[ConstantFold][SVE] Fix constant folding for shufflevector.

[efriedma]

Summary

Don't try to fold away shuffles which can't be folded. Fix creation of shufflevector constant expressions.

[spatel]

Similar to D70985 - LGTM.

I haven't looked at scalable vectors very much though, so I'm not sure if there are still problems for any other opcodes. Eg, ConstantFoldSelectInstruction can iterate based on number of elements.

llvm/test/Analysis/ConstantFolding/shufflevector.ll

5

Can reduce this to a single shuffle and still get the same crash?

define <vscale x 4 x i32> @shufflevector_scalable_constant() {
  %r = shufflevector <vscale x 4 x i32> undef, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
  ret <vscale x 4 x i32> %r
}

[efriedma]

I haven't looked at scalable vectors very much though, so I'm not sure if there are still problems for any other opcodes.

There are; Huihui is going to look into this. This particular case happened to be blocking some other work I'm doing.

llvm/test/Analysis/ConstantFolding/shufflevector.ll

5

We'll probably fold that to undef at some point.

llvm/lib/IR/ConstantFold.cpp

Constant *llvm::ConstantFoldShuffleVectorInstruction(Constant *V1,
 
   // Undefined shuffle mask -> undefined value.
   if (isa<UndefValue>(Mask))
     return UndefValue::get(VectorType::get(EltTy, MaskNumElts));
 
   // Don't break the bitcode reader hack.
   if (isa<ConstantExpr>(Mask)) return nullptr;
 
+  // Do not iterate on scalable vector. The num of elements is unknown at
+  // compile-time.
+  VectorType *ValTy = cast<VectorType>(V1->getType());
+  if (ValTy->isScalable())
+    return nullptr;
+
   unsigned SrcNumElts = V1->getType()->getVectorNumElements();
 
   // Loop over the shuffle mask, evaluating each element.
   SmallVector<Constant*, 32> Result;
   for (unsigned i = 0; i != MaskNumElts; ++i) {
     int Elt = ShuffleVectorInst::getMaskValue(Mask, i);
     if (Elt == -1) {
       Result.push_back(UndefValue::get(EltTy));

llvm/lib/IR/Constants.cpp

 Constant *ConstantExpr::getShuffleVector(Constant *V1, Constant *V2,
                                          Constant *Mask, Type *OnlyIfReducedTy) {
   assert(ShuffleVectorInst::isValidOperands(V1, V2, Mask) &&
          "Invalid shuffle vector constant expr operands!");
 
   if (Constant *FC = ConstantFoldShuffleVectorInstruction(V1, V2, Mask))
     return FC;          // Fold a few common cases.
 
-  unsigned NElts = Mask->getType()->getVectorNumElements();
+  ElementCount NElts = Mask->getType()->getVectorElementCount();
   Type *EltTy = V1->getType()->getVectorElementType();
   Type *ShufTy = VectorType::get(EltTy, NElts);
 
   if (OnlyIfReducedTy == ShufTy)
     return nullptr;
 
   // Look up the constant in the table first to ensure uniqueness
   Constant *ArgVec[] = { V1, V2, Mask };

llvm/lib/IR/ConstantsContext.h

 /// ShuffleVectorConstantExpr - This class is private to
 /// Constants.cpp, and is used behind the scenes to implement
 /// shufflevector constant exprs.
 class ShuffleVectorConstantExpr : public ConstantExpr {
 public:
   ShuffleVectorConstantExpr(Constant *C1, Constant *C2, Constant *C3)
   : ConstantExpr(VectorType::get(
                    cast<VectorType>(C1->getType())->getElementType(),
-                   cast<VectorType>(C3->getType())->getNumElements()),
+                   cast<VectorType>(C3->getType())->getElementCount()),
                  Instruction::ShuffleVector,
                  &Op<0>(), 3) {
     Op<0>() = C1;
     Op<1>() = C2;
     Op<2>() = C3;
   }
 
   // allocate space for exactly three operands

llvm/test/Analysis/ConstantFolding/shufflevector.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -constprop -S | FileCheck %s
+
+define <vscale x 4 x i32> @shufflevector_scalable_constant() {
+; CHECK-LABEL: @shufflevector_scalable_constant(
+; CHECK-NEXT:    ret <vscale x 4 x i32> shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> undef, i32 1, i32 0), <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer)
+;
+  %i = insertelement <vscale x 4 x i32> undef, i32 1, i32 0
+  %i2 = shufflevector <vscale x 4 x i32> %i, <vscale x 4 x i32> undef, <vscale x 4 x i32> zeroinitializer
+  ret <vscale x 4 x i32> %i2
+}