[llvm] r267597 - Masked Store in Loop Vectorizer - bugfix
Elena Demikhovsky via llvm-commits
llvm-commits at lists.llvm.org
Tue Apr 26 13:18:05 PDT 2016
Author: delena
Date: Tue Apr 26 15:18:04 2016
New Revision: 267597
URL: http://llvm.org/viewvc/llvm-project?rev=267597&view=rev
Log:
Masked Store in Loop Vectorizer - bugfix
Fixed a bug in loop vectorization with conditional store.
Differential Revision: http://reviews.llvm.org/D19532
Added:
llvm/trunk/test/Transforms/LoopVectorize/X86/max-mstore.ll
Modified:
llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp
Modified: llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp?rev=267597&r1=267596&r2=267597&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp (original)
+++ llvm/trunk/lib/Transforms/Vectorize/LoopVectorize.cpp Tue Apr 26 15:18:04 2016
@@ -4909,24 +4909,20 @@ bool LoopVectorizationLegality::blockCan
if (!SI)
return false;
+ // Build a masked store if it is legal for the target.
+ if (isLegalMaskedStore(SI->getValueOperand()->getType(),
+ SI->getPointerOperand()) ||
+ isLegalMaskedScatter(SI->getValueOperand()->getType())) {
+ MaskedOp.insert(SI);
+ continue;
+ }
+
bool isSafePtr = (SafePtrs.count(SI->getPointerOperand()) != 0);
bool isSinglePredecessor = SI->getParent()->getSinglePredecessor();
if (++NumPredStores > NumberOfStoresToPredicate || !isSafePtr ||
- !isSinglePredecessor) {
- // Build a masked store if it is legal for the target, otherwise
- // scalarize the block.
- bool isLegalMaskedOp =
- isLegalMaskedStore(SI->getValueOperand()->getType(),
- SI->getPointerOperand()) ||
- isLegalMaskedScatter(SI->getValueOperand()->getType());
- if (isLegalMaskedOp) {
- --NumPredStores;
- MaskedOp.insert(SI);
- continue;
- }
+ !isSinglePredecessor)
return false;
- }
}
if (it->mayThrow())
return false;
Added: llvm/trunk/test/Transforms/LoopVectorize/X86/max-mstore.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoopVectorize/X86/max-mstore.ll?rev=267597&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LoopVectorize/X86/max-mstore.ll (added)
+++ llvm/trunk/test/Transforms/LoopVectorize/X86/max-mstore.ll Tue Apr 26 15:18:04 2016
@@ -0,0 +1,46 @@
+; RUN: opt -basicaa -loop-vectorize -force-vector-interleave=1 -S -mcpu=core-avx2
+
+target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+target triple = "x86_64-unknown-linux-gnu"
+
+ at b = common global [256 x i32] zeroinitializer, align 16
+ at a = common global [256 x i32] zeroinitializer, align 16
+
+; unsigned int a[256], b[256];
+; void foo() {
+; for (i = 0; i < 256; i++) {
+; if (b[i] > a[i])
+; a[i] = b[i];
+; }
+; }
+
+; CHECK-LABEL: foo
+; CHECK: load <8 x i32>
+; CHECK: icmp ugt <8 x i32>
+; CHECK: masked.store
+
+define void @foo() {
+entry:
+ br label %for.body
+
+for.body: ; preds = %for.inc, %entry
+ %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.inc ]
+ %arrayidx = getelementptr inbounds [256 x i32], [256 x i32]* @b, i64 0, i64 %indvars.iv
+ %0 = load i32, i32* %arrayidx, align 4
+ %arrayidx2 = getelementptr inbounds [256 x i32], [256 x i32]* @a, i64 0, i64 %indvars.iv
+ %1 = load i32, i32* %arrayidx2, align 4
+ %cmp3 = icmp ugt i32 %0, %1
+ br i1 %cmp3, label %if.then, label %for.inc
+
+if.then: ; preds = %for.body
+ store i32 %0, i32* %arrayidx2, align 4
+ br label %for.inc
+
+for.inc: ; preds = %for.body, %if.then
+ %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+ %exitcond = icmp eq i64 %indvars.iv.next, 256
+ br i1 %exitcond, label %for.end, label %for.body
+
+for.end: ; preds = %for.inc
+ ret void
+}
More information about the llvm-commits
mailing list