[PATCH] D121452: [SVE][VPlan] Avoid collecting scalars for SVE

[Malhar via Phabricator via llvm-commits]

malharJ updated this revision to Diff 414630.
malharJ added a comment.

- Updated terminology to use scalable vectors instead of SVE


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D121452/new/

https://reviews.llvm.org/D121452

Files:
  llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
  llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll


Index: llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll
===================================================================
--- /dev/null
+++ llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll
@@ -0,0 +1,34 @@
+; RUN: opt -mtriple=aarch64 -loop-vectorize -S < %s | FileCheck %s
+
+target triple = "aarch64-unknown-linux-gnu"
+
+; CHECK-NOT: Assertion {{.*}} "Can't scalarize a scalable vector"' failed
+
+; The test checks that scalarized code is not generated for SVE.
+; It creates a scenario where the gep instruction is used outside
+; the loop, preventing it (and consequently the loop induction
+; update variable) from being classified as 'uniform'.
+
+define void @test_no_scalarization(i64* %a) #0 {
+L.entry:
+  %idx = alloca i32, align 4
+  store i32 100, i32* %idx, align 4
+  %0 = load i32, i32* %idx
+  br label %L.LB19_336
+
+L.LB19_336:                                       ; preds = %L.LB19_336, %L.entry
+  %indvars.iv = phi i32 [ %indvars.iv.next, %L.LB19_336 ], [ %0, %L.entry ]
+  %indvars.iv.next = add nsw i32 %indvars.iv, -1
+  %1 = getelementptr i64, i64* %a, i32 %indvars.iv
+  %2 = bitcast i64* %1 to double*
+  %3 = load double, double* %2, align 8
+  %4 = icmp sgt i32 %indvars.iv.next, 1
+  br i1 %4, label %L.LB19_336, label %L.LB19_337
+
+L.LB19_337:                                       ; preds = %L.LB19_336
+  store i64 1, i64* %1, align 8
+  ret void
+}
+
+attributes #0 = { nofree norecurse noreturn nosync nounwind "target-features"="+neon,+v8a,+sve" }
+
Index: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
===================================================================
--- llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -4636,6 +4636,14 @@
   assert(VF.isVector() && Scalars.find(VF) == Scalars.end() &&
          "This function should not be visited twice for the same VF");
 
+  // This avoids any chances of creating a REPLICATE recipe during planning
+  // since that would result in generation of scalarized code during execution,
+  // which is not supported for scalable vectors.
+  if (VF.isScalable()) {
+    Scalars[VF].insert(Uniforms[VF].begin(), Uniforms[VF].end());
+    return;
+  }
+
   SmallSetVector<Instruction *, 8> Worklist;
 
   // These sets are used to seed the analysis with pointers used by memory


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D121452.414630.patch
Type: text/x-patch
Size: 2379 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20220311/52c69c64/attachment.bin>