Skip to content

Commit 27b3b4a

Browse files
IcohedronIcohedron
authored
[DirectX] Move the scalarizer pass to before dxil-flatten-arrays (#146800)
Fixes #145924 and #140416 Depends on #146173 being merged first. This PR moves the scalarizer pass to immediately before the dxil-flatten-arrays pass to allow the dxil-flatten-arrays pass to turn scalar GEPs (including i8 GEPs) into flattened array GEPs where applicable. A number of LLVM DirectX CodeGen tests have been edited to remove scalar GEPs and also correct previously uncaught incorrectly-transformed GEPs. No more validation errors of the form `Access to out-of-bounds memory is disallowed` or `TGSM pointers must originate from an unambiguous TGSM global variable` appear anymore after this PR when compiling DML shaders.
1 parent 7ba0c98 commit 27b3b4a

File tree

5 files changed

+160
-29
lines changed

5 files changed

+160
-29
lines changed

‎llvm/lib/Target/DirectX/DXILDataScalarization.cpp

Lines changed: 6 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -302,7 +302,7 @@ bool DataScalarizerVisitor::visitExtractElementInst(ExtractElementInst &EEI) {
302302

303303
bool DataScalarizerVisitor::visitGetElementPtrInst(GetElementPtrInst &GEPI) {
304304
Value *PtrOperand = GEPI.getPointerOperand();
305-
Type *OrigGEPType = GEPI.getPointerOperandType();
305+
Type *OrigGEPType = GEPI.getSourceElementType();
306306
Type *NewGEPType = OrigGEPType;
307307
bool NeedsTransform = false;
308308

@@ -319,6 +319,11 @@ bool DataScalarizerVisitor::visitGetElementPtrInst(GetElementPtrInst &GEPI) {
319319
}
320320
}
321321

322+
// Scalar geps should remain scalars geps. The dxil-flatten-arrays pass will
323+
// convert these scalar geps into flattened array geps
324+
if (!isa<ArrayType>(OrigGEPType))
325+
NewGEPType = OrigGEPType;
326+
322327
// Note: We bail if this isn't a gep touched via alloca or global
323328
// transformations
324329
if (!NeedsTransform)

‎llvm/lib/Target/DirectX/DirectXTargetMachine.cpp

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -107,10 +107,10 @@ class DirectXPassConfig : public TargetPassConfig {
107107
addPass(createDXILIntrinsicExpansionLegacyPass());
108108
addPass(createDXILCBufferAccessLegacyPass());
109109
addPass(createDXILDataScalarizationLegacyPass());
110-
addPass(createDXILFlattenArraysLegacyPass());
111110
ScalarizerPassOptions DxilScalarOptions;
112111
DxilScalarOptions.ScalarizeLoadStore = true;
113112
addPass(createScalarizerPass(DxilScalarOptions));
113+
addPass(createDXILFlattenArraysLegacyPass());
114114
addPass(createDXILForwardHandleAccessesLegacyPass());
115115
addPass(createDXILLegalizeLegacyPass());
116116
addPass(createDXILResourceImplicitBindingLegacyPass());

‎llvm/test/CodeGen/DirectX/llc-pipeline.ll

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -19,10 +19,12 @@
1919
; CHECK-NEXT: DXIL Intrinsic Expansion
2020
; CHECK-NEXT: DXIL CBuffer Access
2121
; CHECK-NEXT: DXIL Data Scalarization
22-
; CHECK-NEXT: DXIL Array Flattener
2322
; CHECK-NEXT: FunctionPass Manager
2423
; CHECK-NEXT: Dominator Tree Construction
2524
; CHECK-NEXT: Scalarize vector operations
25+
; CHECK-NEXT: DXIL Array Flattener
26+
; CHECK-NEXT: FunctionPass Manager
27+
; CHECK-NEXT: Dominator Tree Construction
2628
; CHECK-NEXT: DXIL Forward Handle Accesses
2729
; CHECK-NEXT: DXIL Legalizer
2830
; CHECK-NEXT: DXIL Resource Binding Analysis

‎llvm/test/CodeGen/DirectX/llc-vector-load-scalarize.ll

Lines changed: 128 additions & 18 deletions
Original file line numberDiff line numberDiff line change
@@ -25,13 +25,13 @@ define <4 x i32> @load_array_vec_test() #0 {
2525
; CHECK-LABEL: define <4 x i32> @load_array_vec_test(
2626
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
2727
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, align 4
28-
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 1), align 4
29-
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 2), align 4
30-
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 3), align 4
28+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 1), align 4
29+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 2), align 4
30+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 3), align 4
3131
; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), align 4
32-
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), i32 1), align 4
33-
; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), i32 2), align 4
34-
; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), i32 3), align 4
32+
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
33+
; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
34+
; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
3535
; CHECK-NEXT: [[DOTI05:%.*]] = add i32 [[TMP2]], [[TMP10]]
3636
; CHECK-NEXT: [[DOTI16:%.*]] = add i32 [[TMP4]], [[TMP12]]
3737
; CHECK-NEXT: [[DOTI27:%.*]] = add i32 [[TMP6]], [[TMP14]]
@@ -53,9 +53,9 @@ define <4 x i32> @load_vec_test() #0 {
5353
; CHECK-LABEL: define <4 x i32> @load_vec_test(
5454
; CHECK-SAME: ) #[[ATTR0]] {
5555
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(3) @vecData.scalarized, align 4
56-
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 1), align 4
57-
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 2), align 4
58-
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 3), align 4
56+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 1), align 4
57+
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 2), align 4
58+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 3), align 4
5959
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0
6060
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[TMP2]], i32 1
6161
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[TMP3]], i32 2
@@ -66,18 +66,60 @@ define <4 x i32> @load_vec_test() #0 {
6666
ret <4 x i32> %1
6767
}
6868

69+
define <4 x i32> @load_vec_from_scalar_gep_test() #0 {
70+
; CHECK-LABEL: define <4 x i32> @load_vec_from_scalar_gep_test(
71+
; CHECK-SAME: ) #[[ATTR0]] {
72+
; CHECK-NEXT: [[DOTI04:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds nuw ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), align 4
73+
; CHECK-NEXT: [[DOTI116:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
74+
; CHECK-NEXT: [[DOTI228:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
75+
; CHECK-NEXT: [[DOTI3310:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
76+
; CHECK-NEXT: [[DOTUPTO011:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI04]], i32 0
77+
; CHECK-NEXT: [[DOTUPTO112:%.*]] = insertelement <4 x i32> [[DOTUPTO011]], i32 [[DOTI116]], i32 1
78+
; CHECK-NEXT: [[DOTUPTO213:%.*]] = insertelement <4 x i32> [[DOTUPTO112]], i32 [[DOTI228]], i32 2
79+
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[DOTUPTO213]], i32 [[DOTI3310]], i32 3
80+
; CHECK-NEXT: ret <4 x i32> [[TMP1]]
81+
;
82+
%1 = getelementptr inbounds nuw i32, ptr addrspace(3) @"arrayofVecData", i32 4
83+
%2 = load <4 x i32>, ptr addrspace(3) %1, align 4
84+
ret <4 x i32> %2
85+
}
86+
87+
define <4 x i32> @load_vec_from_i8_gep_test() #0 {
88+
; CHECK-LABEL: define <4 x i32> @load_vec_from_i8_gep_test(
89+
; CHECK-SAME: ) #[[ATTR0]] {
90+
; CHECK-NEXT: [[DOTI04:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds nuw ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 4), align 4
91+
; CHECK-NEXT: [[DOTI116:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
92+
; CHECK-NEXT: [[DOTI228:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
93+
; CHECK-NEXT: [[DOTI3310:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
94+
; CHECK-NEXT: [[DOTUPTO011:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI04]], i32 0
95+
; CHECK-NEXT: [[DOTUPTO112:%.*]] = insertelement <4 x i32> [[DOTUPTO011]], i32 [[DOTI116]], i32 1
96+
; CHECK-NEXT: [[DOTUPTO213:%.*]] = insertelement <4 x i32> [[DOTUPTO112]], i32 [[DOTI228]], i32 2
97+
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[DOTUPTO213]], i32 [[DOTI3310]], i32 3
98+
; CHECK-NEXT: ret <4 x i32> [[TMP1]]
99+
;
100+
%1 = getelementptr inbounds nuw i8, ptr addrspace(3) @"arrayofVecData", i32 16
101+
%2 = load <4 x i32>, ptr addrspace(3) %1, align 4
102+
ret <4 x i32> %2
103+
}
104+
69105
define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
70106
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_test(
71107
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
72108
; CHECK-NEXT: [[TMP3:%.*]] = mul i32 [[INDEX]], 4
73109
; CHECK-NEXT: [[TMP2:%.*]] = add i32 0, [[TMP3]]
74110
; CHECK-NEXT: [[DOTFLAT:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP2]]
75111
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[DOTFLAT]], align 4
76-
; CHECK-NEXT: [[DOTFLAT_I1:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 1
112+
; CHECK-NEXT: [[TMP4:%.*]] = mul i32 [[INDEX]], 4
113+
; CHECK-NEXT: [[TMP5:%.*]] = add i32 1, [[TMP4]]
114+
; CHECK-NEXT: [[DOTFLAT_I1:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP5]]
77115
; CHECK-NEXT: [[DOTI1:%.*]] = load i32, ptr [[DOTFLAT_I1]], align 4
78-
; CHECK-NEXT: [[DOTFLAT_I2:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 2
116+
; CHECK-NEXT: [[TMP6:%.*]] = mul i32 [[INDEX]], 4
117+
; CHECK-NEXT: [[TMP7:%.*]] = add i32 2, [[TMP6]]
118+
; CHECK-NEXT: [[DOTFLAT_I2:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP7]]
79119
; CHECK-NEXT: [[DOTI2:%.*]] = load i32, ptr [[DOTFLAT_I2]], align 4
80-
; CHECK-NEXT: [[DOTFLAT_I3:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 3
120+
; CHECK-NEXT: [[TMP8:%.*]] = mul i32 [[INDEX]], 4
121+
; CHECK-NEXT: [[TMP9:%.*]] = add i32 3, [[TMP8]]
122+
; CHECK-NEXT: [[DOTFLAT_I3:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP9]]
81123
; CHECK-NEXT: [[DOTI3:%.*]] = load i32, ptr [[DOTFLAT_I3]], align 4
82124
; CHECK-NEXT: [[DOTUPTO01:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
83125
; CHECK-NEXT: [[DOTUPTO12:%.*]] = insertelement <4 x i32> [[DOTUPTO01]], i32 [[DOTI1]], i32 1
@@ -90,17 +132,85 @@ define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
90132
ret <4 x i32> %4
91133
}
92134

135+
define <4 x i32> @load_static_array_of_vec_from_scalar_gep_test(i32 %index) #0 {
136+
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_from_scalar_gep_test(
137+
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
138+
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[INDEX]], 4
139+
; CHECK-NEXT: [[TMP2:%.*]] = mul i32 [[TMP1]], 1
140+
; CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP2]]
141+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP3]]
142+
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[TMP4]], align 4
143+
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP1]], 1
144+
; CHECK-NEXT: [[TMP6:%.*]] = add i32 1, [[TMP5]]
145+
; CHECK-NEXT: [[DOTI14:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP6]]
146+
; CHECK-NEXT: [[DOTI11:%.*]] = load i32, ptr [[DOTI14]], align 4
147+
; CHECK-NEXT: [[TMP7:%.*]] = mul i32 [[TMP1]], 1
148+
; CHECK-NEXT: [[TMP8:%.*]] = add i32 2, [[TMP7]]
149+
; CHECK-NEXT: [[DOTI25:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP8]]
150+
; CHECK-NEXT: [[DOTI22:%.*]] = load i32, ptr [[DOTI25]], align 4
151+
; CHECK-NEXT: [[TMP9:%.*]] = mul i32 [[TMP1]], 1
152+
; CHECK-NEXT: [[TMP10:%.*]] = add i32 3, [[TMP9]]
153+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP10]]
154+
; CHECK-NEXT: [[DOTI33:%.*]] = load i32, ptr [[DOTI36]], align 4
155+
; CHECK-NEXT: [[DOTUPTO07:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
156+
; CHECK-NEXT: [[DOTUPTO18:%.*]] = insertelement <4 x i32> [[DOTUPTO07]], i32 [[DOTI11]], i32 1
157+
; CHECK-NEXT: [[DOTUPTO29:%.*]] = insertelement <4 x i32> [[DOTUPTO18]], i32 [[DOTI22]], i32 2
158+
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[DOTUPTO29]], i32 [[DOTI33]], i32 3
159+
; CHECK-NEXT: ret <4 x i32> [[TMP11]]
160+
;
161+
%2 = mul i32 %index, 4
162+
%3 = getelementptr inbounds i32, ptr @staticArrayOfVecData, i32 %2
163+
%4 = load <4 x i32>, <4 x i32>* %3, align 4
164+
ret <4 x i32> %4
165+
}
166+
167+
define <4 x i32> @load_static_array_of_vec_from_i8_gep_test(i32 %index) #0 {
168+
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_from_i8_gep_test(
169+
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
170+
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[INDEX]], 12
171+
; CHECK-NEXT: [[TMP2:%.*]] = mul i32 [[TMP1]], 1
172+
; CHECK-NEXT: [[TMP12:%.*]] = lshr i32 [[TMP2]], 2
173+
; CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP12]]
174+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP3]]
175+
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[TMP4]], align 4
176+
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP1]], 1
177+
; CHECK-NEXT: [[TMP14:%.*]] = lshr i32 [[TMP5]], 2
178+
; CHECK-NEXT: [[TMP6:%.*]] = add i32 1, [[TMP14]]
179+
; CHECK-NEXT: [[DOTI14:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP6]]
180+
; CHECK-NEXT: [[DOTI11:%.*]] = load i32, ptr [[DOTI14]], align 4
181+
; CHECK-NEXT: [[TMP7:%.*]] = mul i32 [[TMP1]], 1
182+
; CHECK-NEXT: [[TMP15:%.*]] = lshr i32 [[TMP7]], 2
183+
; CHECK-NEXT: [[TMP8:%.*]] = add i32 2, [[TMP15]]
184+
; CHECK-NEXT: [[DOTI25:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP8]]
185+
; CHECK-NEXT: [[DOTI22:%.*]] = load i32, ptr [[DOTI25]], align 4
186+
; CHECK-NEXT: [[TMP9:%.*]] = mul i32 [[TMP1]], 1
187+
; CHECK-NEXT: [[TMP13:%.*]] = lshr i32 [[TMP9]], 2
188+
; CHECK-NEXT: [[TMP10:%.*]] = add i32 3, [[TMP13]]
189+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP10]]
190+
; CHECK-NEXT: [[DOTI33:%.*]] = load i32, ptr [[DOTI36]], align 4
191+
; CHECK-NEXT: [[DOTUPTO07:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
192+
; CHECK-NEXT: [[DOTUPTO18:%.*]] = insertelement <4 x i32> [[DOTUPTO07]], i32 [[DOTI11]], i32 1
193+
; CHECK-NEXT: [[DOTUPTO29:%.*]] = insertelement <4 x i32> [[DOTUPTO18]], i32 [[DOTI22]], i32 2
194+
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[DOTUPTO29]], i32 [[DOTI33]], i32 3
195+
; CHECK-NEXT: ret <4 x i32> [[TMP11]]
196+
;
197+
%2 = mul i32 %index, 12
198+
%3 = getelementptr inbounds i8, ptr @staticArrayOfVecData, i32 %2
199+
%4 = load <4 x i32>, <4 x i32>* %3, align 4
200+
ret <4 x i32> %4
201+
}
202+
93203
define <4 x i32> @multid_load_test() #0 {
94204
; CHECK-LABEL: define <4 x i32> @multid_load_test(
95205
; CHECK-SAME: ) #[[ATTR0]] {
96206
; CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, align 4
97-
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 1), align 4
98-
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 2), align 4
99-
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 3), align 4
207+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), align 4
208+
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 2), align 4
209+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 3), align 4
100210
; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 16), align 4
101-
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 16), i32 1), align 4
102-
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 16), i32 2), align 4
103-
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) getelementptr (i32, ptr addrspace(3) getelementptr inbounds ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 16), i32 3), align 4
211+
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 17), align 4
212+
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 18), align 4
213+
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 19), align 4
104214
; CHECK-NEXT: [[DOTI08:%.*]] = add i32 [[TMP1]], [[TMP5]]
105215
; CHECK-NEXT: [[DOTI19:%.*]] = add i32 [[TMP2]], [[DOTI13]]
106216
; CHECK-NEXT: [[DOTI210:%.*]] = add i32 [[TMP3]], [[DOTI25]]

0 commit comments

Comments
 (0)