Skip to content

Commit b1206e3

Browse files
IcohedronIcohedron
committed
Move scalarizer to before dxil-flatten-arrays
1 parent 3734f57 commit b1206e3

File tree

5 files changed

+160
-40
lines changed

5 files changed

+160
-40
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 & 29 deletions
Original file line numberDiff line numberDiff line change
@@ -8,30 +8,19 @@
88
@staticArrayOfVecData = internal global [3 x <4 x i32>] [<4 x i32> <i32 1, i32 2, i32 3, i32 4>, <4 x i32> <i32 5, i32 6, i32 7, i32 8>, <4 x i32> <i32 9, i32 10, i32 11, i32 12>], align 4
99
@"groupshared2dArrayofVectors" = local_unnamed_addr addrspace(3) global [3 x [3 x <4 x i32>]] zeroinitializer, align 16
1010

11-
; CHECK: @arrayofVecData.scalarized.1dim = local_unnamed_addr addrspace(3) global [8 x i32] zeroinitializer, align 16
12-
; CHECK: @vecData.scalarized = external addrspace(3) global [4 x i32], align 4
13-
; CHECK: @staticArrayOfVecData.scalarized.1dim = internal global [12 x i32] [i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12], align 4
14-
; CHECK: @groupshared2dArrayofVectors.scalarized.1dim = local_unnamed_addr addrspace(3) global [36 x i32] zeroinitializer, align 16
1511

16-
; CHECK-NOT: @arrayofVecData
17-
; CHECK-NOT: @arrayofVecData.scalarized
18-
; CHECK-NOT: @vecData
19-
; CHECK-NOT: @staticArrayOfVecData
20-
; CHECK-NOT: @staticArrayOfVecData.scalarized
21-
; CHECK-NOT: @groupshared2dArrayofVectors
22-
; CHECK-NOT: @groupshared2dArrayofVectors.scalarized
2312

2413
define <4 x i32> @load_array_vec_test() #0 {
2514
; CHECK-LABEL: define <4 x i32> @load_array_vec_test(
2615
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
2716
; 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
17+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 1), align 4
18+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 2), align 4
19+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 3), align 4
3120
; 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
21+
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
22+
; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
23+
; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
3524
; CHECK-NEXT: [[DOTI05:%.*]] = add i32 [[TMP2]], [[TMP10]]
3625
; CHECK-NEXT: [[DOTI16:%.*]] = add i32 [[TMP4]], [[TMP12]]
3726
; CHECK-NEXT: [[DOTI27:%.*]] = add i32 [[TMP6]], [[TMP14]]
@@ -53,9 +42,9 @@ define <4 x i32> @load_vec_test() #0 {
5342
; CHECK-LABEL: define <4 x i32> @load_vec_test(
5443
; CHECK-SAME: ) #[[ATTR0]] {
5544
; 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
45+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 1), align 4
46+
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 2), align 4
47+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([4 x i32], ptr addrspace(3) @vecData.scalarized, i32 0, i32 3), align 4
5948
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i32 0
6049
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[TMP2]], i32 1
6150
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[TMP3]], i32 2
@@ -66,18 +55,60 @@ define <4 x i32> @load_vec_test() #0 {
6655
ret <4 x i32> %1
6756
}
6857

58+
define <4 x i32> @load_vec_from_scalar_gep_test() #0 {
59+
; CHECK-LABEL: define <4 x i32> @load_vec_from_scalar_gep_test(
60+
; CHECK-SAME: ) #[[ATTR0]] {
61+
; 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
62+
; CHECK-NEXT: [[DOTI116:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
63+
; CHECK-NEXT: [[DOTI228:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
64+
; CHECK-NEXT: [[DOTI3310:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
65+
; CHECK-NEXT: [[DOTUPTO011:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI04]], i32 0
66+
; CHECK-NEXT: [[DOTUPTO112:%.*]] = insertelement <4 x i32> [[DOTUPTO011]], i32 [[DOTI116]], i32 1
67+
; CHECK-NEXT: [[DOTUPTO213:%.*]] = insertelement <4 x i32> [[DOTUPTO112]], i32 [[DOTI228]], i32 2
68+
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[DOTUPTO213]], i32 [[DOTI3310]], i32 3
69+
; CHECK-NEXT: ret <4 x i32> [[TMP1]]
70+
;
71+
%1 = getelementptr inbounds nuw i32, ptr addrspace(3) @"arrayofVecData", i32 4
72+
%2 = load <4 x i32>, ptr addrspace(3) %1, align 4
73+
ret <4 x i32> %2
74+
}
75+
76+
define <4 x i32> @load_vec_from_i8_gep_test() #0 {
77+
; CHECK-LABEL: define <4 x i32> @load_vec_from_i8_gep_test(
78+
; CHECK-SAME: ) #[[ATTR0]] {
79+
; 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
80+
; CHECK-NEXT: [[DOTI116:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 5), align 4
81+
; CHECK-NEXT: [[DOTI228:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 6), align 4
82+
; CHECK-NEXT: [[DOTI3310:%.*]] = load i32, ptr addrspace(3) getelementptr ([8 x i32], ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 0, i32 7), align 4
83+
; CHECK-NEXT: [[DOTUPTO011:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI04]], i32 0
84+
; CHECK-NEXT: [[DOTUPTO112:%.*]] = insertelement <4 x i32> [[DOTUPTO011]], i32 [[DOTI116]], i32 1
85+
; CHECK-NEXT: [[DOTUPTO213:%.*]] = insertelement <4 x i32> [[DOTUPTO112]], i32 [[DOTI228]], i32 2
86+
; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x i32> [[DOTUPTO213]], i32 [[DOTI3310]], i32 3
87+
; CHECK-NEXT: ret <4 x i32> [[TMP1]]
88+
;
89+
%1 = getelementptr inbounds nuw i8, ptr addrspace(3) @"arrayofVecData", i32 16
90+
%2 = load <4 x i32>, ptr addrspace(3) %1, align 4
91+
ret <4 x i32> %2
92+
}
93+
6994
define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
7095
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_test(
7196
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
7297
; CHECK-NEXT: [[TMP3:%.*]] = mul i32 [[INDEX]], 4
7398
; CHECK-NEXT: [[TMP2:%.*]] = add i32 0, [[TMP3]]
7499
; CHECK-NEXT: [[DOTFLAT:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP2]]
75100
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[DOTFLAT]], align 4
76-
; CHECK-NEXT: [[DOTFLAT_I1:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 1
101+
; CHECK-NEXT: [[TMP4:%.*]] = mul i32 [[INDEX]], 4
102+
; CHECK-NEXT: [[TMP5:%.*]] = add i32 1, [[TMP4]]
103+
; CHECK-NEXT: [[DOTFLAT_I1:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP5]]
77104
; CHECK-NEXT: [[DOTI1:%.*]] = load i32, ptr [[DOTFLAT_I1]], align 4
78-
; CHECK-NEXT: [[DOTFLAT_I2:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 2
105+
; CHECK-NEXT: [[TMP6:%.*]] = mul i32 [[INDEX]], 4
106+
; CHECK-NEXT: [[TMP7:%.*]] = add i32 2, [[TMP6]]
107+
; CHECK-NEXT: [[DOTFLAT_I2:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP7]]
79108
; CHECK-NEXT: [[DOTI2:%.*]] = load i32, ptr [[DOTFLAT_I2]], align 4
80-
; CHECK-NEXT: [[DOTFLAT_I3:%.*]] = getelementptr i32, ptr [[DOTFLAT]], i32 3
109+
; CHECK-NEXT: [[TMP8:%.*]] = mul i32 [[INDEX]], 4
110+
; CHECK-NEXT: [[TMP9:%.*]] = add i32 3, [[TMP8]]
111+
; CHECK-NEXT: [[DOTFLAT_I3:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP9]]
81112
; CHECK-NEXT: [[DOTI3:%.*]] = load i32, ptr [[DOTFLAT_I3]], align 4
82113
; CHECK-NEXT: [[DOTUPTO01:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
83114
; CHECK-NEXT: [[DOTUPTO12:%.*]] = insertelement <4 x i32> [[DOTUPTO01]], i32 [[DOTI1]], i32 1
@@ -90,17 +121,85 @@ define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
90121
ret <4 x i32> %4
91122
}
92123

124+
define <4 x i32> @load_static_array_of_vec_from_scalar_gep_test(i32 %index) #0 {
125+
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_from_scalar_gep_test(
126+
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
127+
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[INDEX]], 4
128+
; CHECK-NEXT: [[TMP2:%.*]] = mul i32 [[TMP1]], 1
129+
; CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP2]]
130+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP3]]
131+
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[TMP4]], align 4
132+
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP1]], 1
133+
; CHECK-NEXT: [[TMP6:%.*]] = add i32 1, [[TMP5]]
134+
; CHECK-NEXT: [[DOTI14:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP6]]
135+
; CHECK-NEXT: [[DOTI11:%.*]] = load i32, ptr [[DOTI14]], align 4
136+
; CHECK-NEXT: [[TMP7:%.*]] = mul i32 [[TMP1]], 1
137+
; CHECK-NEXT: [[TMP8:%.*]] = add i32 2, [[TMP7]]
138+
; CHECK-NEXT: [[DOTI25:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP8]]
139+
; CHECK-NEXT: [[DOTI22:%.*]] = load i32, ptr [[DOTI25]], align 4
140+
; CHECK-NEXT: [[TMP9:%.*]] = mul i32 [[TMP1]], 1
141+
; CHECK-NEXT: [[TMP10:%.*]] = add i32 3, [[TMP9]]
142+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP10]]
143+
; CHECK-NEXT: [[DOTI33:%.*]] = load i32, ptr [[DOTI36]], align 4
144+
; CHECK-NEXT: [[DOTUPTO07:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
145+
; CHECK-NEXT: [[DOTUPTO18:%.*]] = insertelement <4 x i32> [[DOTUPTO07]], i32 [[DOTI11]], i32 1
146+
; CHECK-NEXT: [[DOTUPTO29:%.*]] = insertelement <4 x i32> [[DOTUPTO18]], i32 [[DOTI22]], i32 2
147+
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[DOTUPTO29]], i32 [[DOTI33]], i32 3
148+
; CHECK-NEXT: ret <4 x i32> [[TMP11]]
149+
;
150+
%2 = mul i32 %index, 4
151+
%3 = getelementptr inbounds i32, ptr @staticArrayOfVecData, i32 %2
152+
%4 = load <4 x i32>, <4 x i32>* %3, align 4
153+
ret <4 x i32> %4
154+
}
155+
156+
define <4 x i32> @load_static_array_of_vec_from_i8_gep_test(i32 %index) #0 {
157+
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_from_i8_gep_test(
158+
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
159+
; CHECK-NEXT: [[TMP1:%.*]] = mul i32 [[INDEX]], 12
160+
; CHECK-NEXT: [[TMP2:%.*]] = mul i32 [[TMP1]], 1
161+
; CHECK-NEXT: [[TMP12:%.*]] = lshr i32 [[TMP2]], 2
162+
; CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP12]]
163+
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP3]]
164+
; CHECK-NEXT: [[DOTI0:%.*]] = load i32, ptr [[TMP4]], align 4
165+
; CHECK-NEXT: [[TMP5:%.*]] = mul i32 [[TMP1]], 1
166+
; CHECK-NEXT: [[TMP14:%.*]] = lshr i32 [[TMP5]], 2
167+
; CHECK-NEXT: [[TMP6:%.*]] = add i32 1, [[TMP14]]
168+
; CHECK-NEXT: [[DOTI14:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP6]]
169+
; CHECK-NEXT: [[DOTI11:%.*]] = load i32, ptr [[DOTI14]], align 4
170+
; CHECK-NEXT: [[TMP7:%.*]] = mul i32 [[TMP1]], 1
171+
; CHECK-NEXT: [[TMP15:%.*]] = lshr i32 [[TMP7]], 2
172+
; CHECK-NEXT: [[TMP8:%.*]] = add i32 2, [[TMP15]]
173+
; CHECK-NEXT: [[DOTI25:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP8]]
174+
; CHECK-NEXT: [[DOTI22:%.*]] = load i32, ptr [[DOTI25]], align 4
175+
; CHECK-NEXT: [[TMP9:%.*]] = mul i32 [[TMP1]], 1
176+
; CHECK-NEXT: [[TMP13:%.*]] = lshr i32 [[TMP9]], 2
177+
; CHECK-NEXT: [[TMP10:%.*]] = add i32 3, [[TMP13]]
178+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 0, i32 [[TMP10]]
179+
; CHECK-NEXT: [[DOTI33:%.*]] = load i32, ptr [[DOTI36]], align 4
180+
; CHECK-NEXT: [[DOTUPTO07:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI0]], i32 0
181+
; CHECK-NEXT: [[DOTUPTO18:%.*]] = insertelement <4 x i32> [[DOTUPTO07]], i32 [[DOTI11]], i32 1
182+
; CHECK-NEXT: [[DOTUPTO29:%.*]] = insertelement <4 x i32> [[DOTUPTO18]], i32 [[DOTI22]], i32 2
183+
; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x i32> [[DOTUPTO29]], i32 [[DOTI33]], i32 3
184+
; CHECK-NEXT: ret <4 x i32> [[TMP11]]
185+
;
186+
%2 = mul i32 %index, 12
187+
%3 = getelementptr inbounds i8, ptr @staticArrayOfVecData, i32 %2
188+
%4 = load <4 x i32>, <4 x i32>* %3, align 4
189+
ret <4 x i32> %4
190+
}
191+
93192
define <4 x i32> @multid_load_test() #0 {
94193
; CHECK-LABEL: define <4 x i32> @multid_load_test(
95194
; CHECK-SAME: ) #[[ATTR0]] {
96195
; 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
196+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 1), align 4
197+
; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 2), align 4
198+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 3), align 4
100199
; 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
200+
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 17), align 4
201+
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 18), align 4
202+
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) getelementptr ([36 x i32], ptr addrspace(3) @groupshared2dArrayofVectors.scalarized.1dim, i32 0, i32 19), align 4
104203
; CHECK-NEXT: [[DOTI08:%.*]] = add i32 [[TMP1]], [[TMP5]]
105204
; CHECK-NEXT: [[DOTI19:%.*]] = add i32 [[TMP2]], [[DOTI13]]
106205
; CHECK-NEXT: [[DOTI210:%.*]] = add i32 [[TMP3]], [[DOTI25]]

0 commit comments

Comments
 (0)