[X86] Auto upgrade old style VPCOM/VPCOMU intrinsics to generic integer comparisons
We were upgrading these to the new style VPCOM/VPCOMU intrinsics (which includes the condition code immediate), but we'll be getting rid of those shortly, so convert these to generics first.
This causes a couple of changes in the upgrade tests as signed/unsigned eq/ne are equivalent and we constant fold true/false codes, these changes are the same as what we already do for avx512 cmp/ucmp.
Noticed while cleaning up vector integer comparison costs for PR40376.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@351690 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/IR/AutoUpgrade.cpp b/lib/IR/AutoUpgrade.cpp
index 321c8c3..b28e9f5 100644
--- a/lib/IR/AutoUpgrade.cpp
+++ b/lib/IR/AutoUpgrade.cpp
@@ -1052,6 +1052,45 @@
return Res;
}
+static Value *upgradeX86vpcom(IRBuilder<> &Builder, CallInst &CI, unsigned Imm,
+ bool IsSigned) {
+ Type *Ty = CI.getType();
+ Value *LHS = CI.getArgOperand(0);
+ Value *RHS = CI.getArgOperand(1);
+
+ CmpInst::Predicate Pred;
+ switch (Imm) {
+ case 0x0:
+ Pred = IsSigned ? ICmpInst::ICMP_SLT : ICmpInst::ICMP_ULT;
+ break;
+ case 0x1:
+ Pred = IsSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE;
+ break;
+ case 0x2:
+ Pred = IsSigned ? ICmpInst::ICMP_SGT : ICmpInst::ICMP_UGT;
+ break;
+ case 0x3:
+ Pred = IsSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE;
+ break;
+ case 0x4:
+ Pred = ICmpInst::ICMP_EQ;
+ break;
+ case 0x5:
+ Pred = ICmpInst::ICMP_NE;
+ break;
+ case 0x6:
+ return Constant::getNullValue(Ty); // FALSE
+ case 0x7:
+ return Constant::getAllOnesValue(Ty); // TRUE
+ default:
+ llvm_unreachable("Unknown XOP vpcom/vpcomu predicate");
+ }
+
+ Value *Cmp = Builder.CreateICmp(Pred, LHS, RHS);
+ Value *Ext = Builder.CreateSExt(Cmp, Ty);
+ return Ext;
+}
+
static Value *upgradeX86ConcatShift(IRBuilder<> &Builder, CallInst &CI,
bool IsShiftRight, bool ZeroMask) {
Type *Ty = CI.getType();
@@ -1989,23 +2028,13 @@
ResultTy);
Rep = Builder.CreateCall(CSt, { CI->getArgOperand(1), Ptr, MaskVec });
} else if (IsX86 && Name.startswith("xop.vpcom")) {
- Intrinsic::ID intID;
- if (Name.endswith("ub"))
- intID = Intrinsic::x86_xop_vpcomub;
- else if (Name.endswith("uw"))
- intID = Intrinsic::x86_xop_vpcomuw;
- else if (Name.endswith("ud"))
- intID = Intrinsic::x86_xop_vpcomud;
- else if (Name.endswith("uq"))
- intID = Intrinsic::x86_xop_vpcomuq;
- else if (Name.endswith("b"))
- intID = Intrinsic::x86_xop_vpcomb;
- else if (Name.endswith("w"))
- intID = Intrinsic::x86_xop_vpcomw;
- else if (Name.endswith("d"))
- intID = Intrinsic::x86_xop_vpcomd;
- else if (Name.endswith("q"))
- intID = Intrinsic::x86_xop_vpcomq;
+ bool IsSigned;
+ if (Name.endswith("ub") || Name.endswith("uw") || Name.endswith("ud") ||
+ Name.endswith("uq"))
+ IsSigned = false;
+ else if (Name.endswith("b") || Name.endswith("w") || Name.endswith("d") ||
+ Name.endswith("q"))
+ IsSigned = true;
else
llvm_unreachable("Unknown suffix");
@@ -2029,11 +2058,7 @@
Imm = 7;
else
llvm_unreachable("Unknown condition");
-
- Function *VPCOM = Intrinsic::getDeclaration(F->getParent(), intID);
- Rep =
- Builder.CreateCall(VPCOM, {CI->getArgOperand(0), CI->getArgOperand(1),
- Builder.getInt8(Imm)});
+ Rep = upgradeX86vpcom(Builder, *CI, Imm, IsSigned);
} else if (IsX86 && Name.startswith("xop.vpcmov")) {
Value *Sel = CI->getArgOperand(2);
Value *NotSel = Builder.CreateNot(Sel);
diff --git a/test/CodeGen/X86/xop-intrinsics-x86_64-upgrade.ll b/test/CodeGen/X86/xop-intrinsics-x86_64-upgrade.ll
index 0f34836..a9f237a 100644
--- a/test/CodeGen/X86/xop-intrinsics-x86_64-upgrade.ll
+++ b/test/CodeGen/X86/xop-intrinsics-x86_64-upgrade.ll
@@ -129,7 +129,7 @@
define <16 x i8> @test_int_x86_xop_vpcomequb(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomequb:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomequb %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomeqb %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomequb(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -139,7 +139,7 @@
define <4 x i32> @test_int_x86_xop_vpcomequd(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomequd:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomequd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomeqd %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomequd(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -149,7 +149,7 @@
define <2 x i64> @test_int_x86_xop_vpcomequq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomequq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomequq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomeqq %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomequq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -159,7 +159,7 @@
define <8 x i16> @test_int_x86_xop_vpcomequw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomequw:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomequw %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomeqw %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomequw(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
@@ -169,7 +169,7 @@
define <16 x i8> @test_int_x86_xop_vpcomfalseb(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseb:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseb %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomfalseb(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -179,7 +179,7 @@
define <4 x i32> @test_int_x86_xop_vpcomfalsed(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalsed:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalsed %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomfalsed(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -189,7 +189,7 @@
define <2 x i64> @test_int_x86_xop_vpcomfalseq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomfalseq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -199,7 +199,7 @@
define <16 x i8> @test_int_x86_xop_vpcomfalseub(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseub:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseub %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomfalseub(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -209,7 +209,7 @@
define <4 x i32> @test_int_x86_xop_vpcomfalseud(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseud:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseud %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomfalseud(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -219,7 +219,7 @@
define <2 x i64> @test_int_x86_xop_vpcomfalseuq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseuq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseuq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomfalseuq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -229,7 +229,7 @@
define <8 x i16> @test_int_x86_xop_vpcomfalseuw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalseuw:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalseuw %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomfalseuw(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
@@ -239,7 +239,7 @@
define <8 x i16> @test_int_x86_xop_vpcomfalsew(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomfalsew:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomfalsew %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomfalsew(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
@@ -599,7 +599,7 @@
define <16 x i8> @test_int_x86_xop_vpcomneub(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomneub:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomnequb %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomneqb %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomneub(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -609,7 +609,7 @@
define <4 x i32> @test_int_x86_xop_vpcomneud(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomneud:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomnequd %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomneqd %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomneud(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -619,7 +619,7 @@
define <2 x i64> @test_int_x86_xop_vpcomneuq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomneuq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomnequq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomneqq %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomneuq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -629,7 +629,7 @@
define <8 x i16> @test_int_x86_xop_vpcomneuw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomneuw:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomnequw %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcomneqw %xmm1, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomneuw(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
@@ -649,7 +649,7 @@
define <16 x i8> @test_int_x86_xop_vpcomtrueb(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueb:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueb %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomtrueb(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -659,7 +659,7 @@
define <4 x i32> @test_int_x86_xop_vpcomtrued(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrued:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrued %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomtrued(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -669,7 +669,7 @@
define <2 x i64> @test_int_x86_xop_vpcomtrueq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomtrueq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -679,7 +679,7 @@
define <16 x i8> @test_int_x86_xop_vpcomtrueub(<16 x i8> %a0, <16 x i8> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueub:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueub %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <16 x i8> @llvm.x86.xop.vpcomtrueub(<16 x i8> %a0, <16 x i8> %a1) ;
ret <16 x i8> %res
@@ -689,7 +689,7 @@
define <4 x i32> @test_int_x86_xop_vpcomtrueud(<4 x i32> %a0, <4 x i32> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueud:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueud %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <4 x i32> @llvm.x86.xop.vpcomtrueud(<4 x i32> %a0, <4 x i32> %a1) ;
ret <4 x i32> %res
@@ -699,7 +699,7 @@
define <2 x i64> @test_int_x86_xop_vpcomtrueuq(<2 x i64> %a0, <2 x i64> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueuq:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueuq %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <2 x i64> @llvm.x86.xop.vpcomtrueuq(<2 x i64> %a0, <2 x i64> %a1) ;
ret <2 x i64> %res
@@ -709,7 +709,7 @@
define <8 x i16> @test_int_x86_xop_vpcomtrueuw(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtrueuw:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtrueuw %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomtrueuw(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
@@ -719,7 +719,7 @@
define <8 x i16> @test_int_x86_xop_vpcomtruew(<8 x i16> %a0, <8 x i16> %a1) {
; CHECK-LABEL: test_int_x86_xop_vpcomtruew:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpcomtruew %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpcmpeqd %xmm0, %xmm0, %xmm0
; CHECK-NEXT: retq
%res = call <8 x i16> @llvm.x86.xop.vpcomtruew(<8 x i16> %a0, <8 x i16> %a1) ;
ret <8 x i16> %res
