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 shuffle instructions 10// 11 12__m256 x(__m256 a, __m256 b) { 13 // Check if the mask is correct 14 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 8, i32 11, i32 7, i32 6, i32 12, i32 15> 15 return _mm256_shuffle_ps(a, b, 203); 16} 17 18__m128d test_mm_permute_pd(__m128d a) { 19 // Check if the mask is correct 20 // CHECK: shufflevector{{.*}}<i32 1, i32 0> 21 return _mm_permute_pd(a, 1); 22} 23 24__m256d test_mm256_permute_pd(__m256d a) { 25 // Check if the mask is correct 26 // CHECK: shufflevector{{.*}}<i32 1, i32 0, i32 3, i32 2> 27 return _mm256_permute_pd(a, 5); 28} 29 30__m128 test_mm_permute_ps(__m128 a) { 31 // Check if the mask is correct 32 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0> 33 return _mm_permute_ps(a, 0x1b); 34} 35 36// Test case for PR12401 37__m128 test_mm_permute_ps2(__m128 a) { 38 // Check if the mask is correct 39 // CHECK: shufflevector{{.*}}<i32 2, i32 1, i32 2, i32 3> 40 return _mm_permute_ps(a, 0xe6); 41} 42 43__m256 test_mm256_permute_ps(__m256 a) { 44 // Check if the mask is correct 45 // CHECK: shufflevector{{.*}}<i32 3, i32 2, i32 1, i32 0, i32 7, i32 6, i32 5, i32 4> 46 return _mm256_permute_ps(a, 0x1b); 47} 48 49__m256d test_mm256_permute2f128_pd(__m256d a, __m256d b) { 50 // Check if the mask is correct 51 // CHECK: @llvm.x86.avx.vperm2f128.pd.256 52 return _mm256_permute2f128_pd(a, b, 0x31); 53} 54 55__m256 test_mm256_permute2f128_ps(__m256 a, __m256 b) { 56 // Check if the mask is correct 57 // CHECK: @llvm.x86.avx.vperm2f128.ps.256 58 return _mm256_permute2f128_ps(a, b, 0x13); 59} 60 61__m256i test_mm256_permute2f128_si256(__m256i a, __m256i b) { 62 // Check if the mask is correct 63 // CHECK: @llvm.x86.avx.vperm2f128.si.256 64 return _mm256_permute2f128_si256(a, b, 0x20); 65} 66