[llvm] 30ad742 - [ARM] Fix crash for gather of pointer costs.

[David Green via llvm-commits]

Author: David Green
Date: 2020-10-31T13:10:14Z
New Revision: 30ad7426442e6f029419366de8fe56eae529c00c

URL: https://github.com/llvm/llvm-project/commit/30ad7426442e6f029419366de8fe56eae529c00c
DIFF: https://github.com/llvm/llvm-project/commit/30ad7426442e6f029419366de8fe56eae529c00c.diff

LOG: [ARM] Fix crash for gather of pointer costs.

If the elt size is unknown due to it being a pointer, a comparison
against 0 will cause an assert. Make sure the elt size is large enough
before comparing and for the moment just return the scalar cost.

Added: 
    

Modified: 
    llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
    llvm/test/Analysis/CostModel/ARM/mve-gather-scatter-cost.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
index c26a77b46091..5455070f00f2 100644
--- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -1341,7 +1341,7 @@ unsigned ARMTTIImpl::getGatherScatterOpCost(unsigned Opcode, Type *DataTy,
   unsigned ScalarCost =
       NumElems * LT.first + BaseT::getScalarizationOverhead(VTy, {});
 
-  if (Alignment < EltSize / 8)
+  if (EltSize < 8 || Alignment < EltSize / 8)
     return ScalarCost;
 
   unsigned ExtSize = EltSize;

diff  --git a/llvm/test/Analysis/CostModel/ARM/mve-gather-scatter-cost.ll b/llvm/test/Analysis/CostModel/ARM/mve-gather-scatter-cost.ll
index 6d1e22701164..1ec8ea00e371 100644
--- a/llvm/test/Analysis/CostModel/ARM/mve-gather-scatter-cost.ll
+++ b/llvm/test/Analysis/CostModel/ARM/mve-gather-scatter-cost.ll
@@ -28,6 +28,7 @@ define i32 @masked_gather() {
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 136 for instruction: %V8I8 = call <8 x i8> @llvm.masked.gather.v8i8.v8p0i8(<8 x i8*> undef, i32 1, <8 x i1> undef, <8 x i8> undef)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V4I8 = call <4 x i8> @llvm.masked.gather.v4i8.v4p0i8(<4 x i8*> undef, i32 1, <4 x i1> undef, <4 x i8> undef)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I8 = call <2 x i8> @llvm.masked.gather.v2i8.v2p0i8(<2 x i8*> undef, i32 1, <2 x i1> undef, <2 x i8> undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 40 for instruction: %V4I32p = call <4 x i32*> @llvm.masked.gather.v4p0i32.v4p0p0i32(<4 x i32**> undef, i32 4, <4 x i1> undef, <4 x i32*> undef)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret i32 0
 ;
   %V4F64 = call <4 x double> @llvm.masked.gather.v4f64.v4p0f64(<4 x double*> undef, i32 4, <4 x i1> undef, <4 x double> undef)
@@ -62,6 +63,8 @@ define i32 @masked_gather() {
   %V4I8 = call <4 x i8> @llvm.masked.gather.v4i8.v4p0i8(<4 x i8*> undef, i32 1, <4 x i1> undef, <4 x i8> undef)
   %V2I8 = call <2 x i8> @llvm.masked.gather.v2i8.v2p0i8(<2 x i8*> undef, i32 1, <2 x i1> undef, <2 x i8> undef)
 
+  %V4I32p = call <4 x i32*> @llvm.masked.gather.v4p0i32.v4p0p0i32(<4 x i32**> undef, i32 4, <4 x i1> undef, <4 x i32*> undef)
+
   ret i32 0
 }
 
@@ -591,6 +594,7 @@ declare <16 x i32> @llvm.masked.gather.v16i32.v16p0i32(<16 x i32*>, i32, <16 x i
 declare <8 x i32> @llvm.masked.gather.v8i32.v8p0i32(<8 x i32*>, i32, <8 x i1>, <8 x i32>)
 declare <4 x i32> @llvm.masked.gather.v4i32.v4p0i32(<4 x i32*>, i32, <4 x i1>, <4 x i32>)
 declare <2 x i32> @llvm.masked.gather.v2i32.v2p0i32(<2 x i32*>, i32, <2 x i1>, <2 x i32>)
+declare <4 x i32*> @llvm.masked.gather.v4p0i32.v4p0p0i32(<4 x i32**>, i32, <4 x i1>, <4 x i32*>)
 
 declare <16 x i16> @llvm.masked.gather.v16i16.v16p0i16(<16 x i16*>, i32, <16 x i1>, <16 x i16>)
 declare <8 x i16> @llvm.masked.gather.v8i16.v8p0i16(<8 x i16*>, i32, <8 x i1>, <8 x i16>)