[llvm] 49d7313 - [LV] Avoid scalable vectorization for loops containing alloca

[Kerry McLaughlin via llvm-commits]

Author: Kerry McLaughlin
Date: 2021-07-16T11:47:13+01:00
New Revision: 49d73130ca17a19bd68c251451a4ff0c0cdc00e1

URL: https://github.com/llvm/llvm-project/commit/49d73130ca17a19bd68c251451a4ff0c0cdc00e1
DIFF: https://github.com/llvm/llvm-project/commit/49d73130ca17a19bd68c251451a4ff0c0cdc00e1.diff

LOG: [LV] Avoid scalable vectorization for loops containing alloca

This patch returns an Invalid cost from getInstructionCost() for alloca
instructions if the VF is scalable, as otherwise loops which contain
these instructions will crash when attempting to scalarize the alloca.

Reviewed By: sdesmalen

Differential Revision: https://reviews.llvm.org/D105824

Added: 
    llvm/test/Transforms/LoopVectorize/AArch64/scalable-alloca.ll

Modified: 
    llvm/lib/Transforms/Vectorize/LoopVectorize.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 64dce08eacd2..9b5f370b7abd 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -7903,6 +7903,12 @@ LoopVectorizationCostModel::getInstructionCost(Instruction *I, ElementCount VF,
   }
   case Instruction::ExtractValue:
     return TTI.getInstructionCost(I, TTI::TCK_RecipThroughput);
+  case Instruction::Alloca:
+    // We cannot easily widen alloca to a scalable alloca, as
+    // the result would need to be a vector of pointers.
+    if (VF.isScalable())
+      return InstructionCost::getInvalid();
+    LLVM_FALLTHROUGH;
   default:
     // This opcode is unknown. Assume that it is the same as 'mul'.
     return TTI.getArithmeticInstrCost(Instruction::Mul, VectorTy, CostKind);

diff  --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-alloca.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-alloca.ll
new file mode 100644
index 000000000000..aa1bc240ebfd
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-alloca.ll
@@ -0,0 +1,31 @@
+; RUN: opt -S -loop-vectorize -mattr=+sve -mtriple aarch64-unknown-linux-gnu -force-vector-width=2 -scalable-vectorization=preferred -pass-remarks-analysis=loop-vectorize -pass-remarks-missed=loop-vectorize < %s 2>%t | FileCheck %s
+; RUN: FileCheck %s --check-prefix=CHECK-REMARKS < %t
+
+; CHECK-REMARKS: UserVF ignored because of invalid costs.
+; CHECK-REMARKS: Instruction with invalid costs prevented vectorization at VF=(vscale x 1, vscale x 2): alloca
+; CHECK-REMARKS: Instruction with invalid costs prevented vectorization at VF=(vscale x 1): store
+define void @alloca(i32** %vla, i64 %N) {
+; CHECK-LABEL: @alloca(
+; CHECK-NOT: <vscale x
+
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ %iv.next, %for.body ], [ 0, %entry ]
+  %alloca = alloca i32, align 16
+  %arrayidx = getelementptr inbounds i32*, i32** %vla, i64 %iv
+  store i32* %alloca, i32** %arrayidx, align 8
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond.not = icmp eq i64 %iv.next, %N
+  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0
+
+for.end:
+  call void @foo(i32** nonnull %vla)
+  ret void
+}
+
+declare void @foo(i32**)
+
+!0 = !{!0, !1}
+!1 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}