Skip to content

Commit 63a728d

Browse files
bharadwajybharadwajy
committed
[HLSL] Set function optnone attribute appropriately
When optimization is disabled, set optnone attribute - for all module functions when targetting Library shaders - only for entry function when targetting non-Library shaders Update tests in accordance with the change.
1 parent 2464f4b commit 63a728d

File tree

6 files changed

+174
-59
lines changed

6 files changed

+174
-59
lines changed

‎clang/lib/CodeGen/CGHLSLRuntime.cpp‎

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -345,6 +345,9 @@ void clang::CodeGen::CGHLSLRuntime::setHLSLEntryAttributes(
345345
WaveSizeAttr->getPreferred());
346346
Fn->addFnAttr(WaveSizeKindStr, WaveSizeStr);
347347
}
348+
if (CGM.getCodeGenOpts().OptimizationLevel == 0) {
349+
Fn->addFnAttr(llvm::Attribute::OptimizeNone);
350+
}
348351
Fn->addFnAttr(llvm::Attribute::NoInline);
349352
}
350353

@@ -446,6 +449,13 @@ void CGHLSLRuntime::setHLSLFunctionAttributes(const FunctionDecl *FD,
446449
const StringRef ExportAttrKindStr = "hlsl.export";
447450
Fn->addFnAttr(ExportAttrKindStr);
448451
}
452+
llvm::Triple T(Fn->getParent()->getTargetTriple());
453+
if (T.getEnvironment() == llvm::Triple::EnvironmentType::Library) {
454+
if (CGM.getCodeGenOpts().OptimizationLevel == 0) {
455+
Fn->addFnAttr(llvm::Attribute::OptimizeNone);
456+
Fn->addFnAttr(llvm::Attribute::NoInline);
457+
}
458+
}
449459
}
450460

451461
static void gatherFunctions(SmallVectorImpl<Function *> &Fns, llvm::Module &M,

‎clang/test/CodeGenHLSL/GlobalConstructorLib.hlsl‎

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1,5 +1,5 @@
11
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CHECK,NOINLINE
2-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -O0 %s -o - | FileCheck %s --check-prefixes=CHECK,INLINE
2+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -O1 %s -o - | FileCheck %s --check-prefixes=CHECK,INLINE
33

44
// Make sure global variable for ctors exist for lib profile.
55
// CHECK:@llvm.global_ctors
@@ -31,12 +31,12 @@ void SecondEntry() {}
3131
// CHECK: ret void
3232

3333

34-
// Verify the constructor is alwaysinline
35-
// NOINLINE: ; Function Attrs: {{.*}}alwaysinline
34+
// Verify the constructor is optnone
35+
// NOINLINE: ; Function Attrs: {{.*}} optnone
3636
// NOINLINE-NEXT: define linkonce_odr void @_ZN4hlsl8RWBufferIfEC2Ev({{.*}} [[CtorAttr:\#[0-9]+]]
3737

3838
// NOINLINE: ; Function Attrs: {{.*}}alwaysinline
3939
// NOINLINE-NEXT: define internal void @_GLOBAL__sub_I_GlobalConstructorLib.hlsl() [[InitAttr:\#[0-9]+]]
4040

4141
// NOINLINE-DAG: attributes [[InitAttr]] = {{.*}} alwaysinline
42-
// NOINLINE-DAG: attributes [[CtorAttr]] = {{.*}} alwaysinline
42+
// NOINLINE-DAG: attributes [[CtorAttr]] = {{.*}} optnone

‎clang/test/CodeGenHLSL/GlobalDestructors.hlsl‎

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=CS,NOINLINE,CHECK
22
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -disable-llvm-passes %s -o - | FileCheck %s --check-prefixes=LIB,NOINLINE,CHECK
3-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -O0 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK
4-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -O0 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK
3+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -emit-llvm -O1 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK
4+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -emit-llvm -O1 %s -o - | FileCheck %s --check-prefixes=INLINE,CHECK
55

66
// Tests that constructors and destructors are appropriately generated for globals
77
// and that their calls are inlined when AlwaysInline is run

‎clang/test/CodeGenHLSL/inline-constructors.hlsl‎

Lines changed: 10 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -1,9 +1,9 @@
1-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -emit-llvm -o - -disable-llvm-passes %s | FileCheck %s --check-prefixes=CHECK,NOINLINE
2-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -emit-llvm -o - -disable-llvm-passes %s | FileCheck %s --check-prefixes=CHECK,NOINLINE
3-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -emit-llvm -o - -O0 %s | FileCheck %s --check-prefixes=CHECK,INLINE
4-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -emit-llvm -o - -O0 %s | FileCheck %s --check-prefixes=CHECK,INLINE
5-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -emit-llvm -o - -O1 %s | FileCheck %s --check-prefixes=CHECK,INLINE
6-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -emit-llvm -o - -O1 %s | FileCheck %s --check-prefixes=CHECK,INLINE
1+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -disable-llvm-passes %s | FileCheck %s --check-prefixes=CHECK,NOINLINE
2+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -disable-llvm-passes %s | FileCheck %s --check-prefixes=CHECK,NOINLINE
3+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -O0 %s | FileCheck %s --check-prefixes=CHECK,INLINE
4+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -O0 %s | FileCheck %s --check-prefixes=CHECK,NOINLINE_LIB
5+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -O1 %s | FileCheck %s --check-prefixes=CHECK,INLINE
6+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library -std=hlsl202x -Wno-hlsl-extensions -emit-llvm -o - -O1 %s | FileCheck %s --check-prefixes=CHECK,INLINE
77

88
// Tests that implicit constructor calls for user classes will always be inlined.
99

@@ -50,6 +50,10 @@ void NionsDay(int hours) {
5050
// NOINLINE-NEXT: call void @_GLOBAL__sub_I_inline_constructors.hlsl()
5151
// NOINLINE-NEXT: %0 = call i32 @llvm.dx.flattened.thread.id.in.group()
5252
// NOINLINE-NEXT: call void @_Z4mainj(i32 %0)
53+
// NOINLINE_LIB: call void @_ZN4WeedC1Ev
54+
// NOINLINE_LIB-NEXT: call void @_ZN5KittyC1Ev
55+
// NOINLINE_LIB-NEXT: %0 = call i32 @llvm.dx.flattened.thread.id.in.group()
56+
// NOINLINE_LIB-NEXT: call void @_Z4mainj(i32 %0)
5357
// Verify inlining leaves only calls to "llvm." intrinsics
5458
// INLINE-NOT: call {{[^@]*}} @{{[^l][^l][^v][^m][^\.]}}
5559
// CHECK: ret void

‎clang/test/CodeGenHLSL/inline-functions.hlsl‎

Lines changed: 144 additions & 45 deletions
Original file line numberDiff line numberDiff line change
@@ -1,32 +1,60 @@
1-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s --check-prefixes=CHECK,NOINLINE
2-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -O0 -o - | FileCheck %s --check-prefixes=CHECK,INLINE
3-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -O1 -o - | FileCheck %s --check-prefixes=CHECK,INLINE
4-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s --check-prefixes=CHECK,NOINLINE
5-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -O0 -o - | FileCheck %s --check-prefixes=CHECK,INLINE
6-
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -O1 -o - | FileCheck %s --check-prefixes=CHECK,INLINE
7-
8-
// Tests that user functions will always be inlined.
1+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s --check-prefixes=CHECK,CHECK_LIB_OPTNONE
2+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -O0 -o - | FileCheck %s --check-prefixes=CHECK,CHECK_LIB_OPTNONE
3+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.3-library %s -emit-llvm -O1 -o - | FileCheck %s --check-prefixes=CHECK,CHECK_OPT
4+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -disable-llvm-passes -o - | FileCheck %s --check-prefixes=CHECK,CHECK_CS_OPTNONE_NOPASS
5+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -O0 -o - | FileCheck %s --check-prefixes=CHECK,CHECK_CS_OPTNONE_PASS
6+
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.0-compute %s -emit-llvm -O1 -o - | FileCheck %s --check-prefixes=CHECK,CHECK_OPT
7+
8+
// Tests inlining of user functions based on specified optimization options.
99
// This includes exported functions and mangled entry point implementation functions.
10-
// The unmangled entry functions must not be alwaysinlined.
1110

1211
#define MAX 100
1312

1413
float nums[MAX];
1514

16-
// Verify that all functions have the alwaysinline attribute
17-
// NOINLINE: Function Attrs: alwaysinline
18-
// NOINLINE: define void @_Z4swapA100_jjj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %ix1, i32 noundef %ix2) [[IntAttr:\#[0-9]+]]
19-
// NOINLINE: ret void
15+
// Check optnone attribute for library target compilation
16+
// CHECK_LIB_OPTNONE: Function Attrs:{{.*}}optnone
17+
// CHECK_LIB_OPTNONE: define void @_Z4swapA100_jjj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %ix1, i32 noundef %ix2) [[ExtAttr:\#[0-9]+]]
18+
19+
// Check alwaysinline attribute for non-entry functions of compute target compilation
20+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs: alwaysinline
21+
// CHECK_CS_OPTNONE_NOPASS: define void @_Z4swapA100_jjj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %ix1, i32 noundef %ix2) [[ExtAttr:\#[0-9]+]]
22+
23+
// Check alwaysinline attribute for non-entry functions of compute target compilation
24+
// CHECK_CS_OPTNONE_PASS: Function Attrs: alwaysinline
25+
// CHECK_CS_OPTNONE_PASS: define void @_Z4swapA100_jjj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %ix1, i32 noundef %ix2) [[ExtAttr:\#[0-9]+]]
26+
27+
// Check alwaysinline attribute for opt compilation to library target
28+
// CHECK_OPT: Function Attrs: alwaysinline
29+
// CHECK_OPT: define void @_Z4swapA100_jjj(ptr noundef byval([100 x i32]) align 4 captures(none) %Buf, i32 noundef %ix1, i32 noundef %ix2) {{[a-z_ ]*}} [[SwapOptAttr:\#[0-9]+]]
30+
31+
// CHECK: ret void
32+
2033
// Swap the values of Buf at indices ix1 and ix2
2134
void swap(unsigned Buf[MAX], unsigned ix1, unsigned ix2) {
2235
float tmp = Buf[ix1];
2336
Buf[ix1] = Buf[ix2];
2437
Buf[ix2] = tmp;
2538
}
2639

27-
// NOINLINE: Function Attrs: alwaysinline
28-
// NOINLINE: define void @_Z10BubbleSortA100_jj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %size) [[IntAttr]]
29-
// NOINLINE: ret void
40+
// Check optnone attribute for library target compilation
41+
// CHECK_LIB_OPTNONE: Function Attrs:{{.*}}optnone
42+
// CHECK_LIB_OPTNONE: define void @_Z10BubbleSortA100_jj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %size) [[ExtAttr]]
43+
44+
// Check alwaysinline attribute for non-entry functions of compute target compilation
45+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs: alwaysinline
46+
// CHECK_CS_OPTNONE_NOPASS: define void @_Z10BubbleSortA100_jj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %size) [[ExtAttr]]
47+
48+
// Check alwaysinline attribute for non-entry functions of compute target compilation
49+
// CHECK_CS_OPTNONE_PASS: Function Attrs: alwaysinline
50+
// CHECK_CS_OPTNONE_PASS: define void @_Z10BubbleSortA100_jj(ptr noundef byval([100 x i32]) align 4 %Buf, i32 noundef %size) [[ExtAttr]]
51+
52+
// Check alwaysinline attribute for opt compilation to library target
53+
// CHECK_OPT: Function Attrs: alwaysinline
54+
// CHECK_OPT: define void @_Z10BubbleSortA100_jj(ptr noundef readonly byval([100 x i32]) align 4 captures(none) %Buf, i32 noundef %size) {{[a-z_ ]*}} [[BubOptAttr:\#[0-9]+]]
55+
56+
// CHECK: ret void
57+
3058
// Inefficiently sort Buf in place
3159
void BubbleSort(unsigned Buf[MAX], unsigned size) {
3260
bool swapped = true;
@@ -41,12 +69,26 @@ void BubbleSort(unsigned Buf[MAX], unsigned size) {
4169
}
4270
}
4371

44-
// Note ExtAttr is the inlined export set of attribs
45-
// CHECK: Function Attrs: alwaysinline
46-
// CHECK: define noundef i32 @_Z11RemoveDupesA100_jj(ptr {{[a-z_ ]*}}noundef byval([100 x i32]) align 4 {{.*}}%Buf, i32 noundef %size) {{[a-z_ ]*}}[[ExtAttr:\#[0-9]+]]
47-
// CHECK: ret i32
72+
// Check optnone attribute for library target compilation of exported function
73+
// CHECK_LIB_OPTNONE: Function Attrs:{{.*}}optnone
74+
// CHECK_LIB_OPTNONE: define noundef i32 @_Z11RemoveDupesA100_jj(ptr {{[a-z_ ]*}}noundef byval([100 x i32]) align 4 {{.*}}%Buf, i32 noundef %size) [[ExportAttr:\#[0-9]+]]
4875
// Sort Buf and remove any duplicate values
4976
// returns the number of values left
77+
78+
// Check alwaysinline attribute for exported function of compute target compilation
79+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs: alwaysinline
80+
// CHECK_CS_OPTNONE_NOPASS: define noundef i32 @_Z11RemoveDupesA100_jj(ptr {{[a-z_ ]*}}noundef byval([100 x i32]) align 4 {{.*}}%Buf, i32 noundef %size) [[ExportAttr:\#[0-9]+]]
81+
82+
// Check alwaysinline attribute for exported function of compute target compilation
83+
// CHECK_CS_OPTNONE_PASS: Function Attrs: alwaysinline
84+
// CHECK_CS_OPTNONE_PASS: define noundef i32 @_Z11RemoveDupesA100_jj(ptr {{[a-z_ ]*}}noundef byval([100 x i32]) align 4 {{.*}}%Buf, i32 noundef %size) [[ExportAttr:\#[0-9]+]]
85+
86+
// Check alwaysinline attribute for exported function of library target compilation
87+
// CHECK_OPT: Function Attrs: alwaysinline
88+
// CHECK_OPT: define noundef i32 @_Z11RemoveDupesA100_jj(ptr noundef byval([100 x i32]) align 4 captures(none) %Buf, i32 noundef %size) {{[a-z_ ]*}} [[RemOptAttr:\#[0-9]+]]
89+
90+
// CHECK: ret i32
91+
5092
export
5193
unsigned RemoveDupes(unsigned Buf[MAX], unsigned size) {
5294
BubbleSort(Buf, size);
@@ -63,19 +105,44 @@ unsigned RemoveDupes(unsigned Buf[MAX], unsigned size) {
63105

64106
RWBuffer<unsigned> Indices;
65107

66-
// The mangled version of main only remains without inlining
67-
// because it has internal linkage from the start
68-
// Note main functions get the norecurse attrib, which IntAttr reflects
69-
// NOINLINE: Function Attrs: alwaysinline
70-
// NOINLINE: define internal void @_Z4mainj(i32 noundef %GI) [[IntAttr]]
71-
// NOINLINE: ret void
108+
// CHECK_LIB_OPTNONE: Function Attrs:{{.*}}optnone
109+
// Internal function attributes are the same as those of source function's
110+
// CHECK_LIB_OPTNONE: define internal void @_Z4mainj(i32 noundef %GI) [[ExtAttr]]
111+
// CHECK_LIB_OPTNONE: ret void
112+
113+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs: alwaysinline
114+
// Internal function attributes are different from those of source function's
115+
// CHECK_CS_OPTNONE_NOPASS: define internal void @_Z4mainj(i32 noundef %GI) [[ExtAttr]]
116+
// CHECK_CS_OPTNONE_NOPASS: ret void
117+
118+
// Check internal function @_Z4mainj is not generated when LLVM passes enabled
119+
// CHECK_CS_OPTNONE_PASS-NOT: define internal void @_Z4mainj
120+
121+
// Check internal function @_Z4mainj is not generated as it should be inlined
122+
// for opt builds
123+
// CHECK_OPT-NOT: define internal void @_Z4mainj
72124

73125
// The unmangled version is not inlined, EntryAttr reflects that
74-
// CHECK: Function Attrs: {{.*}}noinline
75-
// CHECK: define void @main() {{[a-z_ ]*}}[[EntryAttr:\#[0-9]+]]
76-
// Make sure function calls are inlined when AlwaysInline is run
77-
// This only leaves calls to llvm. intrinsics
78-
// INLINE-NOT: call {{[^@]*}} @{{[^l][^l][^v][^m][^\.]}}
126+
// CHECK_LIB_OPTNONE: Function Attrs: {{.*}}noinline
127+
// CHECK_LIB_OPTNONE: define void @main() {{[a-z_ ]*}}[[EntryAttr:\#[0-9]+]]
128+
// Make sure internal function is not inlined when optimization is disabled
129+
// CHECK_LIB_OPTNONE: call void @_Z4mainj
130+
131+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs:{{.*}}optnone
132+
// CHECK_CS_OPTNONE_NOPASS: define void @main() {{[a-z_ ]*}}[[EntryAttr:\#[0-9]+]]
133+
// Make sure internal function is not inlined when optimization is disabled
134+
// CHECK_CS_OPTNONE_NOPASS: call void @_Z4mainj
135+
136+
// CHECK_CS_OPTNONE_PASS: Function Attrs: {{.*}}noinline
137+
// CHECK_CS_OPTNONE_PASS: define void @main() {{[a-z_ ]*}}[[EntryAttr:\#[0-9]+]]
138+
// Make sure internal function is inlined when LLVM passes are enabled
139+
// CHECK_CS_OPTNONE_PASS: _Z4mainj.exit:
140+
141+
// CHECK_OPT: Function Attrs: {{.*}}noinline
142+
// CHECK_OPT: define void @main() {{[a-z_ ]*}}[[EntryAttr:\#[0-9]+]]
143+
// Make sure internal function is inlined as optimization is enabled
144+
// CHECK_OPT: _Z4mainj.exit:
145+
79146
// CHECK: ret void
80147

81148
[numthreads(1,1,1)]
@@ -90,19 +157,41 @@ void main(unsigned int GI : SV_GroupIndex) {
90157
tmpIndices[i] = Indices[i];
91158
}
92159

93-
// The mangled version of main only remains without inlining
94-
// because it has internal linkage from the start
95-
// Note main functions get the norecurse attrib, which IntAttr reflects
96-
// NOINLINE: Function Attrs: alwaysinline
97-
// NOINLINE: define internal void @_Z6main10v() [[IntAttr]]
98-
// NOINLINE: ret void
160+
// CHECK_LIB_OPTNONE: Function Attrs:{{.*}}optnone
161+
// CHECK_LIB_OPTNONE: define internal void @_Z6main10v() [[ExtAttr]]
162+
// CHECK_LIB_OPTNONE: ret void
163+
164+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs:{{.*}}alwaysinline
165+
// CHECK_CS_OPTNONE_NOPASS: define internal void @_Z6main10v() [[ExtAttr]]
166+
// CHECK_CS_OPTNONE_NOPASS: ret void
167+
168+
// Check internal function @_Z6main10v is not generated when LLVM passes are enabled
169+
// CHECK_CS_OPTNONE_PASS-NOT: define internal void @_Z6main10v
170+
171+
// Check internal function @_Z6main10v is not generated as it should be inlined
172+
// CHECK_OPT-NOT: define internal void @_Z6main10v
99173

100174
// The unmangled version is not inlined, EntryAttr reflects that
101-
// CHECK: Function Attrs: {{.*}}noinline
102-
// CHECK: define void @main10() {{[a-z_ ]*}}[[EntryAttr]]
103-
// Make sure function calls are inlined when AlwaysInline is run
104-
// This only leaves calls to llvm. intrinsics
105-
// INLINE-NOT: call {{[^@]*}} @{{[^l][^l][^v][^m][^\.]}}
175+
// CHECK_LIB_OPTNONE: Function Attrs: {{.*}}noinline
176+
// CHECK_LIB_OPTNONE: define void @main10() {{[a-z_ ]*}}[[EntryAttr]]
177+
// Make sure internal function is not inlined when optimization is disabled
178+
// CHECK_LIB_OPTNONE: call void @_Z6main10v
179+
180+
// CHECK_CS_OPTNONE_NOPASS: Function Attrs: {{.*}}noinline
181+
// CHECK_CS_OPTNONE_NOPASS: define void @main10() {{[a-z_ ]*}}[[EntryAttr]]
182+
// Make sure internal function is not inlined when optimization is disabled
183+
// CHECK_CS_OPTNONE_NOPASS: call void @_Z6main10v
184+
185+
// CHECK_CS_OPTNONE_PASS: Function Attrs: {{.*}}noinline
186+
// CHECK_CS_OPTNONE_PASS: define void @main10() {{[a-z_ ]*}}[[EntryAttr]]
187+
// Check internal function is inlined as optimization is enabled when LLVM passes
188+
// are enabled
189+
// CHECK_CS_OPTNONE_PASS: _Z6main10v.exit:
190+
191+
// CHECK_OPT: Function Attrs: {{.*}}noinline
192+
// CHECK_OPT: define void @main10() {{[a-z_ ]*}}[[EntryAttr]]
193+
// Make sure internal function is inlined as optimization is enabled
194+
// CHECK_OPT: _Z6main10v.exit:
106195
// CHECK: ret void
107196

108197
[numthreads(1,1,1)]
@@ -111,6 +200,16 @@ void main10() {
111200
main(10);
112201
}
113202

114-
// NOINLINE: attributes [[IntAttr]] = {{.*}} alwaysinline
115-
// CHECK: attributes [[ExtAttr]] = {{.*}} alwaysinline
116-
// CHECK: attributes [[EntryAttr]] = {{.*}} noinline
203+
// CHECK_LIB_OPTNONE: attributes [[ExtAttr]] = {{.*}} optnone
204+
// CHECK_LIB_OPTNONE: attributes [[ExportAttr]] = {{.*}} optnone
205+
206+
// CHECK_CS_OPTNONE_NOPASS: attributes [[ExtAttr]] ={{.*}} alwaysinline
207+
// CHECK_CS_OPTNONE_NOPASS: attributes [[EntryAttr]] = {{.*}} noinline
208+
209+
// CHECK_CS_OPTNONE_PASS: attributes [[ExtAttr]] ={{.*}} alwaysinline
210+
// CHECK_CS_OPTNONE_PASS: attributes [[EntryAttr]] = {{.*}} noinline
211+
212+
// CHECK_OPT: attributes [[SwapOptAttr]] ={{.*}} alwaysinline
213+
// CHECK_OPT: attributes [[BubOptAttr]] ={{.*}} alwaysinline
214+
// CHECK_OPT: attributes [[RemOptAttr]] ={{.*}} alwaysinline
215+
// CHECK_OPT: attributes [[EntryAttr]] ={{.*}} noinline

‎clang/test/CodeGenHLSL/this-assignment-overload.hlsl‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -25,7 +25,7 @@ void main() {
2525
}
2626

2727
// This test makes a probably safe assumption that HLSL 202x includes operator overloading for assignment operators.
28-
// CHECK: define linkonce_odr noundef i32 @_ZN4Pair8getFirstEv(ptr noundef nonnull align 4 dereferenceable(8) %this) #0 align 2 {
28+
// CHECK: define linkonce_odr noundef i32 @_ZN4Pair8getFirstEv(ptr noundef nonnull align 4 dereferenceable(8) %this) [[Attr:\#[0-9]+]] align 2 {
2929
// CHECK-NEXT:entry:
3030
// CHECK-NEXT:%this.addr = alloca ptr, align 4
3131
// CHECK-NEXT:%Another = alloca %struct.Pair, align 4
@@ -42,7 +42,7 @@ void main() {
4242
// CHECK-NEXT:%0 = load i32, ptr %First2, align 4
4343
// CHECK-NEXT:ret i32 %0
4444

45-
// CHECK: define linkonce_odr noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 4 dereferenceable(8) %this) #0 align 2 {
45+
// CHECK: define linkonce_odr noundef i32 @_ZN4Pair9getSecondEv(ptr noundef nonnull align 4 dereferenceable(8) %this) [[Attr]] align 2 {
4646
// CHECK-NEXT:entry:
4747
// CHECK-NEXT:%this.addr = alloca ptr, align 4
4848
// CHECK-NEXT:%agg.tmp = alloca %struct.Pair, align 4
@@ -53,3 +53,5 @@ void main() {
5353
// CHECK-NEXT:%Second = getelementptr inbounds nuw %struct.Pair, ptr %this1, i32 0, i32 1
5454
// CHECK-NEXT:%0 = load i32, ptr %Second, align 4
5555
// CHECK-NEXT:ret i32 %0
56+
57+
// CHECK: attributes [[Attr]] = {{.*}}alwaysinline

0 commit comments

Comments
 (0)