Skip to content

Commit 4fffccc

Browse files
farzonlfarzonl
committed
Address pr comments on test seperation.
1 parent f45ade6 commit 4fffccc

File tree

6 files changed

+199
-36
lines changed

6 files changed

+199
-36
lines changed

‎llvm/lib/Target/DirectX/DXILFlattenArrays.cpp

Lines changed: 2 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -307,12 +307,8 @@ bool DXILFlattenArraysVisitor::visit(Function &F) {
307307
bool MadeChange = false;
308308
ReversePostOrderTraversal<Function *> RPOT(&F);
309309
for (BasicBlock *BB : make_early_inc_range(RPOT)) {
310-
for (Instruction &I : make_early_inc_range(*BB)) {
311-
if (InstVisitor::visit(I) && I.getType()->isVoidTy()) {
312-
I.eraseFromParent();
313-
MadeChange = true;
314-
}
315-
}
310+
for (Instruction &I : make_early_inc_range(*BB))
311+
MadeChange |= InstVisitor::visit(I);
316312
}
317313
finish();
318314
return MadeChange;

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

Lines changed: 153 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,153 @@
1+
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --tool llc --version 5
2+
; RUN: llc %s -mtriple=dxil-pc-shadermodel6.3-library --filetype=asm -o - | FileCheck %s
3+
4+
; Make sure we can load groupshared, static vectors and arrays of vectors
5+
6+
@"arrayofVecData" = local_unnamed_addr addrspace(3) global [2 x <3 x float>] zeroinitializer, align 16
7+
@"vecData" = external addrspace(3) global <4 x i32>, align 4
8+
@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
9+
@"groushared2dArrayofVectors" = local_unnamed_addr addrspace(3) global [3 x [ 3 x <4 x i32>]] zeroinitializer, align 16
10+
11+
; CHECK: @arrayofVecData.scalarized.1dim = local_unnamed_addr addrspace(3) global [6 x float] 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: @groushared2dArrayofVectors.scalarized.1dim = local_unnamed_addr addrspace(3) global [36 x i32] zeroinitializer, align 16
15+
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: @groushared2dArrayofVectors
22+
; CHECK-NOT: @groushared2dArrayofVectors.scalarized
23+
24+
25+
define <4 x i32> @load_array_vec_test() #0 {
26+
; CHECK-LABEL: define <4 x i32> @load_array_vec_test(
27+
; CHECK-SAME: ) #[[ATTR0:[0-9]+]] {
28+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr addrspace(3) @arrayofVecData.scalarized.1dim to ptr addrspace(3)
29+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) [[TMP1]], align 4
30+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 1) to ptr addrspace(3)
31+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) [[TMP3]], align 4
32+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 2) to ptr addrspace(3)
33+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) [[TMP5]], align 4
34+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @arrayofVecData.scalarized.1dim, i32 3) to ptr addrspace(3)
35+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) [[TMP7]], align 4
36+
; CHECK-NEXT: [[TMP9:%.*]] = bitcast ptr addrspace(3) @arrayofVecData.scalarized.1dim to ptr addrspace(3)
37+
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr [2 x [3 x float]], ptr addrspace(3) [[TMP9]], i32 0, i32 1
38+
; CHECK-NEXT: [[TMP11:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
39+
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) [[TMP11]], align 4
40+
; CHECK-NEXT: [[TMP13:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
41+
; CHECK-NEXT: [[DOTI12:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP13]], i32 1
42+
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) [[DOTI12]], align 4
43+
; CHECK-NEXT: [[TMP14:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
44+
; CHECK-NEXT: [[DOTI24:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP14]], i32 2
45+
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) [[DOTI24]], align 4
46+
; CHECK-NEXT: [[TMP15:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
47+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP15]], i32 3
48+
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) [[DOTI36]], align 4
49+
; CHECK-NEXT: [[DOTI08:%.*]] = add i32 [[TMP2]], [[TMP12]]
50+
; CHECK-NEXT: [[DOTI19:%.*]] = add i32 [[TMP4]], [[DOTI13]]
51+
; CHECK-NEXT: [[DOTI210:%.*]] = add i32 [[TMP6]], [[DOTI25]]
52+
; CHECK-NEXT: [[DOTI311:%.*]] = add i32 [[TMP8]], [[DOTI37]]
53+
; CHECK-NEXT: [[DOTUPTO015:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI08]], i64 0
54+
; CHECK-NEXT: [[DOTUPTO116:%.*]] = insertelement <4 x i32> [[DOTUPTO015]], i32 [[DOTI19]], i64 1
55+
; CHECK-NEXT: [[DOTUPTO217:%.*]] = insertelement <4 x i32> [[DOTUPTO116]], i32 [[DOTI210]], i64 2
56+
; CHECK-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> [[DOTUPTO217]], i32 [[DOTI311]], i64 3
57+
; CHECK-NEXT: ret <4 x i32> [[TMP16]]
58+
;
59+
%1 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([2 x <4 x i32>], [2 x <4 x i32>] addrspace(3)* @"arrayofVecData", i32 0, i32 0), align 4
60+
%2 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([2 x <4 x i32>], [2 x <4 x i32>] addrspace(3)* @"arrayofVecData", i32 0, i32 1), align 4
61+
%3 = add <4 x i32> %1, %2
62+
ret <4 x i32> %3
63+
}
64+
65+
66+
define <4 x i32> @load_vec_test() #0 {
67+
; CHECK-LABEL: define <4 x i32> @load_vec_test(
68+
; CHECK-SAME: ) #[[ATTR0]] {
69+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr addrspace(3) @vecData.scalarized to ptr addrspace(3)
70+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) [[TMP1]], align 4
71+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 1) to ptr addrspace(3)
72+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) [[TMP3]], align 4
73+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 2) to ptr addrspace(3)
74+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) [[TMP5]], align 4
75+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @vecData.scalarized, i32 3) to ptr addrspace(3)
76+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) [[TMP7]], align 4
77+
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 0
78+
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[TMP4]], i64 1
79+
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[TMP6]], i64 2
80+
; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x i32> [[DOTUPTO2]], i32 [[TMP8]], i64 3
81+
; CHECK-NEXT: ret <4 x i32> [[TMP9]]
82+
;
83+
%1 = load <4 x i32>, <4 x i32> addrspace(3)* @"vecData", align 4
84+
ret <4 x i32> %1
85+
}
86+
87+
define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
88+
; CHECK-LABEL: define <4 x i32> @load_static_array_of_vec_test(
89+
; CHECK-SAME: i32 [[INDEX:%.*]]) #[[ATTR0]] {
90+
; CHECK-NEXT: [[DOTFLAT:%.*]] = getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 [[INDEX]]
91+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr [[DOTFLAT]] to ptr
92+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4
93+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr [[DOTFLAT]] to ptr
94+
; CHECK-NEXT: [[DOTFLAT_I1:%.*]] = getelementptr i32, ptr [[TMP3]], i32 1
95+
; CHECK-NEXT: [[DOTI1:%.*]] = load i32, ptr [[DOTFLAT_I1]], align 4
96+
; CHECK-NEXT: [[TMP4:%.*]] = bitcast ptr [[DOTFLAT]] to ptr
97+
; CHECK-NEXT: [[DOTFLAT_I2:%.*]] = getelementptr i32, ptr [[TMP4]], i32 2
98+
; CHECK-NEXT: [[DOTI2:%.*]] = load i32, ptr [[DOTFLAT_I2]], align 4
99+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr [[DOTFLAT]] to ptr
100+
; CHECK-NEXT: [[DOTFLAT_I3:%.*]] = getelementptr i32, ptr [[TMP5]], i32 3
101+
; CHECK-NEXT: [[DOTI3:%.*]] = load i32, ptr [[DOTFLAT_I3]], align 4
102+
; CHECK-NEXT: [[DOTUPTO0:%.*]] = insertelement <4 x i32> poison, i32 [[TMP2]], i64 0
103+
; CHECK-NEXT: [[DOTUPTO1:%.*]] = insertelement <4 x i32> [[DOTUPTO0]], i32 [[DOTI1]], i64 1
104+
; CHECK-NEXT: [[DOTUPTO2:%.*]] = insertelement <4 x i32> [[DOTUPTO1]], i32 [[DOTI2]], i64 2
105+
; CHECK-NEXT: [[TMP6:%.*]] = insertelement <4 x i32> [[DOTUPTO2]], i32 [[DOTI3]], i64 3
106+
; CHECK-NEXT: ret <4 x i32> [[TMP6]]
107+
;
108+
%3 = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* @staticArrayOfVecData, i32 0, i32 %index
109+
%4 = load <4 x i32>, <4 x i32>* %3, align 4
110+
ret <4 x i32> %4
111+
}
112+
113+
define <4 x i32> @multid_load_test() #0 {
114+
; CHECK-LABEL: define <4 x i32> @multid_load_test(
115+
; CHECK-SAME: ) #[[ATTR0]] {
116+
; CHECK-NEXT: [[TMP1:%.*]] = bitcast ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim to ptr addrspace(3)
117+
; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(3) [[TMP1]], align 4
118+
; CHECK-NEXT: [[TMP3:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 1) to ptr addrspace(3)
119+
; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(3) [[TMP3]], align 4
120+
; CHECK-NEXT: [[TMP5:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 2) to ptr addrspace(3)
121+
; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr addrspace(3) [[TMP5]], align 4
122+
; CHECK-NEXT: [[TMP7:%.*]] = bitcast ptr addrspace(3) getelementptr (i32, ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim, i32 3) to ptr addrspace(3)
123+
; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr addrspace(3) [[TMP7]], align 4
124+
; CHECK-NEXT: [[TMP9:%.*]] = bitcast ptr addrspace(3) @groushared2dArrayofVectors.scalarized.1dim to ptr addrspace(3)
125+
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr [3 x [3 x [4 x i32]]], ptr addrspace(3) [[TMP9]], i32 0, i32 1, i32 1
126+
; CHECK-NEXT: [[TMP11:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
127+
; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr addrspace(3) [[TMP11]], align 4
128+
; CHECK-NEXT: [[TMP13:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
129+
; CHECK-NEXT: [[DOTI12:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP13]], i32 1
130+
; CHECK-NEXT: [[DOTI13:%.*]] = load i32, ptr addrspace(3) [[DOTI12]], align 4
131+
; CHECK-NEXT: [[TMP14:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
132+
; CHECK-NEXT: [[DOTI24:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP14]], i32 2
133+
; CHECK-NEXT: [[DOTI25:%.*]] = load i32, ptr addrspace(3) [[DOTI24]], align 4
134+
; CHECK-NEXT: [[TMP15:%.*]] = bitcast ptr addrspace(3) [[TMP10]] to ptr addrspace(3)
135+
; CHECK-NEXT: [[DOTI36:%.*]] = getelementptr i32, ptr addrspace(3) [[TMP15]], i32 3
136+
; CHECK-NEXT: [[DOTI37:%.*]] = load i32, ptr addrspace(3) [[DOTI36]], align 4
137+
; CHECK-NEXT: [[DOTI08:%.*]] = add i32 [[TMP2]], [[TMP12]]
138+
; CHECK-NEXT: [[DOTI19:%.*]] = add i32 [[TMP4]], [[DOTI13]]
139+
; CHECK-NEXT: [[DOTI210:%.*]] = add i32 [[TMP6]], [[DOTI25]]
140+
; CHECK-NEXT: [[DOTI311:%.*]] = add i32 [[TMP8]], [[DOTI37]]
141+
; CHECK-NEXT: [[DOTUPTO015:%.*]] = insertelement <4 x i32> poison, i32 [[DOTI08]], i64 0
142+
; CHECK-NEXT: [[DOTUPTO116:%.*]] = insertelement <4 x i32> [[DOTUPTO015]], i32 [[DOTI19]], i64 1
143+
; CHECK-NEXT: [[DOTUPTO217:%.*]] = insertelement <4 x i32> [[DOTUPTO116]], i32 [[DOTI210]], i64 2
144+
; CHECK-NEXT: [[TMP16:%.*]] = insertelement <4 x i32> [[DOTUPTO217]], i32 [[DOTI311]], i64 3
145+
; CHECK-NEXT: ret <4 x i32> [[TMP16]]
146+
;
147+
%1 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 0, i32 0), align 4
148+
%2 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 1, i32 1), align 4
149+
%3 = add <4 x i32> %1, %2
150+
ret <4 x i32> %3
151+
}
152+
153+
attributes #0 = { convergent norecurse nounwind "hlsl.export"}

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

Lines changed: 30 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,30 @@
1+
; RUN: opt -S -passes='dxil-data-scalarization,scalarizer<load-store>,dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
2+
3+
; Make sure we can store groupshared, static vectors and arrays of vectors
4+
5+
@"arrayofVecData" = local_unnamed_addr addrspace(3) global [2 x <3 x float>] zeroinitializer, align 16
6+
@"vecData" = external addrspace(3) global <4 x i32>, align 4
7+
8+
; CHECK: @arrayofVecData.scalarized = local_unnamed_addr addrspace(3) global [2 x [3 x float]] zeroinitializer, align 16
9+
; CHECK: @vecData.scalarized = external addrspace(3) global [4 x i32], align 4
10+
; CHECK-NOT: @arrayofVecData
11+
; CHECK-NOT: @vecData
12+
13+
; CHECK-LABEL: store_array_vec_test
14+
define void @store_array_vec_test () local_unnamed_addr #0 {
15+
; CHECK-COUNT-6: store float {{1|2|3|4|6}}.000000e+00, ptr addrspace(3) {{(.*@arrayofVecData.scalarized.*|%.*)}}, align {{4|8|16}}
16+
; CHECK-NEXT: ret void
17+
store <3 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00>, ptr addrspace(3) @"arrayofVecData", align 16
18+
store <3 x float> <float 2.000000e+00, float 4.000000e+00, float 6.000000e+00>, ptr addrspace(3) getelementptr inbounds (i8, ptr addrspace(3) @"arrayofVecData", i32 16), align 16
19+
ret void
20+
}
21+
22+
; CHECK-LABEL: store_vec_test
23+
define void @store_vec_test(<4 x i32> %inputVec) #0 {
24+
; CHECK-COUNT-4: store i32 %inputVec.{{.*}}, ptr addrspace(3) {{(@vecData.scalarized|getelementptr \(i32, ptr addrspace\(3\) @vecData.scalarized, i32 .*\)|%.*)}}, align 4
25+
; CHECK-NEXT: ret void
26+
store <4 x i32> %inputVec, <4 x i32> addrspace(3)* @"vecData", align 4
27+
ret void
28+
}
29+
30+
attributes #0 = { convergent norecurse nounwind "hlsl.export"}

‎llvm/test/CodeGen/DirectX/scalar-data.ll

Lines changed: 3 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -1,8 +1,6 @@
1-
; RUN: opt -S -passes='dxil-data-scalarization,function(scalarizer<load-store>),dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s -check-prefix DATACHECK
2-
; RUN: opt -S -passes='dxil-data-scalarization,dxil-flatten-arrays,function(scalarizer<load-store>),dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
31
; RUN: llc %s -mtriple=dxil-pc-shadermodel6.3-library --filetype=asm -o - | FileCheck %s
42

5-
; Make sure we don't touch arrays without vectors and that can recurse multiple-dimension arrays of vectors
3+
; Make sure we don't touch arrays without vectors and that can recurse and flatten multiple-dimension arrays of vectors
64

75
@staticArray = internal global [4 x i32] [i32 1, i32 2, i32 3, i32 4], align 4
86
@"groushared3dArrayofVectors" = local_unnamed_addr addrspace(3) global [3 x [3 x [3 x <4 x i32>]]] zeroinitializer, align 16
@@ -11,7 +9,7 @@
119
; CHECK-NOT: @staticArray.scalarized
1210
; CHECK-NOT: @staticArray.scalarized.1dim
1311
; CHECK-NOT: @staticArray.1dim
14-
; DATACHECK: @groushared3dArrayofVectors.scalarized = local_unnamed_addr addrspace(3) global [3 x [3 x [3 x [4 x i32]]]] zeroinitializer, align 16
12+
1513
; CHECK: @groushared3dArrayofVectors.scalarized.1dim = local_unnamed_addr addrspace(3) global [108 x i32] zeroinitializer, align 16
16-
; DATACHECK-NOT: @groushared3dArrayofVectors
1714
; CHECK-NOT: @groushared3dArrayofVectors.scalarized
15+
; CHECK-NOT: @groushared3dArrayofVectors

‎llvm/test/CodeGen/DirectX/scalar-load.ll

Lines changed: 8 additions & 20 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,4 @@
1-
; RUN: opt -S -passes='dxil-data-scalarization,function(scalarizer<load-store>),dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s -check-prefix DATACHECK
2-
; RUN: opt -S -passes='dxil-data-scalarization,dxil-flatten-arrays,function(scalarizer<load-store>),dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
3-
; RUN: llc %s -mtriple=dxil-pc-shadermodel6.3-library --filetype=asm -o - | FileCheck %s
1+
; RUN: opt -S -passes='dxil-data-scalarization,function(scalarizer<load-store>),dxil-op-lower' -mtriple=dxil-pc-shadermodel6.3-library %s | FileCheck %s
42

53
; Make sure we can load groupshared, static vectors and arrays of vectors
64

@@ -9,26 +7,20 @@
97
@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
108
@"groushared2dArrayofVectors" = local_unnamed_addr addrspace(3) global [3 x [ 3 x <4 x i32>]] zeroinitializer, align 16
119

12-
; DATACHECK: @arrayofVecData.scalarized = local_unnamed_addr addrspace(3) global [2 x [3 x float]] zeroinitializer, align 16
13-
; CHECK: @arrayofVecData.scalarized.1dim = local_unnamed_addr addrspace(3) global [6 x float] zeroinitializer, align 16
10+
; CHECK: @arrayofVecData.scalarized = local_unnamed_addr addrspace(3) global [2 x [3 x float]] zeroinitializer, align 16
1411
; CHECK: @vecData.scalarized = external addrspace(3) global [4 x i32], align 4
15-
; DATACHECK: @staticArrayOfVecData.scalarized = 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
16-
; 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
17-
; DATACHECK: @groushared2dArrayofVectors.scalarized = local_unnamed_addr addrspace(3) global [3 x [3 x [4 x i32]]] zeroinitializer, align 16
18-
; CHECK: @groushared2dArrayofVectors.scalarized.1dim = local_unnamed_addr addrspace(3) global [36 x i32] zeroinitializer, align 16
12+
; CHECK: @staticArrayOfVecData.scalarized = 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
13+
; CHECK: @groushared2dArrayofVectors.scalarized = local_unnamed_addr addrspace(3) global [3 x [3 x [4 x i32]]] zeroinitializer, align 16
1914

2015
; CHECK-NOT: @arrayofVecData
21-
; CHECK-NOT: @arrayofVecData.scalarized
2216
; CHECK-NOT: @vecData
2317
; CHECK-NOT: @staticArrayOfVecData
24-
; CHECK-NOT: @staticArrayOfVecData.scalarized
2518
; CHECK-NOT: @groushared2dArrayofVectors
26-
; CHECK-NOT: @groushared2dArrayofVectors.scalarized
19+
2720

2821
; CHECK-LABEL: load_array_vec_test
2922
define <4 x i32> @load_array_vec_test() #0 {
30-
; CHECK-COUNT-8: load i32, ptr addrspace(3) {{(.*@arrayofVecData.scalarized.1dim.*|%.*)}}, align 4
31-
; CHECK-NOT: load i32, ptr addrspace(3) {{.*}}, align 4
23+
; CHECK-COUNT-8: load i32, ptr addrspace(3) {{(.*@arrayofVecData.scalarized.*|%.*)}}, align 4
3224
%1 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([2 x <4 x i32>], [2 x <4 x i32>] addrspace(3)* @"arrayofVecData", i32 0, i32 0), align 4
3325
%2 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([2 x <4 x i32>], [2 x <4 x i32>] addrspace(3)* @"arrayofVecData", i32 0, i32 1), align 4
3426
%3 = add <4 x i32> %1, %2
@@ -38,26 +30,22 @@ define <4 x i32> @load_array_vec_test() #0 {
3830
; CHECK-LABEL: load_vec_test
3931
define <4 x i32> @load_vec_test() #0 {
4032
; CHECK-COUNT-4: load i32, ptr addrspace(3) {{(@vecData.scalarized|getelementptr \(i32, ptr addrspace\(3\) @vecData.scalarized, i32 .*\)|%.*)}}, align {{.*}}
41-
; CHECK-NOT: load i32, ptr addrspace(3) {{.*}}, align 4
4233
%1 = load <4 x i32>, <4 x i32> addrspace(3)* @"vecData", align 4
4334
ret <4 x i32> %1
4435
}
4536

4637
; CHECK-LABEL: load_static_array_of_vec_test
4738
define <4 x i32> @load_static_array_of_vec_test(i32 %index) #0 {
48-
; DATACHECK: getelementptr [3 x [4 x i32]], ptr @staticArrayOfVecData.scalarized, i32 0, i32 %index
49-
; CHECK: getelementptr [12 x i32], ptr @staticArrayOfVecData.scalarized.1dim, i32 %index
39+
; CHECK: getelementptr [3 x [4 x i32]], ptr @staticArrayOfVecData.scalarized, i32 0, i32 %index
5040
; CHECK-COUNT-4: load i32, ptr {{.*}}, align 4
51-
; CHECK-NOT: load i32, ptr {{.*}}, align 4
5241
%3 = getelementptr inbounds [3 x <4 x i32>], [3 x <4 x i32>]* @staticArrayOfVecData, i32 0, i32 %index
5342
%4 = load <4 x i32>, <4 x i32>* %3, align 4
5443
ret <4 x i32> %4
5544
}
5645

5746
; CHECK-LABEL: multid_load_test
5847
define <4 x i32> @multid_load_test() #0 {
59-
; CHECK-COUNT-8: load i32, ptr addrspace(3) {{(.*@groushared2dArrayofVectors.scalarized.1dim.*|%.*)}}, align 4
60-
; CHECK-NOT: load i32, ptr addrspace(3) {{.*}}, align 4
48+
; CHECK-COUNT-8: load i32, ptr addrspace(3) {{(.*@groushared2dArrayofVectors.scalarized.*|%.*)}}, align 4
6149
%1 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 0, i32 0), align 4
6250
%2 = load <4 x i32>, <4 x i32> addrspace(3)* getelementptr inbounds ([3 x [3 x <4 x i32>]], [3 x [3 x <4 x i32>]] addrspace(3)* @"groushared2dArrayofVectors", i32 0, i32 1, i32 1), align 4
6351
%3 = add <4 x i32> %1, %2

0 commit comments

Comments
 (0)