maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[X86] When commuting SSE immediate blend, make sure that the new blend mask is a valid imm8. 

Example:
define <4 x i32> @test(<4 x i32> %a, <4 x i32> %b) {
  %shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 4, i32 5, i32 6, i32 3>
  ret <4 x i32> %shuffle
}

Before llc (-mattr=+sse4.1), produced the following assembly instruction:
  pblendw $4294967103, %xmm1, %xmm0

After
  pblendw $63, %xmm1, %xmm0

llvm-svn: 221455
This commit is contained in:
Andrea Di Biagio 2014-11-06 14:36:45 +00:00
parent e77ffe35bf
commit 7ecd22ca4a
2 changed files with 15 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/lib/Target/X86/X86InstrInfo.cpp
View File

@ -2449,7 +2449,8 @@ X86InstrInfo::commuteInstruction(MachineInstr *MI, bool NewMI) const {
case X86::VPBLENDDYrri: Mask = 0xFF; break;
case X86::VPBLENDWYrri: Mask = 0xFF; break;
}
unsigned Imm = MI->getOperand(3).getImm();
// Only the least significant bits of Imm are used.
unsigned Imm = MI->getOperand(3).getImm() & Mask;
if (NewMI) {
MachineFunction &MF = *MI->getParent()->getParent();
MI = MF.CloneMachineInstr(MI);

13
llvm/test/CodeGen/X86/commuted-blend-mask.ll Normal file
View File

@ -0,0 +1,13 @@
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 < %s | FileCheck %s
; When commuting the operands of a SSE blend, make sure that the resulting blend
; mask can be encoded as a imm8.
; Before, when commuting the operands to the shuffle in function @test, the backend
; produced the following assembly:
; pblendw $4294967103, %xmm1, %xmm0
define <4 x i32> @test(<4 x i32> %a, <4 x i32> %b) {
;CHECK: pblendw $63, %xmm1, %xmm0
%shuffle = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 4, i32 5, i32 6, i32 3>
ret <4 x i32> %shuffle
}