aboutsummaryrefslogtreecommitdiff
path: root/java/java3d
diff options
context:
space:
mode:
authorMikhail Teterin <mi@FreeBSD.org>2026-10-06 04:18:46 +0000
committerGleb Popov <arrowd@FreeBSD.org>2026-10-10 13:01:14 +0000
commitc8ddf6fc600d7a0922a170d8d3ad850d2281ff0a (patch)
tree4e26ba158f9559c20bc3f474d09415429b258f5c /java/java3d
parent413b5df516808981eb9dce86e6fe32a85566ec6b (diff)
multimedia/x265: update 4.2 -> 4.3
Upstream changes: * New: LoongArch64 support; foveated encoding (--fovea-gaze, --fovea-delta, --fovea-sigma, --fovea-gaze-file), which spends fewer bits away from a gaze point; selective, noise-based MCSTF (--selective-mcstf, --mcstf-ref-range). * MCSTF reworked: multi-threaded motion estimation and bilateral filtering, AVX2 kernels (about 2x faster with threading). * More AArch64 Neon/SVE/SVE2 optimizations; early-exit diamond search in threaded ME. * Fixes for ASAN/UBSAN and TSAN findings, threaded-ME and MCSTF crashes, --pools thread counts, scenecut with --bframes 0, a bitstream error with VUI timing info disabled, and CPU-feature detection on FreeBSD (elf_aux_info). * API: the temporalFilterStrength and searchRangeForLayer[012] fields are gone from x265_param; X265_BUILD is now 217. https://github.com/Multicorewareinc/x265/blob/4.3/doc/reST/releasenotes.rst The project has moved to GitHub; distfiles are now published as release assets there. While here, pull in the fixes committed upstream since 4.3: a crash when x265 is run without arguments, uninitialized DTS fields, out-of-range mvd values in the bitstream, a motion-search offset error for small rectangular PUs, recon-file bit-depth handling in multi-lib builds, repeated x265_encoder_headers() calls accumulating NALs, stack underflows in the AArch64 and RISC-V intra filters, an AArch64 execute-only fix, __atomic builtins (fixing 32-bit builds), and removal of a 25% AArch64 slowdown caused by a mutex added in 4.3. The version string now counts these on top of the release and names the newest of them: 4.3+14-d978725. Also make x265_param_free() accept NULL. With SVT-HEVC enabled, every CLI error exit crashed: the CLI frees its parameters, clears the pointer, and then hands the NULL to x265_param_free() a second time. And build the 10- and 12-bit libraries with the same VMAF setting as the main one. ENABLE_LIBVMAF changes the layout of x265_api, so with the VMAF option on, encoding at those depths with --csv called through a NULL pointer and crashed. PR: 299194 Co-Authored-By: Claude Opus 5.5 <effort=high>
Diffstat (limited to 'java/java3d')
0 files changed, 0 insertions, 0 deletions