AMDGPU: Fold rcp/rsq of undef to undef

llvm-svn: 298725

diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
index c259e7e..845cf10 100644 (file)
--- a/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCalls.cpp
@@ -3065,9 +3065,14 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
 
     break;
   }
-
   case Intrinsic::amdgcn_rcp: {
-    if (const ConstantFP *C = dyn_cast<ConstantFP>(II->getArgOperand(0))) {
+    Value *Src = II->getArgOperand(0);
+
+    // TODO: Move to ConstantFolding/InstSimplify?
+    if (isa<UndefValue>(Src))
+      return replaceInstUsesWith(CI, Src);
+
+    if (const ConstantFP *C = dyn_cast<ConstantFP>(Src)) {
       const APFloat &ArgVal = C->getValueAPF();
       APFloat Val(ArgVal.getSemantics(), 1.0);
       APFloat::opStatus Status = Val.divide(ArgVal,
@@ -3080,6 +3085,14 @@ Instruction *InstCombiner::visitCallInst(CallInst &CI) {
 
     break;
   }
+  case Intrinsic::amdgcn_rsq: {
+    Value *Src = II->getArgOperand(0);
+
+    // TODO: Move to ConstantFolding/InstSimplify?
+    if (isa<UndefValue>(Src))
+      return replaceInstUsesWith(CI, Src);
+    break;
+  }
   case Intrinsic::amdgcn_frexp_mant:
   case Intrinsic::amdgcn_frexp_exp: {
     Value *Src = II->getArgOperand(0);

diff --git a/llvm/test/Transforms/InstCombine/amdgcn-intrinsics.ll b/llvm/test/Transforms/InstCombine/amdgcn-intrinsics.ll
index 1afc22d..deae550 100644 (file)
--- a/llvm/test/Transforms/InstCombine/amdgcn-intrinsics.ll
+++ b/llvm/test/Transforms/InstCombine/amdgcn-intrinsics.ll
@@ -7,6 +7,12 @@
 declare float @llvm.amdgcn.rcp.f32(float) nounwind readnone
 declare double @llvm.amdgcn.rcp.f64(double) nounwind readnone
 
+; CHECK-LABEL: @test_constant_fold_rcp_f32_undef
+; CHECK-NEXT: ret float undef
+define float @test_constant_fold_rcp_f32_undef() nounwind {
+  %val = call float @llvm.amdgcn.rcp.f32(float undef) nounwind readnone
+  ret float %val
+}
 
 ; CHECK-LABEL: @test_constant_fold_rcp_f32_1
 ; CHECK-NEXT: ret float 1.000000e+00
@@ -50,6 +56,18 @@ define double @test_constant_fold_rcp_f64_43() nounwind {
   ret double %val
 }
 
+; --------------------------------------------------------------------
+; llvm.amdgcn.rsq
+; --------------------------------------------------------------------
+
+declare float @llvm.amdgcn.rsq.f32(float) nounwind readnone
+
+; CHECK-LABEL: @test_constant_fold_rsq_f32_undef
+; CHECK-NEXT: ret float undef
+define float @test_constant_fold_rsq_f32_undef() nounwind {
+  %val = call float @llvm.amdgcn.rsq.f32(float undef) nounwind readnone
+  ret float %val
+}
 
 ; --------------------------------------------------------------------
 ; llvm.amdgcn.frexp.mant