Skip to content

Commit 36c1194

Browse files
MatzeBMatzeB
authored
Remove optimization flags from clang codegen tests (llvm#113714)
- Remove an -O3 flag from a couple of clang x86 codegen tests so the tests do not need to be updated when optimizations in LLVM change. - Change the tests to use utils/update_cc_test_checks.sh - Change from apple/darwin triples to generic x86_64-- and i386-- because it was not relevant to the test but `update_cc_test_checks` seems to be unable to handle platforms that prepend `_` to function names.
1 parent 0eb5c9d commit 36c1194

File tree

3 files changed

+348
-116
lines changed

3 files changed

+348
-116
lines changed

‎clang/test/CodeGen/X86/avx-cmp-builtins.c

Lines changed: 83 additions & 21 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,6 @@
1-
// RUN: %clang_cc1 -flax-vector-conversions=none -ffreestanding %s -O3 -triple=x86_64-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s
2-
// RUN: %clang_cc1 -flax-vector-conversions=none -ffreestanding %s -O3 -triple=i386-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s
3-
// FIXME: The shufflevector instructions in test_cmpgt_sd are relying on O3 here.
1+
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
2+
// RUN: %clang_cc1 -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-- -target-feature +avx -disable-O0-optnone -emit-llvm -o - | opt -S -passes=mem2reg | FileCheck %s
3+
// RUN: %clang_cc1 -flax-vector-conversions=none -ffreestanding %s -triple=i386-- -target-feature +avx -disable-O0-optnone -emit-llvm -o - | opt -S -passes=mem2reg | FileCheck %s
44

55

66
#include <immintrin.h>
@@ -9,62 +9,124 @@
99
// Test LLVM IR codegen of cmpXY instructions
1010
//
1111

12+
// CHECK-LABEL: define dso_local <2 x double> @test_cmp_sd(
13+
// CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0:[0-9]+]] {
14+
// CHECK-NEXT: [[ENTRY:.*:]]
15+
// CHECK-NEXT: [[TMP0:%.*]] = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[A]], <2 x double> [[B]], i8 13)
16+
// CHECK-NEXT: ret <2 x double> [[TMP0]]
17+
//
1218
__m128d test_cmp_sd(__m128d a, __m128d b) {
1319
// Expects that the third argument in LLVM IR is immediate expression
14-
// CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 13)
1520
return _mm_cmp_sd(a, b, _CMP_GE_OS);
1621
}
1722

23+
// CHECK-LABEL: define dso_local <4 x float> @test_cmp_ss(
24+
// CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] {
25+
// CHECK-NEXT: [[ENTRY:.*:]]
26+
// CHECK-NEXT: [[TMP0:%.*]] = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[A]], <4 x float> [[B]], i8 13)
27+
// CHECK-NEXT: ret <4 x float> [[TMP0]]
28+
//
1829
__m128 test_cmp_ss(__m128 a, __m128 b) {
1930
// Expects that the third argument in LLVM IR is immediate expression
20-
// CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 13)
2131
return _mm_cmp_ss(a, b, _CMP_GE_OS);
2232
}
2333

34+
// CHECK-LABEL: define dso_local <4 x float> @test_cmpgt_ss(
35+
// CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] {
36+
// CHECK-NEXT: [[ENTRY:.*:]]
37+
// CHECK-NEXT: [[TMP0:%.*]] = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[B]], <4 x float> [[A]], i8 1)
38+
// CHECK-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x float> [[A]], <4 x float> [[TMP0]], <4 x i32> <i32 4, i32 1, i32 2, i32 3>
39+
// CHECK-NEXT: ret <4 x float> [[SHUFFLE_I]]
40+
//
2441
__m128 test_cmpgt_ss(__m128 a, __m128 b) {
25-
// CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 1)
26-
// CHECK: shufflevector <{{.*}}, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
2742
return _mm_cmpgt_ss(a, b);
2843
}
2944

45+
// CHECK-LABEL: define dso_local <4 x float> @test_cmpge_ss(
46+
// CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] {
47+
// CHECK-NEXT: [[ENTRY:.*:]]
48+
// CHECK-NEXT: [[TMP0:%.*]] = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[B]], <4 x float> [[A]], i8 2)
49+
// CHECK-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x float> [[A]], <4 x float> [[TMP0]], <4 x i32> <i32 4, i32 1, i32 2, i32 3>
50+
// CHECK-NEXT: ret <4 x float> [[SHUFFLE_I]]
51+
//
3052
__m128 test_cmpge_ss(__m128 a, __m128 b) {
31-
// CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 2)
32-
// CHECK: shufflevector <{{.*}}, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
3353
return _mm_cmpge_ss(a, b);
3454
}
3555

56+
// CHECK-LABEL: define dso_local <4 x float> @test_cmpngt_ss(
57+
// CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] {
58+
// CHECK-NEXT: [[ENTRY:.*:]]
59+
// CHECK-NEXT: [[TMP0:%.*]] = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[B]], <4 x float> [[A]], i8 5)
60+
// CHECK-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x float> [[A]], <4 x float> [[TMP0]], <4 x i32> <i32 4, i32 1, i32 2, i32 3>
61+
// CHECK-NEXT: ret <4 x float> [[SHUFFLE_I]]
62+
//
3663
__m128 test_cmpngt_ss(__m128 a, __m128 b) {
37-
// CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 5)
38-
// CHECK: shufflevector <{{.*}}, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
3964
return _mm_cmpngt_ss(a, b);
4065
}
4166

67+
// CHECK-LABEL: define dso_local <4 x float> @test_cmpnge_ss(
68+
// CHECK-SAME: <4 x float> noundef [[A:%.*]], <4 x float> noundef [[B:%.*]]) #[[ATTR0]] {
69+
// CHECK-NEXT: [[ENTRY:.*:]]
70+
// CHECK-NEXT: [[TMP0:%.*]] = call <4 x float> @llvm.x86.sse.cmp.ss(<4 x float> [[B]], <4 x float> [[A]], i8 6)
71+
// CHECK-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <4 x float> [[A]], <4 x float> [[TMP0]], <4 x i32> <i32 4, i32 1, i32 2, i32 3>
72+
// CHECK-NEXT: ret <4 x float> [[SHUFFLE_I]]
73+
//
4274
__m128 test_cmpnge_ss(__m128 a, __m128 b) {
43-
// CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 6)
44-
// CHECK: shufflevector <{{.*}}, <4 x i32> <i32 0, i32 5, i32 6, i32 7>
4575
return _mm_cmpnge_ss(a, b);
4676
}
4777

78+
// CHECK-LABEL: define dso_local <2 x double> @test_cmpgt_sd(
79+
// CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] {
80+
// CHECK-NEXT: [[ENTRY:.*:]]
81+
// CHECK-NEXT: [[TMP0:%.*]] = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[B]], <2 x double> [[A]], i8 1)
82+
// CHECK-NEXT: [[VECEXT_I:%.*]] = extractelement <2 x double> [[TMP0]], i32 0
83+
// CHECK-NEXT: [[VECINIT_I:%.*]] = insertelement <2 x double> poison, double [[VECEXT_I]], i32 0
84+
// CHECK-NEXT: [[VECEXT1_I:%.*]] = extractelement <2 x double> [[A]], i32 1
85+
// CHECK-NEXT: [[VECINIT2_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double [[VECEXT1_I]], i32 1
86+
// CHECK-NEXT: ret <2 x double> [[VECINIT2_I]]
87+
//
4888
__m128d test_cmpgt_sd(__m128d a, __m128d b) {
49-
// CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 1)
50-
// CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3>
5189
return _mm_cmpgt_sd(a, b);
5290
}
5391

92+
// CHECK-LABEL: define dso_local <2 x double> @test_cmpge_sd(
93+
// CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] {
94+
// CHECK-NEXT: [[ENTRY:.*:]]
95+
// CHECK-NEXT: [[TMP0:%.*]] = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[B]], <2 x double> [[A]], i8 2)
96+
// CHECK-NEXT: [[VECEXT_I:%.*]] = extractelement <2 x double> [[TMP0]], i32 0
97+
// CHECK-NEXT: [[VECINIT_I:%.*]] = insertelement <2 x double> poison, double [[VECEXT_I]], i32 0
98+
// CHECK-NEXT: [[VECEXT1_I:%.*]] = extractelement <2 x double> [[A]], i32 1
99+
// CHECK-NEXT: [[VECINIT2_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double [[VECEXT1_I]], i32 1
100+
// CHECK-NEXT: ret <2 x double> [[VECINIT2_I]]
101+
//
54102
__m128d test_cmpge_sd(__m128d a, __m128d b) {
55-
// CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 2)
56-
// CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3>
57103
return _mm_cmpge_sd(a, b);
58104
}
59105

106+
// CHECK-LABEL: define dso_local <2 x double> @test_cmpngt_sd(
107+
// CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] {
108+
// CHECK-NEXT: [[ENTRY:.*:]]
109+
// CHECK-NEXT: [[TMP0:%.*]] = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[B]], <2 x double> [[A]], i8 5)
110+
// CHECK-NEXT: [[VECEXT_I:%.*]] = extractelement <2 x double> [[TMP0]], i32 0
111+
// CHECK-NEXT: [[VECINIT_I:%.*]] = insertelement <2 x double> poison, double [[VECEXT_I]], i32 0
112+
// CHECK-NEXT: [[VECEXT1_I:%.*]] = extractelement <2 x double> [[A]], i32 1
113+
// CHECK-NEXT: [[VECINIT2_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double [[VECEXT1_I]], i32 1
114+
// CHECK-NEXT: ret <2 x double> [[VECINIT2_I]]
115+
//
60116
__m128d test_cmpngt_sd(__m128d a, __m128d b) {
61-
// CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 5)
62-
// CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3>
63117
return _mm_cmpngt_sd(a, b);
64118
}
65119

120+
// CHECK-LABEL: define dso_local <2 x double> @test_cmpnge_sd(
121+
// CHECK-SAME: <2 x double> noundef [[A:%.*]], <2 x double> noundef [[B:%.*]]) #[[ATTR0]] {
122+
// CHECK-NEXT: [[ENTRY:.*:]]
123+
// CHECK-NEXT: [[TMP0:%.*]] = call <2 x double> @llvm.x86.sse2.cmp.sd(<2 x double> [[B]], <2 x double> [[A]], i8 6)
124+
// CHECK-NEXT: [[VECEXT_I:%.*]] = extractelement <2 x double> [[TMP0]], i32 0
125+
// CHECK-NEXT: [[VECINIT_I:%.*]] = insertelement <2 x double> poison, double [[VECEXT_I]], i32 0
126+
// CHECK-NEXT: [[VECEXT1_I:%.*]] = extractelement <2 x double> [[A]], i32 1
127+
// CHECK-NEXT: [[VECINIT2_I:%.*]] = insertelement <2 x double> [[VECINIT_I]], double [[VECEXT1_I]], i32 1
128+
// CHECK-NEXT: ret <2 x double> [[VECINIT2_I]]
129+
//
66130
__m128d test_cmpnge_sd(__m128d a, __m128d b) {
67-
// CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 6)
68-
// CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3>
69131
return _mm_cmpnge_sd(a, b);
70132
}

0 commit comments

Comments
 (0)