[llvm-commits] [llvm] r145376 - in /llvm/trunk: lib/Target/X86/X86InstrInfo.cpp lib/Target/X86/X86InstrSSE.td test/CodeGen/X86/avx-cast.ll test/CodeGen/X86/avx-intrinsics-x86.ll test/CodeGen/X86/avx-splat.ll test/CodeGen/X86/avx2-intrinsics-x86.ll
Craig Topper
craig.topper at gmail.com
Mon Nov 28 21:37:58 PST 2011
Author: ctopper
Date: Mon Nov 28 23:37:58 2011
New Revision: 145376
URL: http://llvm.org/viewvc/llvm-project?rev=145376&view=rev
Log:
Fix VINSERTF128/VEXTRACTF128 to be marked as FP instructions. Allow execution dependency fix pass to convert them to their integer equivalents when AVX2 is enabled.
Modified:
llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
llvm/trunk/lib/Target/X86/X86InstrSSE.td
llvm/trunk/test/CodeGen/X86/avx-cast.ll
llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
llvm/trunk/test/CodeGen/X86/avx-splat.ll
llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll
Modified: llvm/trunk/lib/Target/X86/X86InstrInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrInfo.cpp?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrInfo.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86InstrInfo.cpp Mon Nov 28 23:37:58 2011
@@ -3568,10 +3568,14 @@
{ X86::VANDPSYrr, X86::VANDPDYrr, X86::VPANDYrr },
{ X86::VORPSYrm, X86::VORPDYrm, X86::VPORYrm },
{ X86::VORPSYrr, X86::VORPDYrr, X86::VPORYrr },
- { X86::VPERM2F128rm, X86::VPERM2F128rm, X86::VPERM2I128rm },
- { X86::VPERM2F128rr, X86::VPERM2F128rr, X86::VPERM2I128rr },
{ X86::VXORPSYrm, X86::VXORPDYrm, X86::VPXORYrm },
- { X86::VXORPSYrr, X86::VXORPDYrr, X86::VPXORYrr }
+ { X86::VXORPSYrr, X86::VXORPDYrr, X86::VPXORYrr },
+ { X86::VEXTRACTF128mr, X86::VEXTRACTF128mr, X86::VEXTRACTI128mr },
+ { X86::VEXTRACTF128rr, X86::VEXTRACTF128rr, X86::VEXTRACTI128rr },
+ { X86::VINSERTF128rm, X86::VINSERTF128rm, X86::VINSERTI128rm },
+ { X86::VINSERTF128rr, X86::VINSERTF128rr, X86::VINSERTI128rr },
+ { X86::VPERM2F128rm, X86::VPERM2F128rm, X86::VPERM2I128rm },
+ { X86::VPERM2F128rr, X86::VPERM2F128rr, X86::VPERM2I128rr }
};
// FIXME: Some shuffle and unpack instructions have equivalents in different
Modified: llvm/trunk/lib/Target/X86/X86InstrSSE.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86InstrSSE.td?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86InstrSSE.td (original)
+++ llvm/trunk/lib/Target/X86/X86InstrSSE.td Mon Nov 28 23:37:58 2011
@@ -7147,7 +7147,7 @@
//===----------------------------------------------------------------------===//
// VINSERTF128 - Insert packed floating-point values
//
-let neverHasSideEffects = 1 in {
+let neverHasSideEffects = 1, ExeDomain = SSEPackedSingle in {
def VINSERTF128rr : AVXAIi8<0x18, MRMSrcReg, (outs VR256:$dst),
(ins VR256:$src1, VR128:$src2, i8imm:$src3),
"vinsertf128\t{$src3, $src2, $src1, $dst|$dst, $src1, $src2, $src3}",
@@ -7194,7 +7194,7 @@
//===----------------------------------------------------------------------===//
// VEXTRACTF128 - Extract packed floating-point values
//
-let neverHasSideEffects = 1 in {
+let neverHasSideEffects = 1, ExeDomain = SSEPackedSingle in {
def VEXTRACTF128rr : AVXAIi8<0x19, MRMDestReg, (outs VR128:$dst),
(ins VR256:$src1, i8imm:$src2),
"vextractf128\t{$src2, $src1, $dst|$dst, $src1, $src2}",
Modified: llvm/trunk/test/CodeGen/X86/avx-cast.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-cast.ll?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-cast.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-cast.ll Mon Nov 28 23:37:58 2011
@@ -16,7 +16,7 @@
ret <4 x double> %shuffle.i
}
-; CHECK: vpxor
+; CHECK: vxorps
; CHECK-NEXT: vinsertf128 $0
define <4 x i64> @castC(<2 x i64> %m) nounwind uwtable readnone ssp {
entry:
Modified: llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-intrinsics-x86.ll Mon Nov 28 23:37:58 2011
@@ -2140,7 +2140,8 @@
define void @test_x86_avx_storeu_dq_256(i8* %a0, <32 x i8> %a1) {
- ; CHECK: vmovdqu
+ ; FIXME: unfortunately the execution domain fix pass changes this to vmovups and its hard to force with no 256-bit integer instructions
+ ; CHECK: vmovups
; add operation forces the execution domain.
%a2 = add <32 x i8> %a1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
call void @llvm.x86.avx.storeu.dq.256(i8* %a0, <32 x i8> %a2)
Modified: llvm/trunk/test/CodeGen/X86/avx-splat.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx-splat.ll?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx-splat.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx-splat.ll Mon Nov 28 23:37:58 2011
@@ -47,7 +47,7 @@
; shuffle (scalar_to_vector (load (ptr + 4))), undef, <0, 0, 0, 0>
; To:
; shuffle (vload ptr)), undef, <1, 1, 1, 1>
-; CHECK: vmovdqa
+; CHECK: vmovaps
; CHECK-NEXT: vinsertf128 $1
; CHECK-NEXT: vpermilps $-1
define <8 x float> @funcE() nounwind {
Modified: llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll?rev=145376&r1=145375&r2=145376&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/avx2-intrinsics-x86.ll Mon Nov 28 23:37:58 2011
@@ -1046,3 +1046,13 @@
ret <8 x i32> %res
}
declare <8 x i32> @llvm.x86.avx2.psrav.d.256(<8 x i32>, <8 x i32>) nounwind readnone
+
+; This is checked here because the execution dependency fix pass makes it hard to test in AVX mode since we don't have 256-bit integer instructions
+define void @test_x86_avx_storeu_dq_256(i8* %a0, <32 x i8> %a1) {
+ ; CHECK: vmovdqu
+ ; add operation forces the execution domain.
+ %a2 = add <32 x i8> %a1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+ call void @llvm.x86.avx.storeu.dq.256(i8* %a0, <32 x i8> %a2)
+ ret void
+}
+declare void @llvm.x86.avx.storeu.dq.256(i8*, <32 x i8>) nounwind
More information about the llvm-commits
mailing list