[LoopVectorize] Don't create unnecessary vscale intrinsic calls

In quite a few cases in LoopVectorize.cpp we call createStepForVF
with a step value of 0, which leads to unnecessary generation of
llvm.vscale intrinsic calls. I've optimised IRBuilder::CreateVScale
and createStepForVF to return 0 when attempting to multiply
vscale by 0.

Differential Revision: https://reviews.llvm.org/D100763

diff --git a/llvm/lib/IR/IRBuilder.cpp b/llvm/lib/IR/IRBuilder.cpp
index 5317404..91c9789 100644 (file)
--- a/llvm/lib/IR/IRBuilder.cpp
+++ b/llvm/lib/IR/IRBuilder.cpp
@@ -81,8 +81,10 @@ static CallInst *createCallHelper(Function *Callee, ArrayRef<Value *> Ops,
 }
 
 Value *IRBuilderBase::CreateVScale(Constant *Scaling, const Twine &Name) {
-  Module *M = GetInsertBlock()->getParent()->getParent();
   assert(isa<ConstantInt>(Scaling) && "Expected constant integer");
+  if (cast<ConstantInt>(Scaling)->isZero())
+    return Scaling;
+  Module *M = GetInsertBlock()->getParent()->getParent();
   Function *TheFn =
       Intrinsic::getDeclaration(M, Intrinsic::vscale, {Scaling->getType()});
   CallInst *CI = createCallHelper(TheFn, {}, this, Name);

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index d4eced4..6f53055 100644 (file)
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -4773,10 +4773,9 @@ void InnerLoopVectorizer::widenPHIInstruction(Instruction *PN,
              "Currently unsupported for scalable vectors");
       unsigned Lanes = IsUniform ? 1 : State.VF.getFixedValue();
 
-      Value *RuntimeVF = getRuntimeVF(Builder, PtrInd->getType(), VF);
       for (unsigned Part = 0; Part < UF; ++Part) {
-        Value *PartStart = Builder.CreateMul(
-            RuntimeVF, ConstantInt::get(PtrInd->getType(), Part));
+        Value *PartStart = createStepForVF(
+            Builder, ConstantInt::get(PtrInd->getType(), Part), VF);
         for (unsigned Lane = 0; Lane < Lanes; ++Lane) {
           Value *Idx = Builder.CreateAdd(
               PartStart, ConstantInt::get(PtrInd->getType(), Lane));

diff --git a/llvm/unittests/IR/IRBuilderTest.cpp b/llvm/unittests/IR/IRBuilderTest.cpp
index ef3895d..a11b75f 100644 (file)
--- a/llvm/unittests/IR/IRBuilderTest.cpp
+++ b/llvm/unittests/IR/IRBuilderTest.cpp
@@ -180,6 +180,14 @@ TEST_F(IRBuilderTest, IntrinsicsWithScalableVectors) {
     EXPECT_EQ(FTy->getParamType(i), ArgTys[i]->getType());
 }
 
+TEST_F(IRBuilderTest, CreateVScale) {
+  IRBuilder<> Builder(BB);
+
+  Constant *Zero = Builder.getInt32(0);
+  Value *VScale = Builder.CreateVScale(Zero);
+  EXPECT_TRUE(isa<ConstantInt>(VScale) && cast<ConstantInt>(VScale)->isZero());
+}
+
 TEST_F(IRBuilderTest, CreateStepVector) {
   IRBuilder<> Builder(BB);