[SLPVectorizer][SVE] Bail out early for scalable vector.

[huihuiz]

Summary

SLPVectorizer try to vectorize list of scalar instructions of the same type,
instructions already vectorized are rejected through isValidElementType().

Without this patch, tryToVectorizeList() will first try to determine vectorization
factor of a list of Instructions before checking whether each instruction has unsupported
type or not. For instructions already vectorized for SVE, it will crash at getVectorElementSize(),
where it try to return a fixed size.

This patch make sure invalid element types are rejected before trying to get vectorization
factor. This make sure we are not trying to vectorize instructions already vectorized.

[huihuiz]

take test test/Transforms/SLPVectorizer/AArch64/scalable-vector.ll, run opt -slp-vectorizer

you will see it crash at: llvm-top/llvm/include/llvm/Support/TypeSize.h:126: uint64_t llvm::TypeSize::getFixedSize() const: Assertion `!IsScalable && "Request for a fixed size on a scalable object"' failed.

[efriedma]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

403

Should we be using isValidElementType(V->getType()) here, instead of explicitly checking for vector types?

[huihuiz]

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

403

Update patch by changing the checking sequence, reject invalid element type before trying to determine SLP vectorization factor.

If we move isValidElementType() into getSameOpcode(), we will lose the optimization remark missed information.

[efriedma]

LGTM

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

 /// could be vectorized even if its structure is diverse.
 static InstructionsState getSameOpcode(ArrayRef<Value *> VL,
                                        unsigned BaseIndex = 0) {
   // Make sure these are all Instructions.
   if (llvm::any_of(VL, [](Value *V) { return !isa<Instruction>(V); }))
     return InstructionsState(VL[BaseIndex], nullptr, nullptr);
 
   bool IsCastOp = isa<CastInst>(VL[BaseIndex]);
   bool IsBinOp = isa<BinaryOperator>(VL[BaseIndex]);
   unsigned Opcode = cast<Instruction>(VL[BaseIndex])->getOpcode();
   unsigned AltOpcode = Opcode;
   unsigned AltIndex = BaseIndex;
 
   // Check for one alternate opcode from another BinaryOperator.
   // TODO - generalize to support all operators (types, calls etc.).
   for (int Cnt = 0, E = VL.size(); Cnt < E; Cnt++) {
     unsigned InstOpcode = cast<Instruction>(VL[Cnt])->getOpcode();

 bool SLPVectorizerPass::tryToVectorizeList(ArrayRef<Value *> VL, BoUpSLP &R,
                                            int UserCost, bool AllowReorder) {
   if (VL.size() < 2)
     return false;
 
   LLVM_DEBUG(dbgs() << "SLP: Trying to vectorize a list of length = "
                     << VL.size() << ".\n");
 
-  // Check that all of the parts are scalar instructions of the same type,
+  // Check that all of the parts are instructions of the same type,
   // we permit an alternate opcode via InstructionsState.
   InstructionsState S = getSameOpcode(VL);
   if (!S.getOpcode())
     return false;
 
   Instruction *I0 = cast<Instruction>(S.OpValue);
-  unsigned Sz = R.getVectorElementSize(I0);
-  unsigned MinVF = std::max(2U, R.getMinVecRegSize() / Sz);
-  unsigned MaxVF = std::max<unsigned>(PowerOf2Floor(VL.size()), MinVF);
-  if (MaxVF < 2) {
-    R.getORE()->emit([&]() {
-      return OptimizationRemarkMissed(SV_NAME, "SmallVF", I0)
-             << "Cannot SLP vectorize list: vectorization factor "
-             << "less than 2 is not supported";
-    });
-    return false;
-  }
-
+  // Make sure invalid types (including vector type) are rejected before
+  // determining vectorization factor for scalar instructions.
   for (Value *V : VL) {
     Type *Ty = V->getType();
     if (!isValidElementType(Ty)) {
       // NOTE: the following will give user internal llvm type name, which may
       // not be useful.
       R.getORE()->emit([&]() {
         std::string type_str;
         llvm::raw_string_ostream rso(type_str);
         Ty->print(rso);
         return OptimizationRemarkMissed(SV_NAME, "UnsupportedType", I0)
                << "Cannot SLP vectorize list: type "
                << rso.str() + " is unsupported by vectorizer";
       });
       return false;
     }
   }
 
+  unsigned Sz = R.getVectorElementSize(I0);
+  unsigned MinVF = std::max(2U, R.getMinVecRegSize() / Sz);
+  unsigned MaxVF = std::max<unsigned>(PowerOf2Floor(VL.size()), MinVF);
+  if (MaxVF < 2) {
+    R.getORE()->emit([&]() {
+      return OptimizationRemarkMissed(SV_NAME, "SmallVF", I0)
+             << "Cannot SLP vectorize list: vectorization factor "
+             << "less than 2 is not supported";
+    });
+    return false;
+  }
+
   bool Changed = false;
   bool CandidateFound = false;
   int MinCost = SLPCostThreshold;
 
   unsigned NextInst = 0, MaxInst = VL.size();
   for (unsigned VF = MaxVF; NextInst + 1 < MaxInst && VF >= MinVF; VF /= 2) {
     // No actual vectorization should happen, if number of parts is the same as
     // provided vectorization factor (i.e. the scalar type is used for vector

llvm/test/Transforms/SLPVectorizer/AArch64/invalid_type.ll

+; RUN: opt < %s -slp-vectorizer -S -pass-remarks-missed=slp-vectorizer 2>&1 | FileCheck %s
+
+target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64-unknown-linux-gnu"
+
+; This test check that slp vectorizer is not trying to vectorize instructions already vectorized.
+; CHECK: remark: <unknown>:0:0: Cannot SLP vectorize list: type <16 x i8> is unsupported by vectorizer
+
+define void @vector() {
+  %load0 = tail call <16 x i8> @vector.load(<16 x i8> *undef, i32 1)
+  %load1 = tail call <16 x i8> @vector.load(<16 x i8> *undef, i32 2)
+  %add = add <16 x i8> %load1, %load0
+  tail call void @vector.store(<16 x i8> %add, <16 x i8>* undef, i32 1)
+  ret void
+}
+
+declare <16 x i8> @vector.load(<16 x i8>*, i32)
+declare void @vector.store(<16 x i8>, <16 x i8>*, i32)

llvm/test/Transforms/SLPVectorizer/AArch64/scalable-vector.ll

+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt < %s -slp-vectorizer -S | FileCheck %s
+
+target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64-unknown-linux-gnu"
+
+; This test check that we are not crashing or changing the code.
+
+define void @test() {
+; CHECK-LABEL: @test(
+; CHECK-NEXT:    [[LOAD0:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0nxv16i8(<vscale x 16 x i8>* undef, i32 1, <vscale x 16 x i1> undef, <vscale x 16 x i8> undef)
+; CHECK-NEXT:    [[LOAD1:%.*]] = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0nxv16i8(<vscale x 16 x i8>* undef, i32 1, <vscale x 16 x i1> undef, <vscale x 16 x i8> undef)
+; CHECK-NEXT:    [[ADD:%.*]] = add <vscale x 16 x i8> [[LOAD1]], [[LOAD0]]
+; CHECK-NEXT:    tail call void @llvm.masked.store.nxv16i8.p0nxv16i8(<vscale x 16 x i8> [[ADD]], <vscale x 16 x i8>* undef, i32 1, <vscale x 16 x i1> undef)
+; CHECK-NEXT:    ret void
+;
+  %load0 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0nxv16i8(<vscale x 16 x i8> *undef, i32 1, <vscale x 16 x i1> undef, <vscale x 16 x i8> undef)
+  %load1 = tail call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0nxv16i8(<vscale x 16 x i8> *undef, i32 1, <vscale x 16 x i1> undef, <vscale x 16 x i8> undef)
+  %add = add <vscale x 16 x i8> %load1, %load0
+  tail call void @llvm.masked.store.nxv16i8.p0nxv16i8(<vscale x 16 x i8> %add, <vscale x 16 x i8>* undef, i32 1, <vscale x 16 x i1> undef)
+  ret void
+}
+
+declare <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0nxv16i8(<vscale x 16 x i8>*, i32 immarg, <vscale x 16 x i1>, <vscale x 16 x i8>)
+declare void @llvm.masked.store.nxv16i8.p0nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>*, i32 immarg, <vscale x 16 x i1>)