diff options
| author | Manman Ren <mren@apple.com> | 2012-06-07 22:39:10 +0000 |
|---|---|---|
| committer | Manman Ren <mren@apple.com> | 2012-06-07 22:39:10 +0000 |
| commit | 9236362a64ce1609222512fe3f17eeb886a3ddea (patch) | |
| tree | cb91606183c09759e1c7b415ea72c4665025cc7c /test/CodeGen/X86/iabs.ll | |
| parent | fa371eaff3c06b4d5a5ff5706a552f519b469756 (diff) | |
| download | external_llvm-9236362a64ce1609222512fe3f17eeb886a3ddea.zip external_llvm-9236362a64ce1609222512fe3f17eeb886a3ddea.tar.gz external_llvm-9236362a64ce1609222512fe3f17eeb886a3ddea.tar.bz2 | |
X86: optimize generated code for integer ABS
This patch will generate the following for integer ABS:
movl %edi, %eax
negl %eax
cmovll %edi, %eax
INSTEAD OF
movl %edi, %ecx
sarl $31, %ecx
leal (%rdi,%rcx), %eax
xorl %ecx, %eax
There exists a target-independent DAG combine for integer ABS, which converts
integer ABS to sar+add+xor. For X86, we match this pattern back to neg+cmov.
This is implemented in PerformXorCombine.
rdar://10695237
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@158175 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/iabs.ll')
| -rw-r--r-- | test/CodeGen/X86/iabs.ll | 16 |
1 files changed, 10 insertions, 6 deletions
diff --git a/test/CodeGen/X86/iabs.ll b/test/CodeGen/X86/iabs.ll index a8ba015..9196cce 100644 --- a/test/CodeGen/X86/iabs.ll +++ b/test/CodeGen/X86/iabs.ll @@ -1,13 +1,17 @@ -; RUN: llc < %s -march=x86-64 -stats |& \ -; RUN: grep {5 .*Number of machine instrs printed} +; RUN: llc < %s -march=x86-64 | FileCheck %s ;; Integer absolute value, should produce something at least as good as: -;; movl %edi, %ecx -;; sarl $31, %ecx -;; leal (%rdi,%rcx), %eax -;; xorl %ecx, %eax +;; movl %edi, %eax +;; negl %eax +;; cmovll %edi, %eax ;; ret +; rdar://10695237 define i32 @test(i32 %a) nounwind { +; CHECK: test: +; CHECK: mov +; CHECK-NEXT: neg +; CHECK-NEXT: cmov +; CHECK-NEXT: ret %tmp1neg = sub i32 0, %a %b = icmp sgt i32 %a, -1 %abs = select i1 %b, i32 %a, i32 %tmp1neg |