits for a match distance\.
-.
 .IP
 The higher number of increases the chance to find a match which usually improves compression ratio\. It also increases memory requirements for the compressor and decompressor\. The minimum \fIwlog\fR is 10 (1 KiB) and the maximum is 30 (1 GiB) on 32\-bit platforms and 31 (2 GiB) on 64\-bit platforms\.
-.
 .IP
 Note: If \fBwindowLog\fR is set to larger than 27, \fB\-\-long=windowLog\fR or \fB\-\-memory=windowSize\fR needs to be passed to the decompressor\.
-.
 .TP
 \fBhashLog\fR=\fIhlog\fR, \fBhlog\fR=\fIhlog\fR
 Specify the maximum number of bits for a hash table\.
-.
 .IP
-Bigger hash tables cause less collisions which usually makes compression faster, but requires more memory during compression\.
-.
+Bigger hash tables cause fewer collisions which usually makes compression faster, but requires more memory during compression\.
 .IP
-The minimum \fIhlog\fR is 6 (64 B) and the maximum is 30 (1 GiB)\.
-.
+The minimum \fIhlog\fR is 6 (64 entries / 256 B) and the maximum is 30 (1B entries / 4 GiB)\.
 .TP
 \fBchainLog\fR=\fIclog\fR, \fBclog\fR=\fIclog\fR
-Specify the maximum number of bits for a hash chain or a binary tree\.
-.
+Specify the maximum number of bits for the secondary search structure, whose form depends on the selected \fBstrategy\fR\.
 .IP
-Higher numbers of bits increases the chance to find a match which usually improves compression ratio\. It also slows down compression speed and increases memory requirements for compression\. This option is ignored for the ZSTD_fast strategy\.
-.
+Higher numbers of bits increases the chance to find a match which usually improves compression ratio\. It also slows down compression speed and increases memory requirements for compression\. This option is ignored for the \fBZSTD_fast\fR \fBstrategy\fR, which only has the primary hash table\.
 .IP
-The minimum \fIclog\fR is 6 (64 B) and the maximum is 29 (524 Mib) on 32\-bit platforms and 30 (1 Gib) on 64\-bit platforms\.
-.
+The minimum \fIclog\fR is 6 (64 entries / 256 B) and the maximum is 29 (512M entries / 2 GiB) on 32\-bit platforms and 30 (1B entries / 4 GiB) on 64\-bit platforms\.
 .TP
 \fBsearchLog\fR=\fIslog\fR, \fBslog\fR=\fIslog\fR
 Specify the maximum number of searches in a hash chain or a binary tree using logarithmic scale\.
-.
 .IP
 More searches increases the chance to find a match which usually increases compression ratio but decreases compression speed\.
-.
 .IP
-The minimum \fIslog\fR is 1 and the maximum is \'windowLog\' \- 1\.
-.
+The minimum \fIslog\fR is 1 and the maximum is 'windowLog' \- 1\.
 .TP
 \fBminMatch\fR=\fImml\fR, \fBmml\fR=\fImml\fR
 Specify the minimum searched length of a match in a hash table\.
-.
 .IP
 Larger search lengths usually decrease compression ratio but improve decompression speed\.
-.
 .IP
 The minimum \fImml\fR is 3 and the maximum is 7\.
-.
 .TP
 \fBtargetLength\fR=\fItlen\fR, \fBtlen\fR=\fItlen\fR
 The impact of this field vary depending on selected strategy\.
-.
 .IP
-For ZSTD_btopt, ZSTD_btultra and ZSTD_btultra2, it specifies the minimum match length that causes match finder to stop searching\. A larger \fBtargetLength\fR usually improves compression ratio but decreases compression speed\. t For ZSTD_fast, it triggers ultra\-fast mode when > 0\. The value represents the amount of data skipped between match sampling\. Impact is reversed : a larger \fBtargetLength\fR increases compression speed but decreases compression ratio\.
-.
+For \fBZSTD_btopt\fR, \fBZSTD_btultra\fR and \fBZSTD_btultra2\fR, it specifies the minimum match length that causes match finder to stop searching\. A larger \fBtargetLength\fR usually improves compression ratio but decreases compression speed\.
+.IP
+For \fBZSTD_fast\fR, it triggers ultra\-fast mode when > 0\. The value represents the amount of data skipped between match sampling\. Impact is reversed: a larger \fBtargetLength\fR increases compression speed but decreases compression ratio\.
 .IP
 For all other strategies, this field has no impact\.
-.
 .IP
-The minimum \fItlen\fR is 0 and the maximum is 128 Kib\.
-.
+The minimum \fItlen\fR is 0 and the maximum is 128 KiB\.
 .TP
 \fBoverlapLog\fR=\fIovlog\fR, \fBovlog\fR=\fIovlog\fR
 Determine \fBoverlapSize\fR, amount of data reloaded from previous job\. This parameter is only available when multithreading is enabled\. Reloading more data improves compression ratio, but decreases speed\.
-.
 .IP
-The minimum \fIovlog\fR is 0, and the maximum is 9\. 1 means "no overlap", hence completely independent jobs\. 9 means "full overlap", meaning up to \fBwindowSize\fR is reloaded from previous job\. Reducing \fIovlog\fR by 1 reduces the reloaded amount by a factor 2\. For example, 8 means "windowSize/2", and 6 means "windowSize/8"\. Value 0 is special and means "default" : \fIovlog\fR is automatically determined by \fBzstd\fR\. In which case, \fIovlog\fR will range from 6 to 9, depending on selected \fIstrat\fR\.
-.
+The minimum \fIovlog\fR is 0, and the maximum is 9\. 1 means "no overlap", hence completely independent jobs\. 9 means "full overlap", meaning up to \fBwindowSize\fR is reloaded from previous job\. Reducing \fIovlog\fR by 1 reduces the reloaded amount by a factor 2\. For example, 8 means "windowSize/2", and 6 means "windowSize/8"\. Value 0 is special and means "default": \fIovlog\fR is automatically determined by \fBzstd\fR\. In which case, \fIovlog\fR will range from 6 to 9, depending on selected \fIstrat\fR\.
 .TP
 \fBldmHashLog\fR=\fIlhlog\fR, \fBlhlog\fR=\fIlhlog\fR
 Specify the maximum size for a hash table used for long distance matching\.
-.
 .IP
 This option is ignored unless long distance matching is enabled\.
-.
 .IP
 Bigger hash tables usually improve compression ratio at the expense of more memory during compression and a decrease in compression speed\.
-.
 .IP
 The minimum \fIlhlog\fR is 6 and the maximum is 30 (default: 20)\.
-.
 .TP
 \fBldmMinMatch\fR=\fIlmml\fR, \fBlmml\fR=\fIlmml\fR
 Specify the minimum searched length of a match for long distance matching\.
-.
 .IP
 This option is ignored unless long distance matching is enabled\.
-.
 .IP
 Larger/very small values usually decrease compression ratio\.
-.
 .IP
 The minimum \fIlmml\fR is 4 and the maximum is 4096 (default: 64)\.
-.
 .TP
 \fBldmBucketSizeLog\fR=\fIlblog\fR, \fBlblog\fR=\fIlblog\fR
 Specify the size of each bucket for the hash table used for long distance matching\.
-.
 .IP
 This option is ignored unless long distance matching is enabled\.
-.
 .IP
 Larger bucket sizes improve collision resolution but decrease compression speed\.
-.
 .IP
 The minimum \fIlblog\fR is 1 and the maximum is 8 (default: 3)\.
-.
 .TP
 \fBldmHashRateLog\fR=\fIlhrlog\fR, \fBlhrlog\fR=\fIlhrlog\fR
 Specify the frequency of inserting entries into the long distance matching hash table\.
-.
 .IP
 This option is ignored unless long distance matching is enabled\.
-.
 .IP
 Larger values will improve compression speed\. Deviating far from the default value will likely result in a decrease in compression ratio\.
-.
 .IP
 The default value is \fBwlog \- lhlog\fR\.
-.
 .SS "Example"
 The following parameters sets advanced compression options to something similar to predefined level 19 for files bigger than 256 KB:
-.
 .P
 \fB\-\-zstd\fR=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6
-.
+.SS "\-B#:"
+Specify the size of each compression job\. This parameter is only available when multi\-threading is enabled\. Each compression job is run in parallel, so this value indirectly impacts the nb of active threads\. Default job size varies depending on compression level (generally \fB4 * windowSize\fR)\. \fB\-B#\fR makes it possible to manually select a custom size\. Note that job size must respect a minimum value which is enforced transparently\. This minimum is either 512 KB, or \fBoverlapSize\fR, whichever is largest\. Different job sizes will lead to non\-identical compressed frames\.
+.SH "DICTIONARY BUILDER"
+\fBzstd\fR offers \fIdictionary\fR compression, which greatly improves efficiency on small files and messages\. It's possible to train \fBzstd\fR with a set of samples, the result of which is saved into a file called a \fBdictionary\fR\. Then, during compression and decompression, reference the same dictionary, using command \fB\-D dictionaryFileName\fR\. Compression of small files similar to the sample set will be greatly improved\.
+.TP
+\fB\-\-train FILEs\fR
+Use FILEs as training set to create a dictionary\. The training set should ideally contain a lot of samples (> 100), and weight typically 100x the target dictionary size (for example, ~10 MB for a 100 KB dictionary)\. \fB\-\-train\fR can be combined with \fB\-r\fR to indicate a directory rather than listing all the files, which can be useful to circumvent shell expansion limits\.
+.IP
+Since dictionary compression is mostly effective for small files, the expectation is that the training set will only contain small files\. In the case where some samples happen to be large, only the first 128 KiB of these samples will be used for training\.
+.IP
+\fB\-\-train\fR supports multithreading if \fBzstd\fR is compiled with threading support (default)\. Additional advanced parameters can be specified with \fB\-\-train\-fastcover\fR\. The legacy dictionary builder can be accessed with \fB\-\-train\-legacy\fR\. The slower cover dictionary builder can be accessed with \fB\-\-train\-cover\fR\. Default \fB\-\-train\fR is equivalent to \fB\-\-train\-fastcover=d=8,steps=4\fR\.
+.TP
+\fB\-o FILE\fR
+Dictionary saved into \fBFILE\fR (default name: dictionary)\.
+.TP
+\fB\-\-maxdict=#\fR
+Limit dictionary to specified size (default: 112640 bytes)\. As usual, quantities are expressed in bytes by default, and it's possible to employ suffixes (like \fBKB\fR or \fBMB\fR) to specify larger values\.
+.TP
+\fB\-#\fR
+Use \fB#\fR compression level during training (optional)\. Will generate statistics more tuned for selected compression level, resulting in a \fIsmall\fR compression ratio improvement for this level\.
+.TP
+\fB\-B#\fR
+Split input files into blocks of size # (default: no split)
+.TP
+\fB\-M#\fR, \fB\-\-memory=#\fR
+Limit the amount of sample data loaded for training (default: 2 GB)\. Note that the default (2 GB) is also the maximum\. This parameter can be useful in situations where the training set size is not well controlled and could be potentially very large\. Since speed of the training process is directly correlated to the size of the training sample set, a smaller sample set leads to faster training\.
+.IP
+In situations where the training set is larger than maximum memory, the CLI will randomly select samples among the available ones, up to the maximum allowed memory budget\. This is meant to improve dictionary relevance by mitigating the potential impact of clustering, such as selecting only files from the beginning of a list sorted by modification date, or sorted by alphabetical order\. The randomization process is deterministic, so training of the same list of files with the same parameters will lead to the creation of the same dictionary\.
+.TP
+\fB\-\-dictID=#\fR
+A dictionary ID is a locally unique ID\. The decoder will use this value to verify it is using the right dictionary\. By default, zstd will create a 4\-bytes random number ID\. It's possible to provide an explicit number ID instead\. It's up to the dictionary manager to not assign twice the same ID to 2 different dictionaries\. Note that short numbers have an advantage: an ID < 256 will only need 1 byte in the compressed frame header, and an ID < 65536 will only need 2 bytes\. This compares favorably to 4 bytes default\.
+.IP
+Note that RFC8878 reserves IDs less than 32768 and greater than or equal to 2\e^31, so they should not be used in public\.
+.TP
+\fB\-\-train\-cover[=k#,d=#,steps=#,split=#,shrink[=#]]\fR
+Select parameters for the default dictionary builder algorithm named cover\. If \fId\fR is not specified, then it tries \fId\fR = 6 and \fId\fR = 8\. If \fIk\fR is not specified, then it tries \fIsteps\fR values in the range [50, 2000]\. If \fIsteps\fR is not specified, then the default value of 40 is used\. If \fIsplit\fR is not specified or split <= 0, then the default value of 100 is used\. Requires that \fId\fR <= \fIk\fR\. If \fIshrink\fR flag is not used, then the default value for \fIshrinkDict\fR of 0 is used\. If \fIshrink\fR is not specified, then the default value for \fIshrinkDictMaxRegression\fR of 1 is used\.
+.IP
+Selects segments of size \fIk\fR with highest score to put in the dictionary\. The score of a segment is computed by the sum of the frequencies of all the subsegments of size \fId\fR\. Generally \fId\fR should be in the range [6, 8], occasionally up to 16, but the algorithm will run faster with d <= \fI8\fR\. Good values for \fIk\fR vary widely based on the input data, but a safe range is [2 * \fId\fR, 2000]\. If \fIsplit\fR is 100, all input samples are used for both training and testing to find optimal \fId\fR and \fIk\fR to build dictionary\. Supports multithreading if \fBzstd\fR is compiled with threading support\. Having \fIshrink\fR enabled takes a truncated dictionary of minimum size and doubles in size until compression ratio of the truncated dictionary is at most \fIshrinkDictMaxRegression%\fR worse than the compression ratio of the largest dictionary\.
+.IP
+Examples:
+.IP
+\fBzstd \-\-train\-cover FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=k=50,d=8 FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=d=8,steps=500 FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=k=50 FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=k=50,split=60 FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=shrink FILEs\fR
+.IP
+\fBzstd \-\-train\-cover=shrink=2 FILEs\fR
+.TP
+\fB\-\-train\-fastcover[=k#,d=#,f=#,steps=#,split=#,accel=#]\fR
+Same as cover but with extra parameters \fIf\fR and \fIaccel\fR and different default value of split If \fIsplit\fR is not specified, then it tries \fIsplit\fR = 75\. If \fIf\fR is not specified, then it tries \fIf\fR = 20\. Requires that 0 < \fIf\fR < 32\. If \fIaccel\fR is not specified, then it tries \fIaccel\fR = 1\. Requires that 0 < \fIaccel\fR <= 10\. Requires that \fId\fR = 6 or \fId\fR = 8\.
+.IP
+\fIf\fR is log of size of array that keeps track of frequency of subsegments of size \fId\fR\. The subsegment is hashed to an index in the range [0,2^\fIf\fR \- 1]\. It is possible that 2 different subsegments are hashed to the same index, and they are considered as the same subsegment when computing frequency\. Using a higher \fIf\fR reduces collision but takes longer\.
+.IP
+Examples:
+.IP
+\fBzstd \-\-train\-fastcover FILEs\fR
+.IP
+\fBzstd \-\-train\-fastcover=d=8,f=15,accel=2 FILEs\fR
+.TP
+\fB\-\-train\-legacy[=selectivity=#]\fR
+Use legacy dictionary builder algorithm with the given dictionary \fIselectivity\fR (default: 9)\. The smaller the \fIselectivity\fR value, the denser the dictionary, improving its efficiency but reducing its achievable maximum size\. \fB\-\-train\-legacy=s=#\fR is also accepted\.
+.IP
+Examples:
+.IP
+\fBzstd \-\-train\-legacy FILEs\fR
+.IP
+\fBzstd \-\-train\-legacy=selectivity=8 FILEs\fR
+.SH "BENCHMARK"
+The \fBzstd\fR CLI provides a benchmarking mode that can be used to easily find suitable compression parameters, or alternatively to benchmark a computer's performance\. \fBzstd \-b [FILE(s)]\fR will benchmark \fBzstd\fR for both compression and decompression using default compression level\. Note that results are very dependent on the content being compressed\. It's possible to pass multiple files to the benchmark, and even a directory with \fB\-r DIRECTORY\fR\. When no \fBFILE\fR is provided, the benchmark will use a procedurally generated \fBlorem ipsum\fR text\.
+.IP "\(bu" 4
+\fB\-b#\fR: benchmark file(s) using compression level #
+.IP "\(bu" 4
+\fB\-e#\fR: benchmark file(s) using multiple compression levels, from \fB\-b#\fR to \fB\-e#\fR (inclusive)
+.IP "\(bu" 4
+\fB\-d\fR: benchmark decompression speed only (requires providing a zstd\-compressed content)
+.IP "\(bu" 4
+\fB\-i#\fR: minimum evaluation time, in seconds (default: 3s), benchmark mode only
+.IP "\(bu" 4
+\fB\-B#\fR, \fB\-\-block\-size=#\fR: cut file(s) into independent chunks of size # (default: no chunking)
+.IP "\(bu" 4
+\fB\-S\fR: output one benchmark result per input file (default: consolidated result)
+.IP "\(bu" 4
+\fB\-D dictionary\fR benchmark using dictionary
+.IP "\(bu" 4
+\fB\-\-priority=rt\fR: set process priority to real\-time (Windows)
+.IP "" 0
+.P
+Beyond compression levels, benchmarking is also compatible with other parameters, such as number of threads (\fB\-T#\fR), advanced compression parameters (\fB\-\-zstd=###\fR), dictionary compression (\fB\-D dictionary\fR), or even disabling checksum verification for example\.
+.P
+\fBOutput Format:\fR CompressionLevel#Filename: InputSize \-> OutputSize (CompressionRatio), CompressionSpeed, DecompressionSpeed
+.P
+\fBMethodology:\fR For speed measurement, the entire input is compressed/decompressed in\-memory to measure speed\. A run lasts at least 1 sec, so when files are small, they are compressed/decompressed several times per run, in order to improve measurement accuracy\.
+.SH "SEE ALSO"
+\fBzstdgrep\fR(1), \fBzstdless\fR(1), \fBgzip\fR(1), \fBxz\fR(1)
+.P
+The \fIzstandard\fR format is specified in Y\. Collet, "Zstandard Compression and the 'application/zstd' Media Type", https://www\.ietf\.org/rfc/rfc8878\.txt, Internet RFC 8878 (February 2021)\.
 .SH "BUGS"
 Report bugs at: https://github\.com/facebook/zstd/issues
-.
 .SH "AUTHOR"
 Yann Collet
diff --git a/sys/contrib/zstd/programs/zstd.1.md b/sys/contrib/zstd/programs/zstd.1.md
index e343ec0448b4..e5c1b7fd215b 100644
--- a/sys/contrib/zstd/programs/zstd.1.md
+++ b/sys/contrib/zstd/programs/zstd.1.md
@@ -4,7 +4,7 @@ zstd(1) -- zstd, zstdmt, unzstd, zstdcat - Compress or decompress .zst files
 SYNOPSIS
 --------
 
-`zstd` [*OPTIONS*] [-|_INPUT-FILE_] [-o _OUTPUT-FILE_]
+`zstd` [<OPTIONS>] [-|<INPUT-FILE>] [-o <OUTPUT-FILE>]
 
 `zstdmt` is equivalent to `zstd -T0`
 
@@ -16,32 +16,33 @@ SYNOPSIS
 DESCRIPTION
 -----------
 `zstd` is a fast lossless compression algorithm and data compression tool,
-with command line syntax similar to `gzip (1)` and `xz (1)`.
+with command line syntax similar to `gzip`(1) and `xz`(1).
 It is based on the **LZ77** family, with further FSE & huff0 entropy stages.
 `zstd` offers highly configurable compression speed,
-with fast modes at > 200 MB/s per core,
-and strong modes nearing lzma compression ratios.
-It also features a very fast decoder, with speeds > 500 MB/s per core.
+from fast modes at > 200 MB/s per core,
+to strong modes with excellent compression ratios.
+It also features a very fast decoder, with speeds > 500 MB/s per core,
+which remains roughly stable at all compression settings.
 
 `zstd` command line syntax is generally similar to gzip,
-but features the following differences :
+but features the following few differences:
 
   - Source files are preserved by default.
     It's possible to remove them automatically by using the `--rm` command.
   - When compressing a single file, `zstd` displays progress notifications
     and result summary by default.
     Use `-q` to turn them off.
-  - `zstd` does not accept input from console,
-    but it properly accepts `stdin` when it's not the console.
   - `zstd` displays a short help page when command line is an error.
     Use `-q` to turn it off.
+  - `zstd` does not accept input from console,
+    though it does accept `stdin` when it's not the console.
+  - `zstd` does not store the input's filename or attributes, only its contents.
 
-`zstd` compresses or decompresses each _file_ according to the selected
-operation mode.
+`zstd` processes each _file_ according to the selected operation mode.
 If no _files_ are given or _file_ is `-`, `zstd` reads from standard input
 and writes the processed data to standard output.
 `zstd` will refuse to write compressed data to standard output
-if it is a terminal : it will display an error message and skip the _file_.
+if it is a terminal: it will display an error message and skip the file.
 Similarly, `zstd` will refuse to read compressed data from standard input
 if it is a terminal.
 
@@ -53,14 +54,15 @@ whose name is derived from the source _file_ name:
 * When decompressing, the `.zst` suffix is removed from the source filename to
   get the target filename
 
-### Concatenation with .zst files
-It is possible to concatenate `.zst` files as is.
-`zstd` will decompress such files as if they were a single `.zst` file.
+### Concatenation with .zst Files
+It is possible to concatenate multiple `.zst` files. `zstd` will decompress
+such agglomerated file as if it was a single `.zst` file.
 
 OPTIONS
 -------
 
-### Integer suffixes and special values
+### Integer Suffixes and Special Values
+
 In most places where an integer argument is expected,
 an optional suffix is supported to easily indicate large integers.
 There must be no space between the integer and the suffix.
@@ -72,7 +74,8 @@ There must be no space between the integer and the suffix.
     Multiply the integer by 1,048,576 (2\^20).
     `Mi`, `M`, and `MB` are accepted as synonyms for `MiB`.
 
-### Operation mode
+### Operation Mode
+
 If multiple operation mode options are given,
 the last one takes effect.
 
@@ -85,23 +88,29 @@ the last one takes effect.
     Decompress.
 * `-t`, `--test`:
     Test the integrity of compressed _files_.
-    This option is equivalent to `--decompress --stdout` except that the
-    decompressed data is discarded instead of being written to standard output.
+    This option is equivalent to `--decompress --stdout > /dev/null`,
+    decompressed data is discarded and checksummed for errors.
     No files are created or removed.
 * `-b#`:
-    Benchmark file(s) using compression level #
-* `--train FILEs`:
-    Use FILEs as a training set to create a dictionary.
+    Benchmark file(s) using compression level _#_.
+    See _BENCHMARK_ below for a description of this operation.
+* `--train FILES`:
+    Use _FILES_ as a training set to create a dictionary.
     The training set should contain a lot of small files (> 100).
+    See _DICTIONARY BUILDER_ below for a description of this operation.
 * `-l`, `--list`:
     Display information related to a zstd compressed file, such as size, ratio, and checksum.
     Some of these fields may not be available.
-    This command can be augmented with the `-v` modifier.
+    This command's output can be augmented with the `-v` modifier.
 
-### Operation modifiers
+### Operation Modifiers
 
 * `-#`:
-    `#` compression level \[1-19] (default: 3)
+    selects `#` compression level \[1-19\] (default: 3).
+    Higher compression levels *generally* produce higher compression ratio at the expense of speed and memory.
+    A rough rule of thumb is that compression speed is expected to be divided by 2 every 2 levels.
+    Technically, each level is mapped to a set of advanced parameters (that can also be modified individually, see below).
+    Because the compressor's behavior highly depends on the content to compress, there's no guarantee of a smooth progression from one level to another.
 * `--ultra`:
     unlocks high compression levels 20+ (maximum 22), using a lot more memory.
     Note that decompression will also require more memory when using these levels.
@@ -119,27 +128,31 @@ the last one takes effect.
     which is either 64 in 32-bit mode, or 256 for 64-bit environments.
     This modifier does nothing if `zstd` is compiled without multithread support.
 * `--single-thread`:
-    Does not spawn a thread for compression, use a single thread for both I/O and compression.
-    In this mode, compression is serialized with I/O, which is slightly slower.
-    (This is different from `-T1`, which spawns 1 compression thread in parallel of I/O).
-    This mode is the only one available when multithread support is disabled.
-    Single-thread mode features lower memory usage.
-    Final compressed result is slightly different from `-T1`.
+    Use a single thread for both I/O and compression.
+    As compression is serialized with I/O, this can be slightly slower.
+    Single-thread mode features significantly lower memory usage,
+    which can be useful for systems with limited amount of memory, such as 32-bit systems.
+
+    Note 1: this mode is the only available one when multithread support is disabled.
+
+    Note 2: this mode is different from `-T1`, which spawns 1 compression thread in parallel with I/O.
+    Final compressed result is also slightly different from `-T1`.
 * `--auto-threads={physical,logical} (default: physical)`:
     When using a default amount of threads via `-T0`, choose the default based on the number
     of detected physical or logical cores.
-* `--adapt[=min=#,max=#]` :
+* `--adapt[=min=#,max=#]`:
     `zstd` will dynamically adapt compression level to perceived I/O conditions.
     Compression level adaptation can be observed live by using command `-v`.
     Adaptation can be constrained between supplied `min` and `max` levels.
     The feature works when combined with multi-threading and `--long` mode.
     It does not work with `--single-thread`.
-    It sets window size to 8 MB by default (can be changed manually, see `wlog`).
+    It sets window size to 8 MiB by default (can be changed manually, see `wlog`).
     Due to the chaotic nature of dynamic adaptation, compressed result is not reproducible.
-    _note_ : at the time of this writing, `--adapt` can remain stuck at low speed
+
+    _Note_: at the time of this writing, `--adapt` can remain stuck at low speed
     when combined with multiple worker threads (>=2).
 * `--long[=#]`:
-    enables long distance matching with `#` `windowLog`, if not `#` is not
+    enables long distance matching with `#` `windowLog`, if `#` is not
     present it defaults to `27`.
     This increases the window size (`windowLog`) and memory usage for both the
     compressor and decompressor.
@@ -148,27 +161,37 @@ the last one takes effect.
 
     Note: If `windowLog` is set to larger than 27, `--long=windowLog` or
     `--memory=windowSize` needs to be passed to the decompressor.
+* `--max`:
+    set advanced parameters to maximum compression.
+    warning: this setting is very slow and uses a lot of resources.
+    It's inappropriate for 32-bit mode and therefore disabled in this mode.
 * `-D DICT`:
     use `DICT` as Dictionary to compress or decompress FILE(s)
 * `--patch-from FILE`:
     Specify the file to be used as a reference point for zstd's diff engine.
     This is effectively dictionary compression with some convenient parameter
-    selection, namely that windowSize > srcSize.
+    selection, namely that _windowSize_ > _srcSize_.
 
-    Note: cannot use both this and -D together
-    Note: `--long` mode will be automatically activated if chainLog < fileLog
-        (fileLog being the windowLog required to cover the whole file). You
+    Note: cannot use both this and `-D` together.
+
+    Note: `--long` mode will be automatically activated if _chainLog_ < _fileLog_
+        (_fileLog_ being the _windowLog_ required to cover the whole file). You
         can also manually force it.
-	Node: for all levels, you can use --patch-from in --single-thread mode
-		to improve compression ratio at the cost of speed
+
+    Note: up to level 15, you can use `--patch-from` in `--single-thread` mode
+        to improve compression ratio marginally at the cost of speed. Using
+        '--single-thread' above level 15  will lead to lower compression
+        ratios.
+
     Note: for level 19, you can get increased compression ratio at the cost
         of speed by specifying `--zstd=targetLength=` to be something large
-        (i.e 4096), and by setting a large `--zstd=chainLog=`
-* `--rsyncable` :
+        (i.e. 4096), and by setting a large `--zstd=chainLog=`.
+* `--rsyncable`:
     `zstd` will periodically synchronize the compression state to make the
-    compressed file more rsync-friendly. There is a negligible impact to
-    compression ratio, and the faster compression levels will see a small
-    compression speed hit.
+    compressed file more rsync-friendly.
+    There is a negligible impact to compression ratio,
+    and a potential impact to compression speed, perceptible at higher speeds,
+    for example when combining `--rsyncable` with many parallel worker threads.
     This feature does not work with `--single-thread`. You probably don't want
     to use it with long range mode, since it will decrease the effectiveness of
     the synchronization points, but your mileage may vary.
@@ -177,24 +200,24 @@ the last one takes effect.
 * `--[no-]content-size`:
     enable / disable whether or not the original size of the file is placed in
     the header of the compressed file. The default option is
-    --content-size (meaning that the original size will be placed in the header).
+    `--content-size` (meaning that the original size will be placed in the header).
 * `--no-dictID`:
     do not store dictionary ID within frame header (dictionary compression).
     The decoder will have to rely on implicit knowledge about which dictionary to use,
     it won't be able to check if it's correct.
 * `-M#`, `--memory=#`:
-    Set a memory usage limit. By default, Zstandard uses 128 MB for decompression
+    Set a memory usage limit. By default, `zstd` uses 128 MiB for decompression
     as the maximum amount of memory the decompressor is allowed to use, but you can
-    override this manually if need be in either direction (ie. you can increase or
+    override this manually if need be in either direction (i.e. you can increase or
     decrease it).
 
-    This is also used during compression when using with --patch-from=. In this case,
-    this parameter overrides that maximum size allowed for a dictionary. (128 MB).
+    This is also used during compression when using with `--patch-from=`. In this case,
+    this parameter overrides that maximum size allowed for a dictionary. (128 MiB).
 
     Additionally, this can be used to limit memory for dictionary training. This parameter
-    overrides the default limit of 2 GB. zstd will load training samples up to the memory limit
+    overrides the default limit of 2 GiB. zstd will load training samples up to the memory limit
     and ignore the rest.
-* `--stream-size=#` :
+* `--stream-size=#`:
     Sets the pledged source size of input coming from a stream. This value must be exact, as it
     will be included in the produced frame header. Incorrect stream sizes will cause an error.
     This information will be used to better optimize compression parameters, resulting in
@@ -206,13 +229,24 @@ the last one takes effect.
     expected. This feature allows for controlling the guess when needed.
     Exact guesses result in better compression ratios. Overestimates result in slightly
     degraded compression ratios, while underestimates may result in significant degradation.
-* `-o FILE`:
-    save result into `FILE`
+* `--target-compressed-block-size=#`:
+    Attempt to produce compressed blocks of approximately this size.
+    This will split larger blocks in order to approach this target.
+    This feature is notably useful for improved latency, when the receiver can leverage receiving early incomplete data.
+    This parameter defines a loose target: compressed blocks will target this size "on average", but individual blocks can still be larger or smaller.
+    Enabling this feature can decrease compression speed by up to ~10% at level 1.
+    Higher levels will see smaller relative speed regression, becoming invisible at higher settings.
 * `-f`, `--force`:
     disable input and output checks. Allows overwriting existing files, input
     from console, output to stdout, operating on links, block devices, etc.
+    During decompression and when the output destination is stdout, pass-through
+    unrecognized formats as-is.
 * `-c`, `--stdout`:
-    write to standard output (even if it is the console)
+    write to standard output (even if it is the console); keep original files (disable `--rm`).
+* `-o FILE`:
+    save result into `FILE`.
+    Note that this operation is in conflict with `-c`.
+    If both operations are present on the command line, the last expressed one wins.
 * `--[no-]sparse`:
     enable / disable sparse FS support,
     to make files with many zeroes smaller on disk.
@@ -221,9 +255,17 @@ the last one takes effect.
     default: enabled when output is into a file,
     and disabled when output is stdout.
     This setting overrides default and can force sparse mode over stdout.
+* `--[no-]pass-through`
+    enable / disable passing through uncompressed files as-is. During
+    decompression when pass-through is enabled, unrecognized formats will be
+    copied as-is from the input to the output. By default, pass-through will
+    occur when the output destination is stdout and the force (`-f`) option is
+    set.
 * `--rm`:
-    remove source file(s) after successful compression or decompression. If used in combination with
-    -o, will trigger a confirmation prompt (which can be silenced with -f), as this is a destructive operation.
+    remove source file(s) after successful compression or decompression.
+    This command is silently ignored if output is `stdout`.
+    If used in combination with `-o`,
+    triggers a confirmation prompt (which can be silenced with `-f`), as this is a destructive operation.
 * `-k`, `--keep`:
     keep source file(s) after successful compression or decompression.
     This is the default behavior.
@@ -261,10 +303,11 @@ the last one takes effect.
 * `-h`/`-H`, `--help`:
     display help/long help and exit
 * `-V`, `--version`:
-    display version number and exit.
-    Advanced : `-vV` also displays supported formats.
+    display version number and immediately exit.
+    note that, since it exits, flags specified after `-V` are effectively ignored.
+    Advanced: `-vV` also displays supported formats.
     `-vvV` also displays POSIX support.
-    `-q` will only display the version number, suitable for machine reading.
+    `-qV` will only display the version number, suitable for machine reading.
 * `-v`, `--verbose`:
     verbose mode, display more information
 * `-q`, `--quiet`:
@@ -273,19 +316,30 @@ the last one takes effect.
 * `--no-progress`:
     do not display the progress bar, but keep all other messages.
 * `--show-default-cparams`:
-    Shows the default compression parameters that will be used for a
-    particular src file. If the provided src file is not a regular file
-    (eg. named pipe), the cli will just output the default parameters.
-    That is, the parameters that are used when the src size is unknown.
+    shows the default compression parameters that will be used for a particular input file, based on the provided compression level and the input size.
+    If the provided file is not a regular file (e.g. a pipe), this flag will output the parameters used for inputs of unknown size.
+* `--exclude-compressed`:
+    only compress files that are not already compressed.
 * `--`:
     All arguments after `--` are treated as files
 
-### Restricted usage of Environment Variables
 
-Using environment variables to set parameters has security implications.
-Therefore, this avenue is intentionally restricted.
+### gzip Operation Modifiers
+When invoked via a `gzip` symlink, `zstd` will support further
+options that intend to mimic the `gzip` behavior:
+
+* `-n`, `--no-name`:
+    do not store the original filename and timestamps when compressing
+    a file. This is the default behavior and hence a no-op.
+* `--best`:
+    alias to the option `-9`.
+
+
+### Environment Variables
+Employing environment variables to set parameters has security implications.
+Therefore, this avenue is intentionally limited.
 Only `ZSTD_CLEVEL` and `ZSTD_NBTHREADS` are currently supported.
-They set the compression level and number of threads to use during compression, respectively.
+They set the default compression level and number of threads to use during compression, respectively.
 
 `ZSTD_CLEVEL` can be used to set the level between 1 and 19 (the "normal" range).
 If the value of `ZSTD_CLEVEL` is not a valid integer, it will be ignored with a warning message.
@@ -293,40 +347,209 @@ If the value of `ZSTD_CLEVEL` is not a valid integer, it will be ignored with a
 
 `ZSTD_NBTHREADS` can be used to set the number of threads `zstd` will attempt to use during compression.
 If the value of `ZSTD_NBTHREADS` is not a valid unsigned integer, it will be ignored with a warning message.
-`ZSTD_NBTHREADS` has a default value of (`1`), and is capped at ZSTDMT_NBWORKERS_MAX==200. `zstd` must be
-compiled with multithread support for this to have any effect.
+`ZSTD_NBTHREADS` has a default value of `max(1, min(4, nbCores/4))`, and is capped at ZSTDMT_NBWORKERS_MAX==200.
+`zstd` must be compiled with multithread support for this variable to have any effect.
 
 They can both be overridden by corresponding command line arguments:
 `-#` for compression level and `-T#` for number of compression threads.
 
 
+ADVANCED COMPRESSION OPTIONS
+----------------------------
+`zstd` provides 22 predefined regular compression levels plus the fast levels.
+A compression level is translated internally into multiple advanced parameters that control the behavior of the compressor
+(one can observe the result of this translation with `--show-default-cparams`).
+These advanced parameters can be overridden using advanced compression options.
+
+### --zstd[=options]:
+The _options_ are provided as a comma-separated list.
+You may specify only the options you want to change and the rest will be
+taken from the selected or default compression level.
+The list of available _options_:
+
+- `strategy`=_strat_, `strat`=_strat_:
+    Specify a strategy used by a match finder.
+
+    There are 9 strategies numbered from 1 to 9, from fastest to strongest:
+    1=`ZSTD_fast`, 2=`ZSTD_dfast`, 3=`ZSTD_greedy`,
+    4=`ZSTD_lazy`, 5=`ZSTD_lazy2`, 6=`ZSTD_btlazy2`,
+    7=`ZSTD_btopt`, 8=`ZSTD_btultra`, 9=`ZSTD_btultra2`.
+
+- `windowLog`=_wlog_, `wlog`=_wlog_:
+    Specify the maximum number of bits for a match distance.
+
+    The higher number of increases the chance to find a match which usually
+    improves compression ratio.
+    It also increases memory requirements for the compressor and decompressor.
+    The minimum _wlog_ is 10 (1 KiB) and the maximum is 30 (1 GiB) on 32-bit
+    platforms and 31 (2 GiB) on 64-bit platforms.
+
+    Note: If `windowLog` is set to larger than 27, `--long=windowLog` or
+    `--memory=windowSize` needs to be passed to the decompressor.
+
+- `hashLog`=_hlog_, `hlog`=_hlog_:
+    Specify the maximum number of bits for a hash table.
+
+    Bigger hash tables cause fewer collisions which usually makes compression
+    faster, but requires more memory during compression.
+
+    The minimum _hlog_ is 6 (64 entries / 256 B) and the maximum is 30 (1B entries / 4 GiB).
+
+- `chainLog`=_clog_, `clog`=_clog_:
+    Specify the maximum number of bits for the secondary search structure,
+    whose form depends on the selected `strategy`.
+
+    Higher numbers of bits increases the chance to find a match which usually
+    improves compression ratio.
+    It also slows down compression speed and increases memory requirements for
+    compression.
+    This option is ignored for the `ZSTD_fast` `strategy`, which only has the primary hash table.
+
+    The minimum _clog_ is 6 (64 entries / 256 B) and the maximum is 29 (512M entries / 2 GiB) on 32-bit platforms
+    and 30 (1B entries / 4 GiB) on 64-bit platforms.
+
+- `searchLog`=_slog_, `slog`=_slog_:
+    Specify the maximum number of searches in a hash chain or a binary tree
+    using logarithmic scale.
+
+    More searches increases the chance to find a match which usually increases
+    compression ratio but decreases compression speed.
+
+    The minimum _slog_ is 1 and the maximum is 'windowLog' - 1.
+
+- `minMatch`=_mml_, `mml`=_mml_:
+    Specify the minimum searched length of a match in a hash table.
+
+    Larger search lengths usually decrease compression ratio but improve
+    decompression speed.
+
+    The minimum _mml_ is 3 and the maximum is 7.
+
+- `targetLength`=_tlen_, `tlen`=_tlen_:
+    The impact of this field vary depending on selected strategy.
+
+    For `ZSTD_btopt`, `ZSTD_btultra` and `ZSTD_btultra2`, it specifies
+    the minimum match length that causes match finder to stop searching.
+    A larger `targetLength` usually improves compression ratio
+    but decreases compression speed.
+
+    For `ZSTD_fast`, it triggers ultra-fast mode when > 0.
+    The value represents the amount of data skipped between match sampling.
+    Impact is reversed: a larger `targetLength` increases compression speed
+    but decreases compression ratio.
+
+    For all other strategies, this field has no impact.
+
+    The minimum _tlen_ is 0 and the maximum is 128 KiB.
+
+- `overlapLog`=_ovlog_,  `ovlog`=_ovlog_:
+    Determine `overlapSize`, amount of data reloaded from previous job.
+    This parameter is only available when multithreading is enabled.
+    Reloading more data improves compression ratio, but decreases speed.
+
+    The minimum _ovlog_ is 0, and the maximum is 9.
+    1 means "no overlap", hence completely independent jobs.
+    9 means "full overlap", meaning up to `windowSize` is reloaded from previous job.
+    Reducing _ovlog_ by 1 reduces the reloaded amount by a factor 2.
+    For example, 8 means "windowSize/2", and 6 means "windowSize/8".
+    Value 0 is special and means "default": _ovlog_ is automatically determined by `zstd`.
+    In which case, _ovlog_ will range from 6 to 9, depending on selected _strat_.
+
+- `ldmHashRateLog`=_lhrlog_, `lhrlog`=_lhrlog_:
+    Specify the frequency of inserting entries into the long distance matching
+    hash table.
+
+    This option is ignored unless long distance matching is enabled.
+
+    Larger values will improve compression speed. Deviating far from the
+    default value will likely result in a decrease in compression ratio.
+
+    The default value varies between 4 and 7, depending on `strategy`.
+
+- `ldmHashLog`=_lhlog_, `lhlog`=_lhlog_:
+    Specify the maximum size for a hash table used for long distance matching.
+
+    This option is ignored unless long distance matching is enabled.
+
+    Bigger hash tables usually improve compression ratio at the expense of more
+    memory during compression and a decrease in compression speed.
+
+    The minimum _lhlog_ is 6 and the maximum is 30 (default: `windowLog - ldmHashRateLog`).
+
+- `ldmMinMatch`=_lmml_, `lmml`=_lmml_:
+    Specify the minimum searched length of a match for long distance matching.
+
+    This option is ignored unless long distance matching is enabled.
+
+    Larger/very small values usually decrease compression ratio.
+
+    The minimum _lmml_ is 4 and the maximum is 4096 (default: 32 to 64, depending on `strategy`).
+
+- `ldmBucketSizeLog`=_lblog_, `lblog`=_lblog_:
+    Specify the size of each bucket for the hash table used for long distance
+    matching.
+
+    This option is ignored unless long distance matching is enabled.
+
+    Larger bucket sizes improve collision resolution but decrease compression
+    speed.
+
+    The minimum _lblog_ is 1 and the maximum is 8 (default: 4 to 8, depending on `strategy`).
+
+
+### Example
+The following parameters sets advanced compression options to something
+similar to predefined level 19 for files bigger than 256 KB:
+
+`--zstd`=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6
+
+### -B#:
+Specify the size of each compression job.
+This parameter is only available when multi-threading is enabled.
+Each compression job is run in parallel, so this value indirectly impacts the nb of active threads.
+Default job size varies depending on compression level (generally  `4 * windowSize`).
+`-B#` makes it possible to manually select a custom size.
+Note that job size must respect a minimum value which is enforced transparently.
+This minimum is either 512 KB, or `overlapSize`, whichever is largest.
+Different job sizes will lead to non-identical compressed frames.
+
+
 DICTIONARY BUILDER
 ------------------
 `zstd` offers _dictionary_ compression,
 which greatly improves efficiency on small files and messages.
 It's possible to train `zstd` with a set of samples,
 the result of which is saved into a file called a `dictionary`.
-Then during compression and decompression, reference the same dictionary,
+Then, during compression and decompression, reference the same dictionary,
 using command `-D dictionaryFileName`.
 Compression of small files similar to the sample set will be greatly improved.
 
 * `--train FILEs`:
     Use FILEs as training set to create a dictionary.
-    The training set should contain a lot of small files (> 100),
+    The training set should ideally contain a lot of samples (> 100),
     and weight typically 100x the target dictionary size
-    (for example, 10 MB for a 100 KB dictionary).
+    (for example, ~10 MB for a 100 KB dictionary).
     `--train` can be combined with `-r` to indicate a directory rather than listing all the files,
     which can be useful to circumvent shell expansion limits.
 
+    Since dictionary compression is mostly effective for small files,
+    the expectation is that the training set will only contain small files.
+    In the case where some samples happen to be large,
+    only the first 128 KiB of these samples will be used for training.
+
     `--train` supports multithreading if `zstd` is compiled with threading support (default).
-    Additional parameters can be specified with `--train-fastcover`.
+    Additional advanced parameters can be specified with `--train-fastcover`.
     The legacy dictionary builder can be accessed with `--train-legacy`.
     The slower cover dictionary builder can be accessed with `--train-cover`.
-    Default is equivalent to `--train-fastcover=d=8,steps=4`.
-* `-o file`:
-    Dictionary saved into `file` (default name: dictionary).
+    Default `--train` is equivalent to `--train-fastcover=d=8,steps=4`.
+
+* `-o FILE`:
+    Dictionary saved into `FILE` (default name: dictionary).
 * `--maxdict=#`:
-    Limit dictionary to specified size (default: 112640).
+    Limit dictionary to specified size (default: 112640 bytes).
+    As usual, quantities are expressed in bytes by default,
+    and it's possible to employ suffixes (like `KB` or `MB`)
+    to specify larger values.
 * `-#`:
     Use `#` compression level during training (optional).
     Will generate statistics more tuned for selected compression level,
@@ -334,17 +557,39 @@ Compression of small files similar to the sample set will be greatly improved.
 * `-B#`:
     Split input files into blocks of size # (default: no split)
 * `-M#`, `--memory=#`:
-    Limit the amount of sample data loaded for training (default: 2 GB). See above for details.
+    Limit the amount of sample data loaded for training (default: 2 GB).
+    Note that the default (2 GB) is also the maximum.
+    This parameter can be useful in situations where the training set size
+    is not well controlled and could be potentially very large.
+    Since speed of the training process is directly correlated to
+    the size of the training sample set,
+    a smaller sample set leads to faster training.
+
+    In situations where the training set is larger than maximum memory,
+    the CLI will randomly select samples among the available ones,
+    up to the maximum allowed memory budget.
+    This is meant to improve dictionary relevance
+    by mitigating the potential impact of clustering,
+    such as selecting only files from the beginning of a list
+    sorted by modification date, or sorted by alphabetical order.
+    The randomization process is deterministic, so
+    training of the same list of files with the same parameters
+    will lead to the creation of the same dictionary.
+
 * `--dictID=#`:
-    A dictionary ID is a locally unique ID
-    that a decoder can use to verify it is using the right dictionary.
+    A dictionary ID is a locally unique ID.
+    The decoder will use this value to verify it is using the right dictionary.
     By default, zstd will create a 4-bytes random number ID.
-    It's possible to give a precise number instead.
-    Short numbers have an advantage : an ID < 256 will only need 1 byte in the
-    compressed frame header, and an ID < 65536 will only need 2 bytes.
-    This compares favorably to 4 bytes default.
-    However, it's up to the dictionary manager to not assign twice the same ID to
+    It's possible to provide an explicit number ID instead.
+    It's up to the dictionary manager to not assign twice the same ID to
     2 different dictionaries.
+    Note that short numbers have an advantage:
+    an ID < 256 will only need 1 byte in the compressed frame header,
+    and an ID < 65536 will only need 2 bytes.
+    This compares favorably to 4 bytes default.
+
+    Note that RFC8878 reserves IDs less than 32768 and greater than or equal to 2\^31, so they should not be used in public.
+
 * `--train-cover[=k#,d=#,steps=#,split=#,shrink[=#]]`:
     Select parameters for the default dictionary builder algorithm named cover.
     If _d_ is not specified, then it tries _d_ = 6 and _d_ = 8.
@@ -409,7 +654,7 @@ Compression of small files similar to the sample set will be greatly improved.
     Use legacy dictionary builder algorithm with the given dictionary
     _selectivity_ (default: 9).
     The smaller the _selectivity_ value, the denser the dictionary,
-    improving its efficiency but reducing its possible maximum size.
+    improving its efficiency but reducing its achievable maximum size.
     `--train-legacy=s=#` is also accepted.
 
     Examples:
@@ -421,177 +666,44 @@ Compression of small files similar to the sample set will be greatly improved.
 
 BENCHMARK
 ---------
+The `zstd` CLI provides a benchmarking mode that can be used to easily find suitable compression parameters, or alternatively to benchmark a computer's performance.
+`zstd -b [FILE(s)]` will benchmark `zstd` for both compression and decompression using default compression level.
+Note that results are very dependent on the content being compressed.
+
+It's possible to pass multiple files to the benchmark, and even a directory with `-r DIRECTORY`.
+When no `FILE` is provided, the benchmark will use a procedurally generated `lorem ipsum` text.
+
+Benchmarking will employ `max(1, min(4, nbCores/4))` worker threads by default in order to match the behavior of the normal CLI I/O.
 
 * `-b#`:
     benchmark file(s) using compression level #
 * `-e#`:
     benchmark file(s) using multiple compression levels, from `-b#` to `-e#` (inclusive)
+* `-d`:
+    benchmark decompression speed only (requires providing a zstd-compressed content)
 * `-i#`:
     minimum evaluation time, in seconds (default: 3s), benchmark mode only
 * `-B#`, `--block-size=#`:
-    cut file(s) into independent blocks of size # (default: no block)
+    cut file(s) into independent chunks of size # (default: no chunking)
+* `-S`:
+    output one benchmark result per input file (default: consolidated result)
+* `-D dictionary`
+    benchmark using dictionary
 * `--priority=rt`:
-    set process priority to real-time
-
-**Output Format:** CompressionLevel#Filename : IntputSize -> OutputSize (CompressionRatio), CompressionSpeed, DecompressionSpeed
-
-**Methodology:** For both compression and decompression speed, the entire input is compressed/decompressed in-memory to measure speed. A run lasts at least 1 sec, so when files are small, they are compressed/decompressed several times per run, in order to improve measurement accuracy.
-
-ADVANCED COMPRESSION OPTIONS
-----------------------------
-### -B#:
-Select the size of each compression job.
-This parameter is only available when multi-threading is enabled.
-Each compression job is run in parallel, so this value indirectly impacts the nb of active threads.
-Default job size varies depending on compression level (generally  `4 * windowSize`).
-`-B#` makes it possible to manually select a custom size.
-Note that job size must respect a minimum value which is enforced transparently.
-This minimum is either 512 KB, or `overlapSize`, whichever is largest.
-Different job sizes will lead to (slightly) different compressed frames.
-
-### --zstd[=options]:
-`zstd` provides 22 predefined compression levels.
-The selected or default predefined compression level can be changed with
-advanced compression options.
-The _options_ are provided as a comma-separated list.
-You may specify only the options you want to change and the rest will be
-taken from the selected or default compression level.
-The list of available _options_:
-
-- `strategy`=_strat_, `strat`=_strat_:
-    Specify a strategy used by a match finder.
-
-    There are 9 strategies numbered from 1 to 9, from faster to stronger:
-    1=ZSTD\_fast, 2=ZSTD\_dfast, 3=ZSTD\_greedy,
-    4=ZSTD\_lazy, 5=ZSTD\_lazy2, 6=ZSTD\_btlazy2,
-    7=ZSTD\_btopt, 8=ZSTD\_btultra, 9=ZSTD\_btultra2.
-
-- `windowLog`=_wlog_, `wlog`=_wlog_:
-    Specify the maximum number of bits for a match distance.
-
-    The higher number of increases the chance to find a match which usually
-    improves compression ratio.
-    It also increases memory requirements for the compressor and decompressor.
-    The minimum _wlog_ is 10 (1 KiB) and the maximum is 30 (1 GiB) on 32-bit
-    platforms and 31 (2 GiB) on 64-bit platforms.
-
-    Note: If `windowLog` is set to larger than 27, `--long=windowLog` or
-    `--memory=windowSize` needs to be passed to the decompressor.
-
-- `hashLog`=_hlog_, `hlog`=_hlog_:
-    Specify the maximum number of bits for a hash table.
-
-    Bigger hash tables cause less collisions which usually makes compression
-    faster, but requires more memory during compression.
-
-    The minimum _hlog_ is 6 (64 B) and the maximum is 30 (1 GiB).
-
-- `chainLog`=_clog_, `clog`=_clog_:
-    Specify the maximum number of bits for a hash chain or a binary tree.
-
-    Higher numbers of bits increases the chance to find a match which usually
-    improves compression ratio.
-    It also slows down compression speed and increases memory requirements for
-    compression.
-    This option is ignored for the ZSTD_fast strategy.
-
-    The minimum _clog_ is 6 (64 B) and the maximum is 29 (524 Mib) on 32-bit platforms
-    and 30 (1 Gib) on 64-bit platforms.
-
-- `searchLog`=_slog_, `slog`=_slog_:
-    Specify the maximum number of searches in a hash chain or a binary tree
-    using logarithmic scale.
-
-    More searches increases the chance to find a match which usually increases
-    compression ratio but decreases compression speed.
-
-    The minimum _slog_ is 1 and the maximum is 'windowLog' - 1.
-
-- `minMatch`=_mml_, `mml`=_mml_:
-    Specify the minimum searched length of a match in a hash table.
-
-    Larger search lengths usually decrease compression ratio but improve
-    decompression speed.
-
-    The minimum _mml_ is 3 and the maximum is 7.
-
-- `targetLength`=_tlen_, `tlen`=_tlen_:
-    The impact of this field vary depending on selected strategy.
-
-    For ZSTD\_btopt, ZSTD\_btultra and ZSTD\_btultra2, it specifies
-    the minimum match length that causes match finder to stop searching.
-    A larger `targetLength` usually improves compression ratio
-    but decreases compression speed.
-t
-    For ZSTD\_fast, it triggers ultra-fast mode when > 0.
-    The value represents the amount of data skipped between match sampling.
-    Impact is reversed : a larger `targetLength` increases compression speed
-    but decreases compression ratio.
-
-    For all other strategies, this field has no impact.
-
-    The minimum _tlen_ is 0 and the maximum is 128 Kib.
-
-- `overlapLog`=_ovlog_,  `ovlog`=_ovlog_:
-    Determine `overlapSize`, amount of data reloaded from previous job.
-    This parameter is only available when multithreading is enabled.
-    Reloading more data improves compression ratio, but decreases speed.
-
-    The minimum _ovlog_ is 0, and the maximum is 9.
-    1 means "no overlap", hence completely independent jobs.
-    9 means "full overlap", meaning up to `windowSize` is reloaded from previous job.
-    Reducing _ovlog_ by 1 reduces the reloaded amount by a factor 2.
-    For example, 8 means "windowSize/2", and 6 means "windowSize/8".
-    Value 0 is special and means "default" : _ovlog_ is automatically determined by `zstd`.
-    In which case, _ovlog_ will range from 6 to 9, depending on selected _strat_.
-
-- `ldmHashLog`=_lhlog_, `lhlog`=_lhlog_:
-    Specify the maximum size for a hash table used for long distance matching.
+    set process priority to real-time (Windows)
 
-    This option is ignored unless long distance matching is enabled.
-
-    Bigger hash tables usually improve compression ratio at the expense of more
-    memory during compression and a decrease in compression speed.
-
-    The minimum _lhlog_ is 6 and the maximum is 30 (default: 20).
-
-- `ldmMinMatch`=_lmml_, `lmml`=_lmml_:
-    Specify the minimum searched length of a match for long distance matching.
-
-    This option is ignored unless long distance matching is enabled.
-
-    Larger/very small values usually decrease compression ratio.
-
-    The minimum _lmml_ is 4 and the maximum is 4096 (default: 64).
-
-- `ldmBucketSizeLog`=_lblog_, `lblog`=_lblog_:
-    Specify the size of each bucket for the hash table used for long distance
-    matching.
-
-    This option is ignored unless long distance matching is enabled.
-
-    Larger bucket sizes improve collision resolution but decrease compression
-    speed.
-
-    The minimum _lblog_ is 1 and the maximum is 8 (default: 3).
-
-- `ldmHashRateLog`=_lhrlog_, `lhrlog`=_lhrlog_:
-    Specify the frequency of inserting entries into the long distance matching
-    hash table.
+Beyond compression levels, benchmarking is also compatible with other parameters, such as number of threads (`-T#`), advanced compression parameters (`--zstd=###`), dictionary compression (`-D dictionary`), or even disabling checksum verification for example.
 
-    This option is ignored unless long distance matching is enabled.
+**Output Format:** CompressionLevel#Filename: InputSize -> OutputSize (CompressionRatio), CompressionSpeed, DecompressionSpeed
 
-    Larger values will improve compression speed. Deviating far from the
-    default value will likely result in a decrease in compression ratio.
+**Methodology:** For speed measurement, the entire input is compressed/decompressed in-memory to measure speed. A run lasts at least 1 sec, so when files are small, they are compressed/decompressed several times per run, in order to improve measurement accuracy.
 
-    The default value is `wlog - lhlog`.
 
-### Example
-The following parameters sets advanced compression options to something
-similar to predefined level 19 for files bigger than 256 KB:
-
-`--zstd`=wlog=23,clog=23,hlog=22,slog=6,mml=3,tlen=48,strat=6
+SEE ALSO
+--------
+`zstdgrep`(1), `zstdless`(1), `gzip`(1), `xz`(1)
 
+The <zstandard> format is specified in Y. Collet, "Zstandard Compression and the 'application/zstd' Media Type", https://www.ietf.org/rfc/rfc8878.txt, Internet RFC 8878 (February 2021).
 
 BUGS
 ----
diff --git a/sys/contrib/zstd/programs/zstdcli.c b/sys/contrib/zstd/programs/zstdcli.c
index bfe18c0c1ba3..83d9b881e502 100644
--- a/sys/contrib/zstd/programs/zstdcli.c
+++ b/sys/contrib/zstd/programs/zstdcli.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Yann Collet, Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -8,36 +8,19 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
-
-/*-************************************
-*  Tuning parameters
-**************************************/
-#ifndef ZSTDCLI_CLEVEL_DEFAULT
-#  define ZSTDCLI_CLEVEL_DEFAULT 3
-#endif
-
-#ifndef ZSTDCLI_CLEVEL_MAX
-#  define ZSTDCLI_CLEVEL_MAX 19   /* without using --ultra */
-#endif
-
-#ifndef ZSTDCLI_NBTHREADS_DEFAULT
-#  define ZSTDCLI_NBTHREADS_DEFAULT 1
-#endif
-
 /*-************************************
 *  Dependencies
 **************************************/
-#include "platform.h" /* IS_CONSOLE, PLATFORM_POSIX_VERSION */
-#include "util.h"     /* UTIL_HAS_CREATEFILELIST, UTIL_createFileList */
+#include "platform.h" /* PLATFORM_POSIX_VERSION */
+#include "util.h"     /* UTIL_HAS_CREATEFILELIST, UTIL_createFileList, UTIL_isConsole */
 #include <stdlib.h>   /* getenv */
 #include <string.h>   /* strcmp, strlen */
 #include <stdio.h>    /* fprintf(), stdin, stdout, stderr */
-#include <errno.h>    /* errno */
 #include <assert.h>   /* assert */
 
 #include "fileio.h"   /* stdinmark, stdoutmark, ZSTD_EXTENSION */
 #ifndef ZSTD_NOBENCH
-#  include "benchzstd.h"  /* BMK_benchFiles */
+#  include "benchzstd.h"  /* BMK_benchFilesAdvanced */
 #endif
 #ifndef ZSTD_NODICT
 #  include "dibio.h"  /* ZDICT_cover_params_t, DiB_trainFromFiles() */
@@ -46,17 +29,35 @@
 #  include "zstdcli_trace.h"
 #endif
 #include "../lib/zstd.h"  /* ZSTD_VERSION_STRING, ZSTD_minCLevel, ZSTD_maxCLevel */
+#include "fileio_asyncio.h"
+#include "fileio_common.h"
+
+/*-************************************
+*  Tuning parameters
+**************************************/
+#ifndef ZSTDCLI_CLEVEL_DEFAULT
+#  define ZSTDCLI_CLEVEL_DEFAULT 3
+#endif
+
+#ifndef ZSTDCLI_CLEVEL_MAX
+#  define ZSTDCLI_CLEVEL_MAX 19   /* without using --ultra */
+#endif
+
+#ifndef ZSTDCLI_NBTHREADS_DEFAULT
+#define ZSTDCLI_NBTHREADS_DEFAULT MAX(1, MIN(4, UTIL_countLogicalCores() / 4))
+#endif
+
 
 
 /*-************************************
 *  Constants
 **************************************/
-#define COMPRESSOR_NAME "zstd command line interface"
+#define COMPRESSOR_NAME "Zstandard CLI"
 #ifndef ZSTD_VERSION
 #  define ZSTD_VERSION "v" ZSTD_VERSION_STRING
 #endif
 #define AUTHOR "Yann Collet"
-#define WELCOME_MESSAGE "*** %s %i-bits %s, by %s ***\n", COMPRESSOR_NAME, (int)(sizeof(size_t)*8), ZSTD_VERSION, AUTHOR
+#define WELCOME_MESSAGE "*** %s (%i-bit) %s, by %s ***\n", COMPRESSOR_NAME, (int)(sizeof(size_t)*8), ZSTD_VERSION, AUTHOR
 
 #define ZSTD_ZSTDMT "zstdmt"
 #define ZSTD_UNZSTD "unzstd"
@@ -99,9 +100,7 @@ typedef enum { cover, fastCover, legacy } dictType;
 /*-************************************
 *  Display Macros
 **************************************/
-#define DISPLAY_F(f, ...)    fprintf((f), __VA_ARGS__)
-#define DISPLAYOUT(...)      DISPLAY_F(stdout, __VA_ARGS__)
-#define DISPLAY(...)         DISPLAY_F(stderr, __VA_ARGS__)
+#undef DISPLAYLEVEL
 #define DISPLAYLEVEL(l, ...) { if (g_displayLevel>=l) { DISPLAY(__VA_ARGS__); } }
 static int g_displayLevel = DISPLAY_LEVEL_DEFAULT;   /* 0 : no display,  1: errors,  2 : + result + interaction + warnings,  3 : + progression,  4 : + information */
 
@@ -124,159 +123,201 @@ static void checkLibVersion(void)
 }
 
 
+/*! exeNameMatch() :
+    @return : a non-zero value if exeName matches test, excluding the extension
+   */
+static int exeNameMatch(const char* exeName, const char* test)
+{
+    return !strncmp(exeName, test, strlen(test)) &&
+        (exeName[strlen(test)] == '\0' || exeName[strlen(test)] == '.');
+}
+
 /*-************************************
 *  Command Line
 **************************************/
 /* print help either in `stderr` or `stdout` depending on originating request
- * error (badusage) => stderr
- * help (usage_advanced) => stdout
+ * error (badUsage) => stderr
+ * help (usageAdvanced) => stdout
  */
 static void usage(FILE* f, const char* programName)
 {
-    DISPLAY_F(f, "Usage : \n");
-    DISPLAY_F(f, "      %s [args] [FILE(s)] [-o file] \n", programName);
+    DISPLAY_F(f, "Compress or decompress the INPUT file(s); reads from STDIN if INPUT is `-` or not provided.\n\n");
+    DISPLAY_F(f, "Usage: %s [OPTIONS...] [INPUT... | -] [-o OUTPUT]\n\n", programName);
+    DISPLAY_F(f, "Options:\n");
+    DISPLAY_F(f, "  -o OUTPUT                     Write output to a single file, OUTPUT.\n");
+    DISPLAY_F(f, "  -k, --keep                    Preserve INPUT file(s). [Default] \n");
+    DISPLAY_F(f, "  --rm                          Remove INPUT file(s) after successful (de)compression.\n");
+#ifdef ZSTD_GZCOMPRESS
+    if (exeNameMatch(programName, ZSTD_GZ)) {     /* behave like gzip */
+        DISPLAY_F(f, "  -n, --no-name                 Do not store original filename when compressing.\n\n");
+    }
+#endif
     DISPLAY_F(f, "\n");
-    DISPLAY_F(f, "FILE    : a filename \n");
-    DISPLAY_F(f, "          with no FILE, or when FILE is - , read standard input\n");
-    DISPLAY_F(f, "Arguments : \n");
 #ifndef ZSTD_NOCOMPRESS
-    DISPLAY_F(f, " -#     : # compression level (1-%d, default: %d) \n", ZSTDCLI_CLEVEL_MAX, ZSTDCLI_CLEVEL_DEFAULT);
+    DISPLAY_F(f, "  -#                            Desired compression level, where `#` is a number between 1 and %d;\n", ZSTDCLI_CLEVEL_MAX);
+    DISPLAY_F(f, "                                lower numbers provide faster compression, higher numbers yield\n");
+    DISPLAY_F(f, "                                better compression ratios. [Default: %d]\n\n", ZSTDCLI_CLEVEL_DEFAULT);
 #endif
 #ifndef ZSTD_NODECOMPRESS
-    DISPLAY_F(f, " -d     : decompression \n");
+    DISPLAY_F(f, "  -d, --decompress              Perform decompression.\n");
 #endif
-    DISPLAY_F(f, " -D DICT: use DICT as Dictionary for compression or decompression \n");
-    DISPLAY_F(f, " -o file: result stored into `file` (only 1 output file) \n");
-    DISPLAY_F(f, " -f     : disable input and output checks. Allows overwriting existing files,\n");
-    DISPLAY_F(f, "          input from console, output to stdout, operating on links,\n");
-    DISPLAY_F(f, "          block devices, etc.\n");
-    DISPLAY_F(f, "--rm    : remove source file(s) after successful de/compression \n");
-    DISPLAY_F(f, " -k     : preserve source file(s) (default) \n");
-    DISPLAY_F(f, " -h/-H  : display help/long help and exit \n");
+    DISPLAY_F(f, "  -D DICT                       Use DICT as the dictionary for compression or decompression.\n\n");
+    DISPLAY_F(f, "  -f, --force                   Disable input and output checks. Allows overwriting existing files,\n");
+    DISPLAY_F(f, "                                receiving input from the console, printing output to STDOUT, and\n");
+    DISPLAY_F(f, "                                operating on links, block devices, etc. Unrecognized formats will be\n");
+    DISPLAY_F(f, "                                passed-through through as-is.\n\n");
+
+    DISPLAY_F(f, "  -h                            Display short usage and exit.\n");
+    DISPLAY_F(f, "  -H, --help                    Display full help and exit.\n");
+    DISPLAY_F(f, "  -V, --version                 Display the program version and exit.\n");
+    DISPLAY_F(f, "\n");
 }
 
-static void usage_advanced(const char* programName)
+static void usageAdvanced(const char* programName)
 {
     DISPLAYOUT(WELCOME_MESSAGE);
+    DISPLAYOUT("\n");
     usage(stdout, programName);
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "Advanced arguments : \n");
-    DISPLAYOUT( " -V     : display Version number and exit \n");
-
-    DISPLAYOUT( " -c     : write to standard output (even if it is the console) \n");
+    DISPLAYOUT("Advanced options:\n");
+    DISPLAYOUT("  -c, --stdout                  Write to STDOUT (even if it is a console) and keep the INPUT file(s).\n\n");
 
-    DISPLAYOUT( " -v     : verbose mode; specify multiple times to increase verbosity \n");
-    DISPLAYOUT( " -q     : suppress warnings; specify twice to suppress errors too \n");
-    DISPLAYOUT( "--[no-]progress : forcibly display, or never display the progress counter.\n");
-    DISPLAYOUT( "                  note: any (de)compressed output to terminal will mix with progress counter text. \n");
+    DISPLAYOUT("  -v, --verbose                 Enable verbose output; pass multiple times to increase verbosity.\n");
+    DISPLAYOUT("  -q, --quiet                   Suppress warnings; pass twice to suppress errors.\n");
+#ifndef ZSTD_NOTRACE
+    DISPLAYOUT("  --trace LOG                   Log tracing information to LOG.\n");
+#endif
+    DISPLAYOUT("\n");
+    DISPLAYOUT("  --[no-]progress               Forcibly show/hide the progress counter. NOTE: Any (de)compressed\n");
+    DISPLAYOUT("                                output to terminal will mix with progress counter text.\n\n");
 
 #ifdef UTIL_HAS_CREATEFILELIST
-    DISPLAYOUT( " -r     : operate recursively on directories \n");
-    DISPLAYOUT( "--filelist FILE : read list of files to operate upon from FILE \n");
-    DISPLAYOUT( "--output-dir-flat DIR : processed files are stored into DIR \n");
+    DISPLAYOUT("  -r                            Operate recursively on directories.\n");
+    DISPLAYOUT("  --filelist LIST               Read a list of files to operate on from LIST.\n");
+    DISPLAYOUT("  --output-dir-flat DIR         Store processed files in DIR.\n");
 #endif
 
 #ifdef UTIL_HAS_MIRRORFILELIST
-    DISPLAYOUT( "--output-dir-mirror DIR : processed files are stored into DIR respecting original directory structure \n");
+    DISPLAYOUT("  --output-dir-mirror DIR       Store processed files in DIR, respecting original directory structure.\n");
 #endif
+    if (AIO_supported())
+        DISPLAYOUT("  --[no-]asyncio                Use asynchronous IO. [Default: Enabled]\n");
 
-
+    DISPLAYOUT("\n");
 #ifndef ZSTD_NOCOMPRESS
-    DISPLAYOUT( "--[no-]check : during compression, add XXH64 integrity checksum to frame (default: enabled)");
+    DISPLAYOUT("  --[no-]check                  Add XXH64 integrity checksums during compression. [Default: Add, Validate]\n");
 #ifndef ZSTD_NODECOMPRESS
-    DISPLAYOUT( ". If specified with -d, decompressor will ignore/validate checksums in compressed frame (default: validate).");
+    DISPLAYOUT("                                If `-d` is present, ignore/validate checksums during decompression.\n");
 #endif
 #else
 #ifdef ZSTD_NOCOMPRESS
-    DISPLAYOUT( "--[no-]check : during decompression, ignore/validate checksums in compressed frame (default: validate).");
+    DISPLAYOUT("  --[no-]check                  Ignore/validate checksums during decompression. [Default: Validate]");
 #endif
 #endif /* ZSTD_NOCOMPRESS */
 
-#ifndef ZSTD_NOTRACE
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "--trace FILE : log tracing information to FILE.");
-#endif
-    DISPLAYOUT( "\n");
-
-    DISPLAYOUT( "--      : All arguments after \"--\" are treated as files \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("  --                            Treat remaining arguments after `--` as files.\n");
 
 #ifndef ZSTD_NOCOMPRESS
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "Advanced compression arguments : \n");
-    DISPLAYOUT( "--ultra : enable levels beyond %i, up to %i (requires more memory) \n", ZSTDCLI_CLEVEL_MAX, ZSTD_maxCLevel());
-    DISPLAYOUT( "--long[=#]: enable long distance matching with given window log (default: %u) \n", g_defaultMaxWindowLog);
-    DISPLAYOUT( "--fast[=#]: switch to very fast compression levels (default: %u) \n", 1);
-    DISPLAYOUT( "--adapt : dynamically adapt compression level to I/O conditions \n");
-    DISPLAYOUT( "--[no-]row-match-finder : force enable/disable usage of fast row-based matchfinder for greedy, lazy, and lazy2 strategies \n");
-    DISPLAYOUT( "--patch-from=FILE : specify the file to be used as a reference point for zstd's diff engine. \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("Advanced compression options:\n");
+    DISPLAYOUT("  --ultra                       Enable levels beyond %i, up to %i; requires more memory.\n", ZSTDCLI_CLEVEL_MAX, ZSTD_maxCLevel());
+    DISPLAYOUT("  --fast[=#]                    Use to very fast compression levels. [Default: %u]\n", 1);
+#ifdef ZSTD_GZCOMPRESS
+    if (exeNameMatch(programName, ZSTD_GZ)) {     /* behave like gzip */
+        DISPLAYOUT("  --best                        Compatibility alias for `-9`.\n");
+    }
+#endif
+    DISPLAYOUT("  --adapt                       Dynamically adapt compression level to I/O conditions.\n");
+    DISPLAYOUT("  --long[=#]                    Enable long distance matching with window log #. [Default: %u]\n", g_defaultMaxWindowLog);
+    DISPLAYOUT("  --patch-from=REF              Use REF as the reference point for Zstandard's diff engine. \n\n");
 # ifdef ZSTD_MULTITHREAD
-    DISPLAYOUT( " -T#    : spawns # compression threads (default: 1, 0==# cores) \n");
-    DISPLAYOUT( " -B#    : select size of each job (default: 0==automatic) \n");
-    DISPLAYOUT( "--single-thread : use a single thread for both I/O and compression (result slightly different than -T1) \n");
-    DISPLAYOUT( "--auto-threads={physical,logical} (default: physical} : use either physical cores or logical cores as default when specifying -T0 \n");
-    DISPLAYOUT( "--rsyncable : compress using a rsync-friendly method (-B sets block size) \n");
+    DISPLAYOUT("  -T#                           Spawn # compression threads. [Default: 1; pass 0 for core count.]\n");
+    DISPLAYOUT("  --single-thread               Share a single thread for I/O and compression (slightly different than `-T1`).\n");
+    DISPLAYOUT("  --auto-threads={physical|logical}\n");
+    DISPLAYOUT("                                Use physical/logical cores when using `-T0`. [Default: Physical]\n\n");
+    DISPLAYOUT("  -B#                           Set job size to #. [Default: 0 (automatic)]\n");
+    DISPLAYOUT("  --rsyncable                   Compress using a rsync-friendly method (`-B` sets block size). \n");
+    DISPLAYOUT("\n");
 # endif
-    DISPLAYOUT( "--exclude-compressed: only compress files that are not already compressed \n");
-    DISPLAYOUT( "--stream-size=# : specify size of streaming input from `stdin` \n");
-    DISPLAYOUT( "--size-hint=# optimize compression parameters for streaming input of approximately this size \n");
-    DISPLAYOUT( "--target-compressed-block-size=# : generate compressed block of approximately targeted size \n");
-    DISPLAYOUT( "--no-dictID : don't write dictID into header (dictionary compression only) \n");
-    DISPLAYOUT( "--[no-]compress-literals : force (un)compressed literals \n");
+    DISPLAYOUT("  --exclude-compressed          Only compress files that are not already compressed.\n\n");
+
+    DISPLAYOUT("  --stream-size=#               Specify size of streaming input from STDIN.\n");
+    DISPLAYOUT("  --size-hint=#                 Optimize compression parameters for streaming input of approximately size #.\n");
+    DISPLAYOUT("  --target-compressed-block-size=#\n");
+    DISPLAYOUT("                                Generate compressed blocks of approximately # size.\n\n");
+    DISPLAYOUT("  --no-dictID                   Don't write `dictID` into the header (dictionary compression only).\n");
+    DISPLAYOUT("  --[no-]compress-literals      Force (un)compressed literals.\n");
+    DISPLAYOUT("  --[no-]row-match-finder       Explicitly enable/disable the fast, row-based matchfinder for\n");
+    DISPLAYOUT("                                the 'greedy', 'lazy', and 'lazy2' strategies.\n");
 
-    DISPLAYOUT( "--format=zstd : compress files to the .zst format (default) \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("  --format=zstd                 Compress files to the `.zst` format. [Default]\n");
+    DISPLAYOUT("  --[no-]mmap-dict              Memory-map dictionary file rather than mallocing and loading all at once\n");
 #ifdef ZSTD_GZCOMPRESS
-    DISPLAYOUT( "--format=gzip : compress files to the .gz format \n");
+    DISPLAYOUT("  --format=gzip                 Compress files to the `.gz` format.\n");
 #endif
 #ifdef ZSTD_LZMACOMPRESS
-    DISPLAYOUT( "--format=xz : compress files to the .xz format \n");
-    DISPLAYOUT( "--format=lzma : compress files to the .lzma format \n");
+    DISPLAYOUT("  --format=xz                   Compress files to the `.xz` format.\n");
+    DISPLAYOUT("  --format=lzma                 Compress files to the `.lzma` format.\n");
 #endif
 #ifdef ZSTD_LZ4COMPRESS
-    DISPLAYOUT( "--format=lz4 : compress files to the .lz4 format \n");
+    DISPLAYOUT( "  --format=lz4                 Compress files to the `.lz4` format.\n");
 #endif
 #endif  /* !ZSTD_NOCOMPRESS */
 
 #ifndef ZSTD_NODECOMPRESS
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "Advanced decompression arguments : \n");
-    DISPLAYOUT( " -l     : print information about zstd compressed files \n");
-    DISPLAYOUT( "--test  : test compressed file integrity \n");
-    DISPLAYOUT( " -M#    : Set a memory usage limit for decompression \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("Advanced decompression options:\n");
+    DISPLAYOUT("  -l                            Print information about Zstandard-compressed files.\n");
+    DISPLAYOUT("  --test                        Test compressed file integrity.\n");
+    DISPLAYOUT("  -M#                           Set the memory usage limit to # megabytes.\n");
 # if ZSTD_SPARSE_DEFAULT
-    DISPLAYOUT( "--[no-]sparse : sparse mode (default: enabled on file, disabled on stdout) \n");
+    DISPLAYOUT("  --[no-]sparse                 Enable sparse mode. [Default: Enabled for files, disabled for STDOUT.]\n");
 # else
-    DISPLAYOUT( "--[no-]sparse : sparse mode (default: disabled) \n");
+    DISPLAYOUT("  --[no-]sparse                 Enable sparse mode. [Default: Disabled]\n");
 # endif
+    {
+        char const* passThroughDefault = "Disabled";
+        if (exeNameMatch(programName, ZSTD_CAT) ||
+            exeNameMatch(programName, ZSTD_ZCAT) ||
+            exeNameMatch(programName, ZSTD_GZCAT)) {
+            passThroughDefault = "Enabled";
+        }
+        DISPLAYOUT("  --[no-]pass-through           Pass through uncompressed files as-is. [Default: %s]\n", passThroughDefault);
+    }
 #endif  /* ZSTD_NODECOMPRESS */
 
 #ifndef ZSTD_NODICT
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "Dictionary builder : \n");
-    DISPLAYOUT( "--train ## : create a dictionary from a training set of files \n");
-    DISPLAYOUT( "--train-cover[=k=#,d=#,steps=#,split=#,shrink[=#]] : use the cover algorithm with optional args \n");
-    DISPLAYOUT( "--train-fastcover[=k=#,d=#,f=#,steps=#,split=#,accel=#,shrink[=#]] : use the fast cover algorithm with optional args \n");
-    DISPLAYOUT( "--train-legacy[=s=#] : use the legacy algorithm with selectivity (default: %u) \n", g_defaultSelectivityLevel);
-    DISPLAYOUT( " -o DICT : DICT is dictionary name (default: %s) \n", g_defaultDictName);
-    DISPLAYOUT( "--maxdict=# : limit dictionary to specified size (default: %u) \n", g_defaultMaxDictSize);
-    DISPLAYOUT( "--dictID=# : force dictionary ID to specified value (default: random) \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("Dictionary builder:\n");
+    DISPLAYOUT("  --train                       Create a dictionary from a training set of files.\n\n");
+    DISPLAYOUT("  --train-cover[=k=#,d=#,steps=#,split=#,shrink[=#]]\n");
+    DISPLAYOUT("                                Use the cover algorithm (with optional arguments).\n");
+    DISPLAYOUT("  --train-fastcover[=k=#,d=#,f=#,steps=#,split=#,accel=#,shrink[=#]]\n");
+    DISPLAYOUT("                                Use the fast cover algorithm (with optional arguments).\n\n");
+    DISPLAYOUT("  --train-legacy[=s=#]          Use the legacy algorithm with selectivity #. [Default: %u]\n", g_defaultSelectivityLevel);
+    DISPLAYOUT("  -o NAME                       Use NAME as dictionary name. [Default: %s]\n", g_defaultDictName);
+    DISPLAYOUT("  --maxdict=#                   Limit dictionary to specified size #. [Default: %u]\n", g_defaultMaxDictSize);
+    DISPLAYOUT("  --dictID=#                    Force dictionary ID to #. [Default: Random]\n");
 #endif
 
 #ifndef ZSTD_NOBENCH
-    DISPLAYOUT( "\n");
-    DISPLAYOUT( "Benchmark arguments : \n");
-    DISPLAYOUT( " -b#    : benchmark file(s), using # compression level (default: %d) \n", ZSTDCLI_CLEVEL_DEFAULT);
-    DISPLAYOUT( " -e#    : test all compression levels successively from -b# to -e# (default: 1) \n");
-    DISPLAYOUT( " -i#    : minimum evaluation time in seconds (default: 3s) \n");
-    DISPLAYOUT( " -B#    : cut file into independent blocks of size # (default: no block) \n");
-    DISPLAYOUT( " -S     : output one benchmark result per input file (default: consolidated result) \n");
-    DISPLAYOUT( "--priority=rt : set process priority to real-time \n");
+    DISPLAYOUT("\n");
+    DISPLAYOUT("Benchmark options:\n");
+    DISPLAYOUT("  -b#                           Perform benchmarking with compression level #. [Default: %d]\n", ZSTDCLI_CLEVEL_DEFAULT);
+    DISPLAYOUT("  -e#                           Test all compression levels up to #; starting level is `-b#`. [Default: 1]\n");
+    DISPLAYOUT("  -i#                           Set the minimum evaluation to time # seconds. [Default: 3]\n");
+    DISPLAYOUT("  -B#                           Cut file into independent chunks of size #. [Default: No chunking]\n");
+    DISPLAYOUT("  -S                            Output one benchmark result per input file. [Default: Consolidated result]\n");
+    DISPLAYOUT("  -D dictionary                 Benchmark using dictionary \n");
+    DISPLAYOUT("  --priority=rt                 Set process priority to real-time.\n");
 #endif
 
 }
 
-static void badusage(const char* programName)
+static void badUsage(const char* programName, const char* parameter)
 {
-    DISPLAYLEVEL(1, "Incorrect parameters \n");
+    DISPLAYLEVEL(1, "Incorrect parameter: %s \n", parameter);
     if (g_displayLevel >= 2) usage(stderr, programName);
 }
 
@@ -296,18 +337,9 @@ static const char* lastNameFromPath(const char* path)
     return name;
 }
 
-/*! exeNameMatch() :
-    @return : a non-zero value if exeName matches test, excluding the extension
-   */
-static int exeNameMatch(const char* exeName, const char* test)
-{
-    return !strncmp(exeName, test, strlen(test)) &&
-        (exeName[strlen(test)] == '\0' || exeName[strlen(test)] == '.');
-}
-
 static void errorOut(const char* msg)
 {
-    DISPLAY("%s \n", msg); exit(1);
+    DISPLAYLEVEL(1, "%s \n", msg); exit(1);
 }
 
 /*! readU32FromCharChecked() :
@@ -556,7 +588,7 @@ static ZDICT_fastCover_params_t defaultFastCoverParams(void)
 
 
 /** parseAdaptParameters() :
- *  reads adapt parameters from *stringPtr (e.g. "--zstd=min=1,max=19) and store them into adaptMinPtr and adaptMaxPtr.
+ *  reads adapt parameters from *stringPtr (e.g. "--adapt=min=1,max=19) and store them into adaptMinPtr and adaptMaxPtr.
  *  Both adaptMinPtr and adaptMaxPtr must be already allocated and correctly initialized.
  *  There is no guarantee that any of these values will be updated.
  *  @return 1 means that parsing was successful,
@@ -603,12 +635,26 @@ static unsigned parseCompressionParameters(const char* stringPtr, ZSTD_compressi
         return 0;
     }
 
-    DISPLAYLEVEL(4, "windowLog=%d, chainLog=%d, hashLog=%d, searchLog=%d \n", params->windowLog, params->chainLog, params->hashLog, params->searchLog);
-    DISPLAYLEVEL(4, "minMatch=%d, targetLength=%d, strategy=%d \n", params->minMatch, params->targetLength, params->strategy);
     if (stringPtr[0] != 0) return 0; /* check the end of string */
     return 1;
 }
 
+static void setMaxCompression(ZSTD_compressionParameters* params)
+{
+    params->windowLog = ZSTD_WINDOWLOG_MAX;
+    params->chainLog = ZSTD_CHAINLOG_MAX;
+    params->hashLog = ZSTD_HASHLOG_MAX;
+    params->searchLog = ZSTD_SEARCHLOG_MAX;
+    params->minMatch = ZSTD_MINMATCH_MIN;
+    params->targetLength = ZSTD_TARGETLENGTH_MAX;
+    params->strategy = ZSTD_STRATEGY_MAX;
+    g_overlapLog = ZSTD_OVERLAPLOG_MAX;
+    g_ldmHashLog = ZSTD_LDM_HASHLOG_MAX;
+    g_ldmHashRateLog = 0; /* automatically derived */
+    g_ldmMinMatch = 16; /* heuristic */
+    g_ldmBucketSizeLog = ZSTD_LDM_BUCKETSIZELOG_MAX;
+}
+
 static void printVersion(void)
 {
     if (g_displayLevel < DISPLAY_LEVEL_DEFAULT) {
@@ -634,6 +680,11 @@ static void printVersion(void)
     #endif
         DISPLAYOUT("\n");
         if (g_displayLevel >= 4) {
+            /* library versions */
+            DISPLAYOUT("zlib version %s\n", FIO_zlibVersion());
+            DISPLAYOUT("lz4 version %s\n", FIO_lz4Version());
+            DISPLAYOUT("lzma version %s\n", FIO_lzmaVersion());
+
             /* posix support */
         #ifdef _POSIX_C_SOURCE
             DISPLAYOUT("_POSIX_C_SOURCE defined: %ldL\n", (long) _POSIX_C_SOURCE);
@@ -658,7 +709,7 @@ static void printDefaultCParams(const char* filename, const char* dictFileName,
     unsigned long long fileSize = UTIL_getFileSize(filename);
     const size_t dictSize = dictFileName != NULL ? (size_t)UTIL_getFileSize(dictFileName) : 0;
     const ZSTD_compressionParameters cParams = ZSTD_getCParams(cLevel, fileSize, dictSize);
-    if (fileSize != UTIL_FILESIZE_UNKNOWN) DISPLAY("%s (%u bytes)\n", filename, (unsigned)fileSize);
+    if (fileSize != UTIL_FILESIZE_UNKNOWN) DISPLAY("%s (%llu bytes)\n", filename, fileSize);
     else DISPLAY("%s (src size unknown)\n", filename);
     DISPLAY(" - windowLog     : %u\n", cParams.windowLog);
     DISPLAY(" - chainLog      : %u\n", cParams.chainLog);
@@ -722,7 +773,7 @@ static int init_cLevel(void) {
 }
 
 #ifdef ZSTD_MULTITHREAD
-static unsigned init_nbThreads(void) {
+static unsigned default_nbThreads(void) {
     const char* const env = getenv(ENV_NBTHREADS);
     if (env != NULL) {
         const char* ptr = env;
@@ -749,20 +800,32 @@ static unsigned init_nbThreads(void) {
     } else {                      \
         argNb++;                  \
         if (argNb >= argCount) {  \
-            DISPLAY("error: missing command argument \n"); \
+            DISPLAYLEVEL(1, "error: missing command argument \n"); \
             CLEAN_RETURN(1);      \
         }                         \
         ptr = argv[argNb];        \
         assert(ptr != NULL);      \
         if (ptr[0]=='-') {        \
-            DISPLAY("error: command cannot be separated from its argument by another command \n"); \
+            DISPLAYLEVEL(1, "error: command cannot be separated from its argument by another command \n"); \
             CLEAN_RETURN(1);      \
 }   }   }
 
-#define NEXT_UINT32(val32) {      \
-    const char* __nb;             \
-    NEXT_FIELD(__nb);             \
+#define NEXT_UINT32(val32) {        \
+    const char* __nb;               \
+    NEXT_FIELD(__nb);               \
     val32 = readU32FromChar(&__nb); \
+    if(*__nb != 0) {                \
+        errorOut("error: only numeric values with optional suffixes K, KB, KiB, M, MB, MiB are allowed"); \
+    }                               \
+}
+
+#define NEXT_TSIZE(valTsize) {           \
+    const char* __nb;                    \
+    NEXT_FIELD(__nb);                    \
+    valTsize = readSizeTFromChar(&__nb); \
+    if(*__nb != 0) {                     \
+        errorOut("error: only numeric values with optional suffixes K, KB, KiB, M, MB, MiB are allowed"); \
+    }                                    \
 }
 
 typedef enum { zom_compress, zom_decompress, zom_test, zom_bench, zom_train, zom_list } zstd_operation_mode;
@@ -788,9 +851,7 @@ int main(int argCount, const char* argv[])
         hasStdout = 0,
         ldmFlag = 0,
         main_pause = 0,
-        nbWorkers = 0,
         adapt = 0,
-        useRowMatchFinder = 0,
         adaptMin = MINCLEVEL,
         adaptMax = MAXCLEVEL,
         rsyncable = 0,
@@ -799,18 +860,22 @@ int main(int argCount, const char* argv[])
         separateFiles = 0,
         setRealTimePrio = 0,
         singleThread = 0,
-#ifdef ZSTD_MULTITHREAD
         defaultLogicalCores = 0,
-#endif
         showDefaultCParams = 0,
         ultra=0,
-        contentSize=1;
-    double compressibility = 0.5;
+        contentSize=1,
+        removeSrcFile=0;
+    ZSTD_ParamSwitch_e mmapDict=ZSTD_ps_auto;
+    ZSTD_ParamSwitch_e useRowMatchFinder = ZSTD_ps_auto;
+    FIO_compressionType_t cType = FIO_zstdCompression;
+    int nbWorkers = -1; /* -1 means unset */
+    double compressibility = -1.0;  /* lorem ipsum generator */
     unsigned bench_nbSeconds = 3;   /* would be better if this value was synchronized from bench */
     size_t blockSize = 0;
 
     FIO_prefs_t* const prefs = FIO_createPreferences();
     FIO_ctx_t* const fCtx = FIO_createContext();
+    FIO_progressSetting_e progress = FIO_ps_auto;
     zstd_operation_mode operation = zom_compress;
     ZSTD_compressionParameters compressionParams;
     int cLevel = init_cLevel();
@@ -831,6 +896,7 @@ int main(int argCount, const char* argv[])
     size_t streamSrcSize = 0;
     size_t targetCBlockSize = 0;
     size_t srcSizeHint = 0;
+    size_t nbInputFileNames = 0;
     int dictCLevel = g_defaultDictCLevel;
     unsigned dictSelect = g_defaultSelectivityLevel;
 #ifndef ZSTD_NODICT
@@ -841,34 +907,33 @@ int main(int argCount, const char* argv[])
 #ifndef ZSTD_NOBENCH
     BMK_advancedParams_t benchParams = BMK_initAdvancedParams();
 #endif
-    ZSTD_paramSwitch_e literalCompressionMode = ZSTD_ps_auto;
-
+    ZSTD_ParamSwitch_e literalCompressionMode = ZSTD_ps_auto;
 
     /* init */
     checkLibVersion();
     (void)recursive; (void)cLevelLast;    /* not used when ZSTD_NOBENCH set */
     (void)memLimit;
     assert(argCount >= 1);
-    if ((filenames==NULL) || (file_of_names==NULL)) { DISPLAY("zstd: allocation error \n"); exit(1); }
+    if ((filenames==NULL) || (file_of_names==NULL)) { DISPLAYLEVEL(1, "zstd: allocation error \n"); exit(1); }
     programName = lastNameFromPath(programName);
-#ifdef ZSTD_MULTITHREAD
-    nbWorkers = init_nbThreads();
-#endif
 
     /* preset behaviors */
     if (exeNameMatch(programName, ZSTD_ZSTDMT)) nbWorkers=0, singleThread=0;
     if (exeNameMatch(programName, ZSTD_UNZSTD)) operation=zom_decompress;
-    if (exeNameMatch(programName, ZSTD_CAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; outFileName=stdoutmark; g_displayLevel=1; }     /* supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_ZCAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; outFileName=stdoutmark; g_displayLevel=1; }    /* behave like zcat, also supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_GZ)) { suffix = GZ_EXTENSION; FIO_setCompressionType(prefs, FIO_gzipCompression); FIO_setRemoveSrcFile(prefs, 1); }        /* behave like gzip */
-    if (exeNameMatch(programName, ZSTD_GUNZIP)) { operation=zom_decompress; FIO_setRemoveSrcFile(prefs, 1); }                                                     /* behave like gunzip, also supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_GZCAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; outFileName=stdoutmark; g_displayLevel=1; }   /* behave like gzcat, also supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_LZMA)) { suffix = LZMA_EXTENSION; FIO_setCompressionType(prefs, FIO_lzmaCompression); FIO_setRemoveSrcFile(prefs, 1); }    /* behave like lzma */
-    if (exeNameMatch(programName, ZSTD_UNLZMA)) { operation=zom_decompress; FIO_setCompressionType(prefs, FIO_lzmaCompression); FIO_setRemoveSrcFile(prefs, 1); } /* behave like unlzma, also supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_XZ)) { suffix = XZ_EXTENSION; FIO_setCompressionType(prefs, FIO_xzCompression); FIO_setRemoveSrcFile(prefs, 1); }          /* behave like xz */
-    if (exeNameMatch(programName, ZSTD_UNXZ)) { operation=zom_decompress; FIO_setCompressionType(prefs, FIO_xzCompression); FIO_setRemoveSrcFile(prefs, 1); }     /* behave like unxz, also supports multiple formats */
-    if (exeNameMatch(programName, ZSTD_LZ4)) { suffix = LZ4_EXTENSION; FIO_setCompressionType(prefs, FIO_lz4Compression); }                                       /* behave like lz4 */
-    if (exeNameMatch(programName, ZSTD_UNLZ4)) { operation=zom_decompress; FIO_setCompressionType(prefs, FIO_lz4Compression); }                                   /* behave like unlz4, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_CAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; FIO_setPassThroughFlag(prefs, 1); outFileName=stdoutmark; g_displayLevel=1; }     /* supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_ZCAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; FIO_setPassThroughFlag(prefs, 1); outFileName=stdoutmark; g_displayLevel=1; }    /* behave like zcat, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_GZ)) {   /* behave like gzip */
+        suffix = GZ_EXTENSION; cType = FIO_gzipCompression; removeSrcFile=1;
+        dictCLevel = cLevel = 6;  /* gzip default is -6 */
+    }
+    if (exeNameMatch(programName, ZSTD_GUNZIP)) { operation=zom_decompress; removeSrcFile=1; }                                                     /* behave like gunzip, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_GZCAT)) { operation=zom_decompress; FIO_overwriteMode(prefs); forceStdout=1; followLinks=1; FIO_setPassThroughFlag(prefs, 1); outFileName=stdoutmark; g_displayLevel=1; }   /* behave like gzcat, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_LZMA)) { suffix = LZMA_EXTENSION; cType = FIO_lzmaCompression; removeSrcFile=1; }    /* behave like lzma */
+    if (exeNameMatch(programName, ZSTD_UNLZMA)) { operation=zom_decompress; cType = FIO_lzmaCompression; removeSrcFile=1; } /* behave like unlzma, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_XZ)) { suffix = XZ_EXTENSION; cType = FIO_xzCompression; removeSrcFile=1; }          /* behave like xz */
+    if (exeNameMatch(programName, ZSTD_UNXZ)) { operation=zom_decompress; cType = FIO_xzCompression; removeSrcFile=1; }     /* behave like unxz, also supports multiple formats */
+    if (exeNameMatch(programName, ZSTD_LZ4)) { suffix = LZ4_EXTENSION; cType = FIO_lz4Compression; }                        /* behave like lz4 */
+    if (exeNameMatch(programName, ZSTD_UNLZ4)) { operation=zom_decompress; cType = FIO_lz4Compression; }                    /* behave like unlz4, also supports multiple formats */
     memset(&compressionParams, 0, sizeof(compressionParams));
 
     /* init crash handler */
@@ -877,6 +942,7 @@ int main(int argCount, const char* argv[])
     /* command switches */
     for (argNb=1; argNb<argCount; argNb++) {
         const char* argument = argv[argNb];
+        const char* const originalArgument = argument;
         if (!argument) continue;   /* Protection if argument empty */
 
         if (nextArgumentsAreFiles) {
@@ -902,46 +968,70 @@ int main(int argCount, const char* argv[])
                 if (!strcmp(argument, "--uncompress")) { operation=zom_decompress; continue; }
                 if (!strcmp(argument, "--force")) { FIO_overwriteMode(prefs); forceStdin=1; forceStdout=1; followLinks=1; allowBlockDevices=1; continue; }
                 if (!strcmp(argument, "--version")) { printVersion(); CLEAN_RETURN(0); }
-                if (!strcmp(argument, "--help")) { usage_advanced(programName); CLEAN_RETURN(0); }
+                if (!strcmp(argument, "--help")) { usageAdvanced(programName); CLEAN_RETURN(0); }
                 if (!strcmp(argument, "--verbose")) { g_displayLevel++; continue; }
                 if (!strcmp(argument, "--quiet")) { g_displayLevel--; continue; }
-                if (!strcmp(argument, "--stdout")) { forceStdout=1; outFileName=stdoutmark; g_displayLevel-=(g_displayLevel==2); continue; }
+                if (!strcmp(argument, "--stdout")) { forceStdout=1; outFileName=stdoutmark; continue; }
                 if (!strcmp(argument, "--ultra")) { ultra=1; continue; }
                 if (!strcmp(argument, "--check")) { FIO_setChecksumFlag(prefs, 2); continue; }
                 if (!strcmp(argument, "--no-check")) { FIO_setChecksumFlag(prefs, 0); continue; }
                 if (!strcmp(argument, "--sparse")) { FIO_setSparseWrite(prefs, 2); continue; }
                 if (!strcmp(argument, "--no-sparse")) { FIO_setSparseWrite(prefs, 0); continue; }
+                if (!strcmp(argument, "--pass-through")) { FIO_setPassThroughFlag(prefs, 1); continue; }
+                if (!strcmp(argument, "--no-pass-through")) { FIO_setPassThroughFlag(prefs, 0); continue; }
                 if (!strcmp(argument, "--test")) { operation=zom_test; continue; }
+                if (!strcmp(argument, "--asyncio")) { FIO_setAsyncIOFlag(prefs, 1); continue;}
+                if (!strcmp(argument, "--no-asyncio")) { FIO_setAsyncIOFlag(prefs, 0); continue;}
                 if (!strcmp(argument, "--train")) { operation=zom_train; if (outFileName==NULL) outFileName=g_defaultDictName; continue; }
                 if (!strcmp(argument, "--no-dictID")) { FIO_setDictIDFlag(prefs, 0); continue; }
-                if (!strcmp(argument, "--keep")) { FIO_setRemoveSrcFile(prefs, 0); continue; }
-                if (!strcmp(argument, "--rm")) { FIO_setRemoveSrcFile(prefs, 1); continue; }
+                if (!strcmp(argument, "--keep")) { removeSrcFile=0; continue; }
+                if (!strcmp(argument, "--rm")) { removeSrcFile=1; continue; }
                 if (!strcmp(argument, "--priority=rt")) { setRealTimePrio = 1; continue; }
                 if (!strcmp(argument, "--show-default-cparams")) { showDefaultCParams = 1; continue; }
                 if (!strcmp(argument, "--content-size")) { contentSize = 1; continue; }
                 if (!strcmp(argument, "--no-content-size")) { contentSize = 0; continue; }
                 if (!strcmp(argument, "--adapt")) { adapt = 1; continue; }
-                if (!strcmp(argument, "--no-row-match-finder")) { useRowMatchFinder = 1; continue; }
-                if (!strcmp(argument, "--row-match-finder")) { useRowMatchFinder = 2; continue; }
-                if (longCommandWArg(&argument, "--adapt=")) { adapt = 1; if (!parseAdaptParameters(argument, &adaptMin, &adaptMax)) { badusage(programName); CLEAN_RETURN(1); } continue; }
+                if (!strcmp(argument, "--no-row-match-finder")) { useRowMatchFinder = ZSTD_ps_disable; continue; }
+                if (!strcmp(argument, "--row-match-finder")) { useRowMatchFinder = ZSTD_ps_enable; continue; }
+                if (longCommandWArg(&argument, "--adapt=")) { adapt = 1; if (!parseAdaptParameters(argument, &adaptMin, &adaptMax)) { badUsage(programName, originalArgument); CLEAN_RETURN(1); } continue; }
                 if (!strcmp(argument, "--single-thread")) { nbWorkers = 0; singleThread = 1; continue; }
-                if (!strcmp(argument, "--format=zstd")) { suffix = ZSTD_EXTENSION; FIO_setCompressionType(prefs, FIO_zstdCompression); continue; }
+                if (!strcmp(argument, "--format=zstd")) { suffix = ZSTD_EXTENSION; cType = FIO_zstdCompression; continue; }
+                if (!strcmp(argument, "--mmap-dict")) { mmapDict = ZSTD_ps_enable; continue; }
+                if (!strcmp(argument, "--no-mmap-dict")) { mmapDict = ZSTD_ps_disable; continue; }
 #ifdef ZSTD_GZCOMPRESS
-                if (!strcmp(argument, "--format=gzip")) { suffix = GZ_EXTENSION; FIO_setCompressionType(prefs, FIO_gzipCompression); continue; }
+                if (!strcmp(argument, "--format=gzip")) { suffix = GZ_EXTENSION; cType = FIO_gzipCompression; continue; }
+                if (exeNameMatch(programName, ZSTD_GZ)) {     /* behave like gzip */
+                    if (!strcmp(argument, "--best")) { dictCLevel = cLevel = 9; continue; }
+                    if (!strcmp(argument, "--no-name")) { /* ignore for now */; continue; }
+                }
 #endif
 #ifdef ZSTD_LZMACOMPRESS
-                if (!strcmp(argument, "--format=lzma")) { suffix = LZMA_EXTENSION; FIO_setCompressionType(prefs, FIO_lzmaCompression);  continue; }
-                if (!strcmp(argument, "--format=xz")) { suffix = XZ_EXTENSION; FIO_setCompressionType(prefs, FIO_xzCompression);  continue; }
+                if (!strcmp(argument, "--format=lzma")) { suffix = LZMA_EXTENSION; cType = FIO_lzmaCompression; continue; }
+                if (!strcmp(argument, "--format=xz")) { suffix = XZ_EXTENSION; cType = FIO_xzCompression; continue; }
 #endif
 #ifdef ZSTD_LZ4COMPRESS
-                if (!strcmp(argument, "--format=lz4")) { suffix = LZ4_EXTENSION; FIO_setCompressionType(prefs, FIO_lz4Compression);  continue; }
+                if (!strcmp(argument, "--format=lz4")) { suffix = LZ4_EXTENSION; cType = FIO_lz4Compression; continue; }
 #endif
                 if (!strcmp(argument, "--rsyncable")) { rsyncable = 1; continue; }
                 if (!strcmp(argument, "--compress-literals")) { literalCompressionMode = ZSTD_ps_enable; continue; }
                 if (!strcmp(argument, "--no-compress-literals")) { literalCompressionMode = ZSTD_ps_disable; continue; }
-                if (!strcmp(argument, "--no-progress")) { FIO_setProgressSetting(FIO_ps_never); continue; }
-                if (!strcmp(argument, "--progress")) { FIO_setProgressSetting(FIO_ps_always); continue; }
+                if (!strcmp(argument, "--no-progress")) { progress = FIO_ps_never; continue; }
+                if (!strcmp(argument, "--progress")) { progress = FIO_ps_always; continue; }
                 if (!strcmp(argument, "--exclude-compressed")) { FIO_setExcludeCompressedFile(prefs, 1); continue; }
+                if (!strcmp(argument, "--fake-stdin-is-console")) { UTIL_fakeStdinIsConsole(); continue; }
+                if (!strcmp(argument, "--fake-stdout-is-console")) { UTIL_fakeStdoutIsConsole(); continue; }
+                if (!strcmp(argument, "--fake-stderr-is-console")) { UTIL_fakeStderrIsConsole(); continue; }
+                if (!strcmp(argument, "--trace-file-stat")) { UTIL_traceFileStat(); continue; }
+
+                if (!strcmp(argument, "--max")) {
+                    if (sizeof(void*)==4) {
+                        DISPLAYLEVEL(2, "--max is incompatible with 32-bit mode \n");
+                        badUsage(programName, originalArgument);
+                        CLEAN_RETURN(1);
+                    }
+                    ultra=1; ldmFlag = 1; setMaxCompression(&compressionParams);
+                    continue;
+                }
 
                 /* long commands with arguments */
 #ifndef ZSTD_NODICT
@@ -952,8 +1042,8 @@ int main(int argCount, const char* argv[])
                   dict = cover;
                   /* Allow optional arguments following an = */
                   if (*argument == 0) { memset(&coverParams, 0, sizeof(coverParams)); }
-                  else if (*argument++ != '=') { badusage(programName); CLEAN_RETURN(1); }
-                  else if (!parseCoverParameters(argument, &coverParams)) { badusage(programName); CLEAN_RETURN(1); }
+                  else if (*argument++ != '=') { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
+                  else if (!parseCoverParameters(argument, &coverParams)) { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
                   continue;
                 }
                 if (longCommandWArg(&argument, "--train-fastcover")) {
@@ -963,8 +1053,8 @@ int main(int argCount, const char* argv[])
                   dict = fastCover;
                   /* Allow optional arguments following an = */
                   if (*argument == 0) { memset(&fastCoverParams, 0, sizeof(fastCoverParams)); }
-                  else if (*argument++ != '=') { badusage(programName); CLEAN_RETURN(1); }
-                  else if (!parseFastCoverParameters(argument, &fastCoverParams)) { badusage(programName); CLEAN_RETURN(1); }
+                  else if (*argument++ != '=') { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
+                  else if (!parseFastCoverParameters(argument, &fastCoverParams)) { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
                   continue;
                 }
                 if (longCommandWArg(&argument, "--train-legacy")) {
@@ -974,8 +1064,8 @@ int main(int argCount, const char* argv[])
                   dict = legacy;
                   /* Allow optional arguments following an = */
                   if (*argument == 0) { continue; }
-                  else if (*argument++ != '=') { badusage(programName); CLEAN_RETURN(1); }
-                  else if (!parseLegacyParameters(argument, &dictSelect)) { badusage(programName); CLEAN_RETURN(1); }
+                  else if (*argument++ != '=') { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
+                  else if (!parseLegacyParameters(argument, &dictSelect)) { badUsage(programName, originalArgument); CLEAN_RETURN(1); }
                   continue;
                 }
 #endif
@@ -983,15 +1073,21 @@ int main(int argCount, const char* argv[])
                 if (longCommandWArg(&argument, "--memlimit")) { NEXT_UINT32(memLimit); continue; }
                 if (longCommandWArg(&argument, "--memory")) { NEXT_UINT32(memLimit); continue; }
                 if (longCommandWArg(&argument, "--memlimit-decompress")) { NEXT_UINT32(memLimit); continue; }
-                if (longCommandWArg(&argument, "--block-size=")) { blockSize = readSizeTFromChar(&argument); continue; }
+                if (longCommandWArg(&argument, "--block-size")) { NEXT_TSIZE(blockSize); continue; }
                 if (longCommandWArg(&argument, "--maxdict")) { NEXT_UINT32(maxDictSize); continue; }
                 if (longCommandWArg(&argument, "--dictID")) { NEXT_UINT32(dictID); continue; }
-                if (longCommandWArg(&argument, "--zstd=")) { if (!parseCompressionParameters(argument, &compressionParams)) { badusage(programName); CLEAN_RETURN(1); } continue; }
-                if (longCommandWArg(&argument, "--stream-size=")) { streamSrcSize = readSizeTFromChar(&argument); continue; }
-                if (longCommandWArg(&argument, "--target-compressed-block-size=")) { targetCBlockSize = readSizeTFromChar(&argument); continue; }
-                if (longCommandWArg(&argument, "--size-hint=")) { srcSizeHint = readSizeTFromChar(&argument); continue; }
-                if (longCommandWArg(&argument, "--output-dir-flat")) { NEXT_FIELD(outDirName); continue; }
-#ifdef ZSTD_MULTITHREAD
+                if (longCommandWArg(&argument, "--zstd=")) { if (!parseCompressionParameters(argument, &compressionParams)) { badUsage(programName, originalArgument); CLEAN_RETURN(1); } ; cType = FIO_zstdCompression; continue; }
+                if (longCommandWArg(&argument, "--stream-size")) { NEXT_TSIZE(streamSrcSize); continue; }
+                if (longCommandWArg(&argument, "--target-compressed-block-size")) { NEXT_TSIZE(targetCBlockSize); continue; }
+                if (longCommandWArg(&argument, "--size-hint")) { NEXT_TSIZE(srcSizeHint); continue; }
+                if (longCommandWArg(&argument, "--output-dir-flat")) {
+                    NEXT_FIELD(outDirName);
+                    if (strlen(outDirName) == 0) {
+                        DISPLAYLEVEL(1, "error: output dir cannot be empty string (did you mean to pass '.' instead?)\n");
+                        CLEAN_RETURN(1);
+                    }
+                    continue;
+                }
                 if (longCommandWArg(&argument, "--auto-threads")) {
                     const char* threadDefault = NULL;
                     NEXT_FIELD(threadDefault);
@@ -999,25 +1095,34 @@ int main(int argCount, const char* argv[])
                         defaultLogicalCores = 1;
                     continue;
                 }
-#endif
 #ifdef UTIL_HAS_MIRRORFILELIST
-                if (longCommandWArg(&argument, "--output-dir-mirror")) { NEXT_FIELD(outMirroredDirName); continue; }
+                if (longCommandWArg(&argument, "--output-dir-mirror")) {
+                    NEXT_FIELD(outMirroredDirName);
+                    if (strlen(outMirroredDirName) == 0) {
+                        DISPLAYLEVEL(1, "error: output dir cannot be empty string (did you mean to pass '.' instead?)\n");
+                        CLEAN_RETURN(1);
+                    }
+                    continue;
+                }
 #endif
 #ifndef ZSTD_NOTRACE
                 if (longCommandWArg(&argument, "--trace")) { char const* traceFile; NEXT_FIELD(traceFile); TRACE_enable(traceFile); continue; }
 #endif
-                if (longCommandWArg(&argument, "--patch-from")) { NEXT_FIELD(patchFromDictFileName); continue; }
+                if (longCommandWArg(&argument, "--patch-from")) { NEXT_FIELD(patchFromDictFileName); ultra = 1; continue; }
                 if (longCommandWArg(&argument, "--long")) {
                     unsigned ldmWindowLog = 0;
                     ldmFlag = 1;
+                    ultra = 1;
                     /* Parse optional window log */
                     if (*argument == '=') {
                         ++argument;
                         ldmWindowLog = readU32FromChar(&argument);
                     } else if (*argument != 0) {
                         /* Invalid character following --long */
-                        badusage(programName);
+                        badUsage(programName, originalArgument);
                         CLEAN_RETURN(1);
+                    } else {
+                        ldmWindowLog = g_defaultMaxWindowLog;
                     }
                     /* Only set windowLog if not already set by --zstd */
                     if (compressionParams.windowLog == 0)
@@ -1036,12 +1141,12 @@ int main(int argCount, const char* argv[])
                         if (fastLevel) {
                             dictCLevel = cLevel = -(int)fastLevel;
                         } else {
-                            badusage(programName);
+                            badUsage(programName, originalArgument);
                             CLEAN_RETURN(1);
                         }
                     } else if (*argument != 0) {
                         /* Invalid character following --fast */
-                        badusage(programName);
+                        badUsage(programName, originalArgument);
                         CLEAN_RETURN(1);
                     } else {
                         cLevel = -1;  /* default for --fast */
@@ -1057,7 +1162,8 @@ int main(int argCount, const char* argv[])
                     continue;
                 }
 
-                /* fall-through, will trigger bad_usage() later on */
+                badUsage(programName, originalArgument);
+                CLEAN_RETURN(1);
             }
 
             argument++;
@@ -1075,8 +1181,8 @@ int main(int argCount, const char* argv[])
                 {
                     /* Display help */
                 case 'V': printVersion(); CLEAN_RETURN(0);   /* Version Only */
-                case 'H':
-                case 'h': usage_advanced(programName); CLEAN_RETURN(0);
+                case 'H': usageAdvanced(programName); CLEAN_RETURN(0);
+                case 'h': usage(stdout, programName); CLEAN_RETURN(0);
 
                      /* Compress */
                 case 'z': operation=zom_compress; argument++; break;
@@ -1092,6 +1198,12 @@ int main(int argCount, const char* argv[])
                     /* Force stdout, even if stdout==console */
                 case 'c': forceStdout=1; outFileName=stdoutmark; argument++; break;
 
+                    /* destination file name */
+                case 'o': argument++; NEXT_FIELD(outFileName); break;
+
+                    /* do not store filename - gzip compatibility - nothing to do */
+                case 'n': argument++; break;
+
                     /* Use file content as dictionary */
                 case 'D': argument++; NEXT_FIELD(dictFileName); break;
 
@@ -1105,7 +1217,7 @@ int main(int argCount, const char* argv[])
                 case 'q': g_displayLevel--; argument++; break;
 
                     /* keep source file (default) */
-                case 'k': FIO_setRemoveSrcFile(prefs, 0); argument++; break;
+                case 'k': removeSrcFile=0; argument++; break;
 
                     /* Checksum */
                 case 'C': FIO_setChecksumFlag(prefs, 2); argument++; break;
@@ -1113,9 +1225,6 @@ int main(int argCount, const char* argv[])
                     /* test compressed file */
                 case 't': operation=zom_test; argument++; break;
 
-                    /* destination file name */
-                case 'o': argument++; NEXT_FIELD(outFileName); break;
-
                     /* limit memory */
                 case 'M':
                     argument++;
@@ -1164,7 +1273,7 @@ int main(int argCount, const char* argv[])
                     /* nb of threads (hidden option) */
                 case 'T':
                     argument++;
-                    nbWorkers = (int)readU32FromChar(&argument);
+                    nbWorkers = readU32FromChar(&argument);
                     break;
 
                     /* Dictionary Selection level */
@@ -1190,7 +1299,12 @@ int main(int argCount, const char* argv[])
                     break;
 
                     /* unknown command */
-                default : badusage(programName); CLEAN_RETURN(1);
+                default :
+                    {   char shortArgument[3] = {'-', 0, 0};
+                        shortArgument[1] = argument[0];
+                        badUsage(programName, shortArgument);
+                        CLEAN_RETURN(1);
+                    }
                 }
             }
             continue;
@@ -1204,18 +1318,31 @@ int main(int argCount, const char* argv[])
     DISPLAYLEVEL(3, WELCOME_MESSAGE);
 
 #ifdef ZSTD_MULTITHREAD
+    if ((operation==zom_decompress) && (nbWorkers > 1)) {
+        DISPLAYLEVEL(2, "Warning : decompression does not support multi-threading\n");
+    }
     if ((nbWorkers==0) && (!singleThread)) {
         /* automatically set # workers based on # of reported cpus */
         if (defaultLogicalCores) {
-            nbWorkers = UTIL_countLogicalCores();
+            nbWorkers = (unsigned)UTIL_countLogicalCores();
             DISPLAYLEVEL(3, "Note: %d logical core(s) detected \n", nbWorkers);
         } else {
-            nbWorkers = UTIL_countPhysicalCores();
+            nbWorkers = (unsigned)UTIL_countPhysicalCores();
             DISPLAYLEVEL(3, "Note: %d physical core(s) detected \n", nbWorkers);
         }
     }
+    /* Resolve to default if nbWorkers is still unset */
+    if (nbWorkers == -1) {
+      if (operation == zom_decompress) {
+        nbWorkers = 1;
+      } else {
+        nbWorkers = default_nbThreads();
+      }
+    }
+    if (operation != zom_bench)
+        DISPLAYLEVEL(4, "Compressing with %u worker threads \n", nbWorkers);
 #else
-    (void)singleThread; (void)nbWorkers;
+    (void)singleThread; (void)nbWorkers; (void)defaultLogicalCores;
 #endif
 
     g_utilDisplayLevel = g_displayLevel;
@@ -1251,6 +1378,8 @@ int main(int argCount, const char* argv[])
         }
     }
 
+    nbInputFileNames = filenames->tableSize; /* saving number of input files */
+
     if (recursive) {  /* at this stage, filenameTable is a list of paths, which can contain both files and directories */
         UTIL_expandFNT(&filenames, followLinks);
     }
@@ -1263,7 +1392,7 @@ int main(int argCount, const char* argv[])
         int const ret = FIO_listMultipleFiles((unsigned)filenames->tableSize, filenames->fileNames, g_displayLevel);
         CLEAN_RETURN(ret);
 #else
-        DISPLAY("file information is not supported \n");
+        DISPLAYLEVEL(1, "file information is not supported \n");
         CLEAN_RETURN(1);
 #endif
     }
@@ -1271,14 +1400,19 @@ int main(int argCount, const char* argv[])
     /* Check if benchmark is selected */
     if (operation==zom_bench) {
 #ifndef ZSTD_NOBENCH
+        if (cType != FIO_zstdCompression) {
+            DISPLAYLEVEL(1, "benchmark mode is only compatible with zstd format \n");
+            CLEAN_RETURN(1);
+        }
         benchParams.blockSize = blockSize;
-        benchParams.nbWorkers = nbWorkers;
+        benchParams.targetCBlockSize = targetCBlockSize;
+        benchParams.nbWorkers = (int)nbWorkers;
         benchParams.realTime = (unsigned)setRealTimePrio;
         benchParams.nbSeconds = bench_nbSeconds;
         benchParams.ldmFlag = ldmFlag;
         benchParams.ldmMinMatch = (int)g_ldmMinMatch;
         benchParams.ldmHashLog = (int)g_ldmHashLog;
-        benchParams.useRowMatchFinder = useRowMatchFinder;
+        benchParams.useRowMatchFinder = (int)useRowMatchFinder;
         if (g_ldmBucketSizeLog != LDM_PARAM_DEFAULT) {
             benchParams.ldmBucketSizeLog = (int)g_ldmBucketSizeLog;
         }
@@ -1287,28 +1421,31 @@ int main(int argCount, const char* argv[])
         }
         benchParams.literalCompressionMode = literalCompressionMode;
 
+        if (benchParams.mode == BMK_decodeOnly) cLevel = cLevelLast = 0;
         if (cLevel > ZSTD_maxCLevel()) cLevel = ZSTD_maxCLevel();
         if (cLevelLast > ZSTD_maxCLevel()) cLevelLast = ZSTD_maxCLevel();
         if (cLevelLast < cLevel) cLevelLast = cLevel;
-        if (cLevelLast > cLevel)
-            DISPLAYLEVEL(3, "Benchmarking levels from %d to %d\n", cLevel, cLevelLast);
+        DISPLAYLEVEL(3, "Benchmarking ");
+        if (filenames->tableSize > 1)
+            DISPLAYLEVEL(3, "%u files ", (unsigned)filenames->tableSize);
+        if (cLevelLast > cLevel) {
+            DISPLAYLEVEL(3, "from level %d to %d ", cLevel, cLevelLast);
+        } else {
+            DISPLAYLEVEL(3, "at level %d ", cLevel);
+        }
+        DISPLAYLEVEL(3, "using %i threads \n", nbWorkers);
         if (filenames->tableSize > 0) {
             if(separateFiles) {
                 unsigned i;
                 for(i = 0; i < filenames->tableSize; i++) {
-                    int c;
-                    DISPLAYLEVEL(3, "Benchmarking %s \n", filenames->fileNames[i]);
-                    for(c = cLevel; c <= cLevelLast; c++) {
-                        BMK_benchFilesAdvanced(&filenames->fileNames[i], 1, dictFileName, c, &compressionParams, g_displayLevel, &benchParams);
-                }   }
+                    operationResult = BMK_benchFilesAdvanced(&filenames->fileNames[i], 1, dictFileName, cLevel, cLevelLast, &compressionParams, g_displayLevel, &benchParams);
+                }
             } else {
-                for(; cLevel <= cLevelLast; cLevel++) {
-                    BMK_benchFilesAdvanced(filenames->fileNames, (unsigned)filenames->tableSize, dictFileName, cLevel, &compressionParams, g_displayLevel, &benchParams);
-            }   }
+                operationResult = BMK_benchFilesAdvanced(filenames->fileNames, (unsigned)filenames->tableSize, dictFileName, cLevel, cLevelLast, &compressionParams, g_displayLevel, &benchParams);
+            }
         } else {
-            for(; cLevel <= cLevelLast; cLevel++) {
-                BMK_syntheticTest(cLevel, compressibility, &compressionParams, g_displayLevel, &benchParams);
-        }   }
+            operationResult = BMK_syntheticTest(compressibility, cLevel, cLevelLast, &compressionParams, g_displayLevel, &benchParams);
+        }
 
 #else
         (void)bench_nbSeconds; (void)blockSize; (void)setRealTimePrio; (void)separateFiles; (void)compressibility;
@@ -1349,24 +1486,34 @@ int main(int argCount, const char* argv[])
     }
 
 #ifndef ZSTD_NODECOMPRESS
-    if (operation==zom_test) { FIO_setTestMode(prefs, 1); outFileName=nulmark; FIO_setRemoveSrcFile(prefs, 0); }  /* test mode */
+    if (operation==zom_test) { FIO_setTestMode(prefs, 1); outFileName=nulmark; removeSrcFile=0; }  /* test mode */
 #endif
 
     /* No input filename ==> use stdin and stdout */
-    if (filenames->tableSize == 0) UTIL_refFilename(filenames, stdinmark);
-    if (!strcmp(filenames->fileNames[0], stdinmark) && !outFileName)
+    if (filenames->tableSize == 0) {
+      /* It is possible that the input
+       was a number of empty directories. In this case
+       stdin and stdout should not be used */
+       if (nbInputFileNames > 0 ){
+        DISPLAYLEVEL(1, "please provide correct input file(s) or non-empty directories -- ignored \n");
+        CLEAN_RETURN(0);
+       }
+       UTIL_refFilename(filenames, stdinmark);
+    }
+
+    if (filenames->tableSize == 1 && !strcmp(filenames->fileNames[0], stdinmark) && !outFileName)
         outFileName = stdoutmark;  /* when input is stdin, default output is stdout */
 
     /* Check if input/output defined as console; trigger an error in this case */
     if (!forceStdin
-     && !strcmp(filenames->fileNames[0], stdinmark)
-     && IS_CONSOLE(stdin) ) {
+     && (UTIL_searchFileNamesTable(filenames, stdinmark) != -1)
+     && UTIL_isConsole(stdin) ) {
         DISPLAYLEVEL(1, "stdin is a console, aborting\n");
         CLEAN_RETURN(1);
     }
-    if ( outFileName && !strcmp(outFileName, stdoutmark)
-      && IS_CONSOLE(stdout)
-      && !strcmp(filenames->fileNames[0], stdinmark)
+    if ( (!outFileName || !strcmp(outFileName, stdoutmark))
+      && UTIL_isConsole(stdout)
+      && (UTIL_searchFileNamesTable(filenames, stdinmark) != -1)
       && !forceStdout
       && operation!=zom_decompress ) {
         DISPLAYLEVEL(1, "stdout is a console, aborting\n");
@@ -1384,25 +1531,35 @@ int main(int argCount, const char* argv[])
 
     if (showDefaultCParams) {
         if (operation == zom_decompress) {
-            DISPLAY("error : can't use --show-default-cparams in decomrpession mode \n");
+            DISPLAYLEVEL(1, "error : can't use --show-default-cparams in decompression mode \n");
             CLEAN_RETURN(1);
         }
     }
 
     if (dictFileName != NULL && patchFromDictFileName != NULL) {
-        DISPLAY("error : can't use -D and --patch-from=# at the same time \n");
+        DISPLAYLEVEL(1, "error : can't use -D and --patch-from=# at the same time \n");
         CLEAN_RETURN(1);
     }
 
     if (patchFromDictFileName != NULL && filenames->tableSize > 1) {
-        DISPLAY("error : can't use --patch-from=# on multiple files \n");
+        DISPLAYLEVEL(1, "error : can't use --patch-from=# on multiple files \n");
         CLEAN_RETURN(1);
     }
 
-    /* No status message in pipe mode (stdin - stdout) */
+    /* No status message by default when output is stdout */
     hasStdout = outFileName && !strcmp(outFileName,stdoutmark);
+    if (hasStdout && (g_displayLevel==2)) g_displayLevel=1;
+
+    /* when stderr is not the console, do not pollute it with progress updates (unless requested) */
+    if (!UTIL_isConsole(stderr) && (progress!=FIO_ps_always)) progress=FIO_ps_never;
+    FIO_setProgressSetting(progress);
 
-    if ((hasStdout || !IS_CONSOLE(stderr)) && (g_displayLevel==2)) g_displayLevel=1;
+    /* don't remove source files when output is stdout */;
+    if (hasStdout && removeSrcFile) {
+        DISPLAYLEVEL(3, "Note: src files are not removed when output is stdout \n");
+        removeSrcFile = 0;
+    }
+    FIO_setRemoveSrcFile(prefs, removeSrcFile);
 
     /* IO Stream/File */
     FIO_setHasStdoutOutput(fCtx, hasStdout);
@@ -1411,6 +1568,7 @@ int main(int argCount, const char* argv[])
     FIO_setNotificationLevel(g_displayLevel);
     FIO_setAllowBlockDevices(prefs, allowBlockDevices);
     FIO_setPatchFromMode(prefs, patchFromDictFileName != NULL);
+    FIO_setMMapDict(prefs, mmapDict);
     if (memLimit == 0) {
         if (compressionParams.windowLog == 0) {
             memLimit = (U32)1 << g_defaultMaxWindowLog;
@@ -1422,8 +1580,9 @@ int main(int argCount, const char* argv[])
     FIO_setMemLimit(prefs, memLimit);
     if (operation==zom_compress) {
 #ifndef ZSTD_NOCOMPRESS
+        FIO_setCompressionType(prefs, cType);
         FIO_setContentSize(prefs, contentSize);
-        FIO_setNbWorkers(prefs, nbWorkers);
+        FIO_setNbWorkers(prefs, (int)nbWorkers);
         FIO_setBlockSize(prefs, (int)blockSize);
         if (g_overlapLog!=OVERLAP_LOG_DEFAULT) FIO_setOverlapLog(prefs, (int)g_overlapLog);
         FIO_setLdmFlag(prefs, (unsigned)ldmFlag);
@@ -1431,8 +1590,8 @@ int main(int argCount, const char* argv[])
         FIO_setLdmMinMatch(prefs, (int)g_ldmMinMatch);
         if (g_ldmBucketSizeLog != LDM_PARAM_DEFAULT) FIO_setLdmBucketSizeLog(prefs, (int)g_ldmBucketSizeLog);
         if (g_ldmHashRateLog != LDM_PARAM_DEFAULT) FIO_setLdmHashRateLog(prefs, (int)g_ldmHashRateLog);
-        FIO_setAdaptiveMode(prefs, (unsigned)adapt);
-        FIO_setUseRowMatchFinder(prefs, useRowMatchFinder);
+        FIO_setAdaptiveMode(prefs, adapt);
+        FIO_setUseRowMatchFinder(prefs, (int)useRowMatchFinder);
         FIO_setAdaptMin(prefs, adaptMin);
         FIO_setAdaptMax(prefs, adaptMax);
         FIO_setRsyncable(prefs, rsyncable);
@@ -1440,6 +1599,7 @@ int main(int argCount, const char* argv[])
         FIO_setTargetCBlockSize(prefs, targetCBlockSize);
         FIO_setSrcSizeHint(prefs, srcSizeHint);
         FIO_setLiteralCompressionMode(prefs, literalCompressionMode);
+        FIO_setSparseWrite(prefs, 0);
         if (adaptMin > cLevel) cLevel = adaptMin;
         if (adaptMax < cLevel) cLevel = adaptMax;
 
@@ -1465,8 +1625,12 @@ int main(int argCount, const char* argv[])
         else
             operationResult = FIO_compressMultipleFilenames(fCtx, prefs, filenames->fileNames, outMirroredDirName, outDirName, outFileName, suffix, dictFileName, cLevel, compressionParams);
 #else
-        (void)contentSize; (void)suffix; (void)adapt; (void)rsyncable; (void)ultra; (void)cLevel; (void)ldmFlag; (void)literalCompressionMode; (void)targetCBlockSize; (void)streamSrcSize; (void)srcSizeHint; (void)ZSTD_strategyMap; (void)useRowMatchFinder; /* not used when ZSTD_NOCOMPRESS set */
-        DISPLAY("Compression not supported \n");
+        /* these variables are only used when compression mode is enabled */
+        (void)contentSize; (void)suffix; (void)adapt; (void)rsyncable;
+        (void)ultra; (void)cLevel; (void)ldmFlag; (void)literalCompressionMode;
+        (void)targetCBlockSize; (void)streamSrcSize; (void)srcSizeHint;
+        (void)ZSTD_strategyMap; (void)useRowMatchFinder; (void)cType;
+        DISPLAYLEVEL(1, "Compression not supported \n");
 #endif
     } else {  /* decompression or test */
 #ifndef ZSTD_NODECOMPRESS
@@ -1476,7 +1640,7 @@ int main(int argCount, const char* argv[])
             operationResult = FIO_decompressMultipleFilenames(fCtx, prefs, filenames->fileNames, outMirroredDirName, outDirName, outFileName, dictFileName);
         }
 #else
-        DISPLAY("Decompression not supported \n");
+        DISPLAYLEVEL(1, "Decompression not supported \n");
 #endif
     }
 
diff --git a/sys/contrib/zstd/programs/zstdcli_trace.c b/sys/contrib/zstd/programs/zstdcli_trace.c
index b3b977feb53b..35075a52c4d0 100644
--- a/sys/contrib/zstd/programs/zstdcli_trace.c
+++ b/sys/contrib/zstd/programs/zstdcli_trace.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/sys/contrib/zstd/programs/zstdcli_trace.h b/sys/contrib/zstd/programs/zstdcli_trace.h
index 38c27dc04c4f..9c135d3ca848 100644
--- a/sys/contrib/zstd/programs/zstdcli_trace.h
+++ b/sys/contrib/zstd/programs/zstdcli_trace.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) Facebook, Inc.
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
diff --git a/sys/contrib/zstd/programs/zstdgrep.1 b/sys/contrib/zstd/programs/zstdgrep.1
index 563696d339ee..d7fda5833981 100644
--- a/sys/contrib/zstd/programs/zstdgrep.1
+++ b/sys/contrib/zstd/programs/zstdgrep.1
@@ -1,5 +1,5 @@
 .
-.TH "ZSTDGREP" "1" "January 2022" "zstd 1.5.2" "User Commands"
+.TH "ZSTDGREP" "1" "March 2024" "zstd 1.5.6" "User Commands"
 .
 .SH "NAME"
 \fBzstdgrep\fR \- print lines matching a pattern in zstandard\-compressed files
@@ -8,19 +8,19 @@
 \fBzstdgrep\fR [\fIgrep\-flags\fR] [\-\-] \fIpattern\fR [\fIfiles\fR \.\.\.]
 .
 .SH "DESCRIPTION"
-\fBzstdgrep\fR runs \fBgrep (1)\fR on files, or \fBstdin\fR if no files argument is given, after decompressing them with \fBzstdcat (1)\fR\.
+\fBzstdgrep\fR runs \fBgrep\fR(1) on files, or \fBstdin\fR if no files argument is given, after decompressing them with \fBzstdcat\fR(1)\.
 .
 .P
-The grep\-flags and pattern arguments are passed on to \fBgrep (1)\fR\. If an \fB\-e\fR flag is found in the \fBgrep\-flags\fR, \fBzstdgrep\fR will not look for a pattern argument\.
+The \fIgrep\-flags\fR and \fIpattern\fR arguments are passed on to \fBgrep\fR(1)\. If an \fB\-e\fR flag is found in the \fIgrep\-flags\fR, \fBzstdgrep\fR will not look for a \fIpattern\fR argument\.
 .
 .P
-Note that modern \fBgrep\fR alternatives such as \fBripgrep\fR (\fBrg\fR) support \fBzstd\fR\-compressed files out of the box, and can prove better alternatives than \fBzstdgrep\fR notably for unsupported complex pattern searches\. Note though that such alternatives may also feature some minor command line differences\.
+Note that modern \fBgrep\fR alternatives such as \fBripgrep\fR (\fBrg\fR(1)) support \fBzstd\fR\-compressed files out of the box, and can prove better alternatives than \fBzstdgrep\fR notably for unsupported complex pattern searches\. Note though that such alternatives may also feature some minor command line differences\.
 .
 .SH "EXIT STATUS"
 In case of missing arguments or missing pattern, 1 will be returned, otherwise 0\.
 .
 .SH "SEE ALSO"
-\fBzstd (1)\fR
+\fBzstd\fR(1)
 .
 .SH "AUTHORS"
 Thomas Klausner \fIwiz@NetBSD\.org\fR
diff --git a/sys/contrib/zstd/programs/zstdgrep.1.md b/sys/contrib/zstd/programs/zstdgrep.1.md
index 35186a4bf02d..6370a81c7ca5 100644
--- a/sys/contrib/zstd/programs/zstdgrep.1.md
+++ b/sys/contrib/zstd/programs/zstdgrep.1.md
@@ -4,16 +4,16 @@ zstdgrep(1) -- print lines matching a pattern in zstandard-compressed files
 SYNOPSIS
 --------
 
-`zstdgrep` [*grep-flags*] [--] _pattern_ [_files_ ...]
+`zstdgrep` [<grep-flags>] [--] <pattern> [<files> ...]
 
 
 DESCRIPTION
 -----------
-`zstdgrep` runs `grep (1)` on files, or `stdin` if no files argument is given, after decompressing them with `zstdcat (1)`.
+`zstdgrep` runs `grep`(1) on files, or `stdin` if no files argument is given, after decompressing them with `zstdcat`(1).
 
-The grep-flags and pattern arguments are passed on to `grep (1)`.  If an `-e` flag is found in the `grep-flags`, `zstdgrep` will not look for a pattern argument.
+The <grep-flags> and <pattern> arguments are passed on to `grep`(1).  If an `-e` flag is found in the <grep-flags>, `zstdgrep` will not look for a <pattern> argument.
 
-Note that modern `grep` alternatives such as `ripgrep` (`rg`) support `zstd`-compressed files out of the box,
+Note that modern `grep` alternatives such as `ripgrep` (`rg`(1)) support `zstd`-compressed files out of the box,
 and can prove better alternatives than `zstdgrep` notably for unsupported complex pattern searches.
 Note though that such alternatives may also feature some minor command line differences.
 
@@ -23,7 +23,7 @@ In case of missing arguments or missing pattern, 1 will be returned, otherwise 0
 
 SEE ALSO
 --------
-`zstd (1)`
+`zstd`(1)
 
 AUTHORS
 -------
diff --git a/sys/contrib/zstd/programs/zstdless b/sys/contrib/zstd/programs/zstdless
index 893799e7d95a..17726a4f6e1e 100755
--- a/sys/contrib/zstd/programs/zstdless
+++ b/sys/contrib/zstd/programs/zstdless
@@ -1,2 +1,8 @@
 #!/bin/sh
-zstdcat "$@" | less
+
+zstd=${ZSTD:-zstd}
+
+# TODO: Address quirks and bugs tied to old versions of less, provide a mechanism to pass flags directly to zstd
+
+export LESSOPEN="|-${zstd} -cdfq %s"
+exec less "$@"
diff --git a/sys/contrib/zstd/programs/zstdless.1 b/sys/contrib/zstd/programs/zstdless.1
index ab38e7a7f45d..7dd65f8fc44a 100644
--- a/sys/contrib/zstd/programs/zstdless.1
+++ b/sys/contrib/zstd/programs/zstdless.1
@@ -1,5 +1,5 @@
 .
-.TH "ZSTDLESS" "1" "January 2022" "zstd 1.5.2" "User Commands"
+.TH "ZSTDLESS" "1" "March 2024" "zstd 1.5.6" "User Commands"
 .
 .SH "NAME"
 \fBzstdless\fR \- view zstandard\-compressed files
@@ -8,7 +8,7 @@
 \fBzstdless\fR [\fIflags\fR] [\fIfile\fR \.\.\.]
 .
 .SH "DESCRIPTION"
-\fBzstdless\fR runs \fBless (1)\fR on files or stdin, if no files argument is given, after decompressing them with \fBzstdcat (1)\fR\.
+\fBzstdless\fR runs \fBless\fR(1) on files or stdin, if no \fIfile\fR argument is given, after decompressing them with \fBzstdcat\fR(1)\.
 .
 .SH "SEE ALSO"
-\fBzstd (1)\fR
+\fBzstd\fR(1)
diff --git a/sys/contrib/zstd/programs/zstdless.1.md b/sys/contrib/zstd/programs/zstdless.1.md
index d91d48abcc71..67c1c76769b9 100644
--- a/sys/contrib/zstd/programs/zstdless.1.md
+++ b/sys/contrib/zstd/programs/zstdless.1.md
@@ -4,13 +4,13 @@ zstdless(1) -- view zstandard-compressed files
 SYNOPSIS
 --------
 
-`zstdless` [*flags*] [_file_ ...]
+`zstdless` [<flags>] [<file> ...]
 
 
 DESCRIPTION
 -----------
-`zstdless` runs `less (1)` on files or stdin, if no files argument is given, after decompressing them with `zstdcat (1)`.
+`zstdless` runs `less`(1) on files or stdin, if no <file> argument is given, after decompressing them with `zstdcat`(1).
 
 SEE ALSO
 --------
-`zstd (1)`
+`zstd`(1)