| John McCall | 64d3e89 | 2011-06-03 00:02:45 +0000 | [diff] [blame] | 1 | // RUN: %clang_cc1 %s -O3 -triple=x86_64-apple-darwin -target-feature +avx -emit-llvm -o - | FileCheck %s |
| 2 | |
| 3 | // Don't include mm_malloc.h, it's system specific. |
| 4 | #define __MM_MALLOC_H |
| 5 | |
| 6 | #include <immintrin.h> |
| 7 | |
| 8 | // |
| 9 | // Test LLVM IR codegen of cmpXY instructions |
| 10 | // |
| 11 | |
| 12 | __m128d test_cmp_pd(__m128d a, __m128d b) { |
| 13 | // Expects that the third argument in LLVM IR is immediate expression |
| 14 | // CHECK: @llvm.x86.sse2.cmp.pd({{.*}}, i8 13) |
| 15 | return _mm_cmp_pd(a, b, _CMP_GE_OS); |
| 16 | } |
| 17 | |
| 18 | __m128d test_cmp_ps(__m128 a, __m128 b) { |
| 19 | // Expects that the third argument in LLVM IR is immediate expression |
| 20 | // CHECK: @llvm.x86.sse.cmp.ps({{.*}}, i8 13) |
| 21 | return _mm_cmp_ps(a, b, _CMP_GE_OS); |
| 22 | } |
| 23 | |
| 24 | __m256d test_cmp_pd256(__m256d a, __m256d b) { |
| 25 | // Expects that the third argument in LLVM IR is immediate expression |
| 26 | // CHECK: @llvm.x86.avx.cmp.pd.256({{.*}}, i8 13) |
| 27 | return _mm256_cmp_pd(a, b, _CMP_GE_OS); |
| 28 | } |
| 29 | |
| 30 | __m256d test_cmp_ps256(__m256 a, __m256 b) { |
| 31 | // Expects that the third argument in LLVM IR is immediate expression |
| 32 | // CHECK: @llvm.x86.avx.cmp.ps.256({{.*}}, i8 13) |
| 33 | return _mm256_cmp_ps(a, b, _CMP_GE_OS); |
| 34 | } |
| 35 | |
| 36 | __m128d test_cmp_sd(__m128d a, __m128d b) { |
| 37 | // Expects that the third argument in LLVM IR is immediate expression |
| 38 | // CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 13) |
| 39 | return _mm_cmp_sd(a, b, _CMP_GE_OS); |
| 40 | } |
| 41 | |
| 42 | __m128d test_cmp_ss(__m128 a, __m128 b) { |
| 43 | // Expects that the third argument in LLVM IR is immediate expression |
| 44 | // CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 13) |
| 45 | return _mm_cmp_ss(a, b, _CMP_GE_OS); |
| 46 | } |
| Manman Ren | 492d84c0 | 2013-06-17 19:42:49 +0000 | [diff] [blame] | 47 | |
| 48 | __m128 test_cmpgt_ss(__m128 a, __m128 b) { |
| 49 | // CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 1) |
| 50 | // CHECK: shufflevector <{{.*}}, <4 x i32> <i32 4, i32 1, i32 2, i32 3> |
| 51 | return _mm_cmpgt_ss(a, b); |
| 52 | } |
| 53 | |
| 54 | __m128 test_cmpge_ss(__m128 a, __m128 b) { |
| 55 | // CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 2) |
| 56 | // CHECK: shufflevector <{{.*}}, <4 x i32> <i32 4, i32 1, i32 2, i32 3> |
| 57 | return _mm_cmpge_ss(a, b); |
| 58 | } |
| 59 | |
| 60 | __m128 test_cmpngt_ss(__m128 a, __m128 b) { |
| 61 | // CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 5) |
| 62 | // CHECK: shufflevector <{{.*}}, <4 x i32> <i32 4, i32 1, i32 2, i32 3> |
| 63 | return _mm_cmpngt_ss(a, b); |
| 64 | } |
| 65 | |
| 66 | __m128 test_cmpnge_ss(__m128 a, __m128 b) { |
| 67 | // CHECK: @llvm.x86.sse.cmp.ss({{.*}}, i8 6) |
| 68 | // CHECK: shufflevector <{{.*}}, <4 x i32> <i32 4, i32 1, i32 2, i32 3> |
| 69 | return _mm_cmpnge_ss(a, b); |
| 70 | } |
| 71 | |
| 72 | __m128d test_cmpgt_sd(__m128d a, __m128d b) { |
| 73 | // CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 1) |
| 74 | // CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3> |
| 75 | return _mm_cmpgt_sd(a, b); |
| 76 | } |
| 77 | |
| 78 | __m128d test_cmpge_sd(__m128d a, __m128d b) { |
| 79 | // CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 2) |
| 80 | // CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3> |
| 81 | return _mm_cmpge_sd(a, b); |
| 82 | } |
| 83 | |
| 84 | __m128d test_cmpngt_sd(__m128d a, __m128d b) { |
| 85 | // CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 5) |
| 86 | // CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3> |
| 87 | return _mm_cmpngt_sd(a, b); |
| 88 | } |
| 89 | |
| 90 | __m128d test_cmpnge_sd(__m128d a, __m128d b) { |
| 91 | // CHECK: @llvm.x86.sse2.cmp.sd({{.*}}, i8 6) |
| 92 | // CHECK: shufflevector <{{.*}}, <2 x i32> <i32 0, i32 3> |
| 93 | return _mm_cmpnge_sd(a, b); |
| 94 | } |