aboutsummaryrefslogtreecommitdiff
path: root/test/CodeGen/X86/clz.ll
diff options
context:
space:
mode:
Diffstat (limited to 'test/CodeGen/X86/clz.ll')
-rw-r--r--test/CodeGen/X86/clz.ll148
1 files changed, 84 insertions, 64 deletions
diff --git a/test/CodeGen/X86/clz.ll b/test/CodeGen/X86/clz.ll
index 6a6f5256f44d..4a094480c931 100644
--- a/test/CodeGen/X86/clz.ll
+++ b/test/CodeGen/X86/clz.ll
@@ -1,4 +1,4 @@
-; RUN: llc < %s -march=x86-64 -mcpu=yonah | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
declare i8 @llvm.cttz.i8(i8, i1)
declare i16 @llvm.cttz.i16(i16, i1)
@@ -10,131 +10,151 @@ declare i32 @llvm.ctlz.i32(i32, i1)
declare i64 @llvm.ctlz.i64(i64, i1)
define i8 @cttz_i8(i8 %x) {
+; CHECK-LABEL: cttz_i8:
+; CHECK: # BB#0:
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: bsfl %eax, %eax
+; CHECK-NEXT: retq
%tmp = call i8 @llvm.cttz.i8( i8 %x, i1 true )
ret i8 %tmp
-; CHECK-LABEL: cttz_i8:
-; CHECK: bsfl
-; CHECK-NOT: cmov
-; CHECK: ret
}
define i16 @cttz_i16(i16 %x) {
+; CHECK-LABEL: cttz_i16:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsfw %di, %ax
+; CHECK-NEXT: retq
%tmp = call i16 @llvm.cttz.i16( i16 %x, i1 true )
ret i16 %tmp
-; CHECK-LABEL: cttz_i16:
-; CHECK: bsfw
-; CHECK-NOT: cmov
-; CHECK: ret
}
define i32 @cttz_i32(i32 %x) {
+; CHECK-LABEL: cttz_i32:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsfl %edi, %eax
+; CHECK-NEXT: retq
%tmp = call i32 @llvm.cttz.i32( i32 %x, i1 true )
ret i32 %tmp
-; CHECK-LABEL: cttz_i32:
-; CHECK: bsfl
-; CHECK-NOT: cmov
-; CHECK: ret
}
define i64 @cttz_i64(i64 %x) {
+; CHECK-LABEL: cttz_i64:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsfq %rdi, %rax
+; CHECK-NEXT: retq
%tmp = call i64 @llvm.cttz.i64( i64 %x, i1 true )
ret i64 %tmp
-; CHECK-LABEL: cttz_i64:
-; CHECK: bsfq
-; CHECK-NOT: cmov
-; CHECK: ret
}
define i8 @ctlz_i8(i8 %x) {
-entry:
+; CHECK-LABEL: ctlz_i8:
+; CHECK: # BB#0:
+; CHECK-NEXT: movzbl %dil, %eax
+; CHECK-NEXT: bsrl %eax, %eax
+; CHECK-NEXT: xorl $7, %eax
+; CHECK-NEXT: retq
%tmp2 = call i8 @llvm.ctlz.i8( i8 %x, i1 true )
ret i8 %tmp2
-; CHECK-LABEL: ctlz_i8:
-; CHECK: bsrl
-; CHECK-NOT: cmov
-; CHECK: xorl $7,
-; CHECK: ret
}
define i16 @ctlz_i16(i16 %x) {
-entry:
+; CHECK-LABEL: ctlz_i16:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsrw %di, %ax
+; CHECK-NEXT: xorl $15, %eax
+; CHECK-NEXT: retq
%tmp2 = call i16 @llvm.ctlz.i16( i16 %x, i1 true )
ret i16 %tmp2
-; CHECK-LABEL: ctlz_i16:
-; CHECK: bsrw
-; CHECK-NOT: cmov
-; CHECK: xorl $15,
-; CHECK: ret
}
define i32 @ctlz_i32(i32 %x) {
+; CHECK-LABEL: ctlz_i32:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsrl %edi, %eax
+; CHECK-NEXT: xorl $31, %eax
+; CHECK-NEXT: retq
%tmp = call i32 @llvm.ctlz.i32( i32 %x, i1 true )
ret i32 %tmp
-; CHECK-LABEL: ctlz_i32:
-; CHECK: bsrl
-; CHECK-NOT: cmov
-; CHECK: xorl $31,
-; CHECK: ret
}
define i64 @ctlz_i64(i64 %x) {
+; CHECK-LABEL: ctlz_i64:
+; CHECK: # BB#0:
+; CHECK-NEXT: bsrq %rdi, %rax
+; CHECK-NEXT: xorq $63, %rax
+; CHECK-NEXT: retq
%tmp = call i64 @llvm.ctlz.i64( i64 %x, i1 true )
ret i64 %tmp
-; CHECK-LABEL: ctlz_i64:
-; CHECK: bsrq
-; CHECK-NOT: cmov
-; CHECK: xorq $63,
-; CHECK: ret
}
-define i32 @ctlz_i32_cmov(i32 %n) {
-entry:
-; Generate a cmov to handle zero inputs when necessary.
-; CHECK-LABEL: ctlz_i32_cmov:
-; CHECK: bsrl
-; CHECK: cmov
-; CHECK: xorl $31,
-; CHECK: ret
+define i32 @ctlz_i32_zero_test(i32 %n) {
+; Generate a test and branch to handle zero inputs because bsr/bsf are very slow.
+
+; CHECK-LABEL: ctlz_i32_zero_test:
+; CHECK: # BB#0:
+; CHECK-NEXT: movl $32, %eax
+; CHECK-NEXT: testl %edi, %edi
+; CHECK-NEXT: je .LBB8_2
+; CHECK-NEXT: # BB#1: # %cond.false
+; CHECK-NEXT: bsrl %edi, %eax
+; CHECK-NEXT: xorl $31, %eax
+; CHECK-NEXT: .LBB8_2: # %cond.end
+; CHECK-NEXT: retq
%tmp1 = call i32 @llvm.ctlz.i32(i32 %n, i1 false)
ret i32 %tmp1
}
define i32 @ctlz_i32_fold_cmov(i32 %n) {
-entry:
; Don't generate the cmovne when the source is known non-zero (and bsr would
; not set ZF).
; rdar://9490949
+; FIXME: The compare and branch are produced late in IR (by CodeGenPrepare), and
+; codegen doesn't know how to delete the movl and je.
+
; CHECK-LABEL: ctlz_i32_fold_cmov:
-; CHECK: bsrl
-; CHECK-NOT: cmov
-; CHECK: xorl $31,
-; CHECK: ret
+; CHECK: # BB#0:
+; CHECK-NEXT: orl $1, %edi
+; CHECK-NEXT: movl $32, %eax
+; CHECK-NEXT: je .LBB9_2
+; CHECK-NEXT: # BB#1: # %cond.false
+; CHECK-NEXT: bsrl %edi, %eax
+; CHECK-NEXT: xorl $31, %eax
+; CHECK-NEXT: .LBB9_2: # %cond.end
+; CHECK-NEXT: retq
%or = or i32 %n, 1
%tmp1 = call i32 @llvm.ctlz.i32(i32 %or, i1 false)
ret i32 %tmp1
}
define i32 @ctlz_bsr(i32 %n) {
-entry:
; Don't generate any xors when a 'ctlz' intrinsic is actually used to compute
; the most significant bit, which is what 'bsr' does natively.
+
; CHECK-LABEL: ctlz_bsr:
-; CHECK: bsrl
-; CHECK-NOT: xorl
-; CHECK: ret
+; CHECK: # BB#0:
+; CHECK-NEXT: bsrl %edi, %eax
+; CHECK-NEXT: retq
%ctlz = call i32 @llvm.ctlz.i32(i32 %n, i1 true)
%bsr = xor i32 %ctlz, 31
ret i32 %bsr
}
-define i32 @ctlz_bsr_cmov(i32 %n) {
-entry:
-; Same as ctlz_bsr, but ensure this happens even when there is a potential
-; zero.
-; CHECK-LABEL: ctlz_bsr_cmov:
-; CHECK: bsrl
-; CHECK-NOT: xorl
-; CHECK: ret
+define i32 @ctlz_bsr_zero_test(i32 %n) {
+; Generate a test and branch to handle zero inputs because bsr/bsf are very slow.
+; FIXME: The compare and branch are produced late in IR (by CodeGenPrepare), and
+; codegen doesn't know how to combine the $32 and $31 into $63.
+
+; CHECK-LABEL: ctlz_bsr_zero_test:
+; CHECK: # BB#0:
+; CHECK-NEXT: movl $32, %eax
+; CHECK-NEXT: testl %edi, %edi
+; CHECK-NEXT: je .LBB11_2
+; CHECK-NEXT: # BB#1: # %cond.false
+; CHECK-NEXT: bsrl %edi, %eax
+; CHECK-NEXT: xorl $31, %eax
+; CHECK-NEXT: .LBB11_2: # %cond.end
+; CHECK-NEXT: xorl $31, %eax
+; CHECK-NEXT: retq
%ctlz = call i32 @llvm.ctlz.i32(i32 %n, i1 false)
%bsr = xor i32 %ctlz, 31
ret i32 %bsr
generated by cgit v1.2.3 (git 2.25.1) at 2024-09-21 19:31:44 +0000