diff options
| author | Daniel Neilson <dneilson@azul.com> | 2018-05-01 15:35:08 +0000 |
|---|---|---|
| committer | Daniel Neilson <dneilson@azul.com> | 2018-05-01 15:35:08 +0000 |
| commit | 9e4bbe801a79030934627ceafca79c75f45eb58a (patch) | |
| tree | 007da645eaf7badfa0b4afe87e1c79c05347a22f /llvm/test/Transforms/LoopVectorize/ARM | |
| parent | 969d63ea8c5820d1325ad27205fc34c4f5030a54 (diff) | |
| download | bcm5719-llvm-9e4bbe801a79030934627ceafca79c75f45eb58a.tar.gz bcm5719-llvm-9e4bbe801a79030934627ceafca79c75f45eb58a.zip | |
[LV] Preserve inbounds on created GEPs
Summary:
This is a fix for PR23997.
The loop vectorizer is not preserving the inbounds property of GEPs that it creates.
This is inhibiting some optimizations. This patch preserves the inbounds property in
the case where a load/store is being fed by an inbounds GEP.
Reviewers: mkuper, javed.absar, hsaito
Reviewed By: hsaito
Subscribers: dcaballe, hsaito, llvm-commits
Differential Revision: https://reviews.llvm.org/D46191
llvm-svn: 331269
Diffstat (limited to 'llvm/test/Transforms/LoopVectorize/ARM')
| -rw-r--r-- | llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll b/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll index 19ed23b01c1..a1cf4b318f3 100644 --- a/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll +++ b/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll @@ -49,18 +49,18 @@ define i32 @test(float* nocapture readonly %x) { ; CHECK-NEXT: [[INDUCTION:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1> ; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[INDEX]], 0 ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, float* [[X:%.*]], i32 [[TMP1]] -; CHECK-NEXT: [[TMP3:%.*]] = getelementptr float, float* [[TMP2]], i32 0 +; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, float* [[TMP2]], i32 0 ; CHECK-NEXT: [[TMP4:%.*]] = bitcast float* [[TMP3]] to <2 x float>* ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, <2 x float>* [[TMP4]], align 4 ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds float, float* [[T4]], i32 [[TMP1]] -; CHECK-NEXT: [[TMP6:%.*]] = getelementptr float, float* [[TMP5]], i32 0 +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, float* [[TMP5]], i32 0 ; CHECK-NEXT: [[TMP7:%.*]] = bitcast float* [[TMP6]] to <2 x float>* ; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <2 x float>, <2 x float>* [[TMP7]], align 4 ; CHECK-NEXT: [[TMP8:%.*]] = fsub fast <2 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]] ; CHECK-NEXT: [[TMP9:%.*]] = fpext <2 x float> [[TMP8]] to <2 x double> ; CHECK-NEXT: [[TMP10:%.*]] = fmul fast <2 x double> [[TMP9]], [[TMP9]] ; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, float* [[T6]], i32 [[TMP1]] -; CHECK-NEXT: [[TMP12:%.*]] = getelementptr float, float* [[TMP11]], i32 0 +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds float, float* [[TMP11]], i32 0 ; CHECK-NEXT: [[TMP13:%.*]] = bitcast float* [[TMP12]] to <2 x float>* ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <2 x float>, <2 x float>* [[TMP13]], align 4 ; CHECK-NEXT: [[TMP14:%.*]] = fpext <2 x float> [[WIDE_LOAD2]] to <2 x double> |
