[PATCH] D98054: [LoopVectorize][SVE] Fix crash when vectorising FP negation

David Sherwood via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Fri Mar 12 08:22:28 PST 2021


david-arm updated this revision to Diff 330250.
david-arm added a comment.

- Attempted to address @sdesmalen 's comments about the cost multiplier N. I created a new patch to remove the multiplier (D98512 <https://reviews.llvm.org/D98512>), since after some investigation it seems that it should always be 1.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98054/new/

https://reviews.llvm.org/D98054

Files:
  llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
  llvm/test/Transforms/LoopVectorize/AArch64/sve-basic-vec.ll


Index: llvm/test/Transforms/LoopVectorize/AArch64/sve-basic-vec.ll
===================================================================
--- llvm/test/Transforms/LoopVectorize/AArch64/sve-basic-vec.ll
+++ llvm/test/Transforms/LoopVectorize/AArch64/sve-basic-vec.ll
@@ -18,8 +18,7 @@
 ; CHECK:         store <vscale x 4 x i32> [[TMP2]], <vscale x 4 x i32>* {{.*}}, align 4
 ;
 entry:
-  %cmp7 = icmp sgt i64 %n, 0
-  br i1 %cmp7, label %for.body, label %for.end
+  br label %for.body
 
 for.body:                                         ; preds = %entry, %for.body
   %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
@@ -50,8 +49,7 @@
 ; CHECK:         store <vscale x 4 x float> [[TMP2]], <vscale x 4 x float>* {{.*}}, align 4
 
 entry:
-  %cmp8 = icmp sgt i64 %n, 0
-  br i1 %cmp8, label %for.body, label %for.end
+  br label %for.body
 
 for.body:                                         ; preds = %entry, %for.body
   %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
@@ -63,7 +61,33 @@
   store float %conv, float* %arrayidx3, align 4
   %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
   %exitcond.not = icmp eq i64 %indvars.iv.next, %n
-  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !6
+  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0
+
+for.end:                                          ; preds = %for.body, %entry
+  ret void
+}
+
+define void @fneg_f32(float* noalias nocapture %a, float* noalias nocapture readonly %b, i64 %n) {
+; CHECK-LABEL: @fneg_f32(
+; CHECK-NEXT:  entry:
+; CHECK:       vector.body:
+; CHECK:         [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, <vscale x 4 x float>* {{.*}}, align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = fneg <vscale x 4 x float> [[WIDE_LOAD]]
+; CHECK:         store <vscale x 4 x float> [[TMP1]], <vscale x 4 x float>* {{.*}}, align 4
+
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds float, float* %b, i64 %indvars.iv
+  %0 = load float, float* %arrayidx, align 4
+  %fneg = fneg float %0
+  %arrayidx3 = getelementptr inbounds float, float* %a, i64 %indvars.iv
+  store float %fneg, float* %arrayidx3, align 4
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %exitcond.not = icmp eq i64 %indvars.iv.next, %n
+  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0
 
 for.end:                                          ; preds = %for.body, %entry
   ret void
@@ -75,4 +99,3 @@
 !3 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
 !4 = !{!"llvm.loop.interleave.count", i32 1}
 !5 = !{!"llvm.loop.vectorize.enable", i1 true}
-!6 = distinct !{!6, !1, !2, !3, !4, !5}
Index: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
===================================================================
--- llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -7379,7 +7379,6 @@
         Op2VK, TargetTransformInfo::OP_None, Op2VP, Operands, I);
   }
   case Instruction::FNeg: {
-    assert(!VF.isScalable() && "VF is assumed to be non scalable.");
     return TTI.getArithmeticInstrCost(
         I->getOpcode(), VectorTy, CostKind, TargetTransformInfo::OK_AnyValue,
         TargetTransformInfo::OK_AnyValue, TargetTransformInfo::OP_None,


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D98054.330250.patch
Type: text/x-patch
Size: 3410 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210312/90bb4f72/attachment.bin>


More information about the llvm-commits mailing list