[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
diff --git a/llvm/lib/Target/X86/X86InstrInfo.cpp b/llvm/lib/Target/X86/X86InstrInfo.cpp
index dd463f1..a49dcc7 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.cpp
+++ b/llvm/lib/Target/X86/X86InstrInfo.cpp
@@ -2449,7 +2449,8 @@
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);
diff --git a/llvm/test/CodeGen/X86/commuted-blend-mask.ll b/llvm/test/CodeGen/X86/commuted-blend-mask.ll
new file mode 100644
index 0000000..e6322cb
--- /dev/null
+++ b/llvm/test/CodeGen/X86/commuted-blend-mask.ll
@@ -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
+}