94 lines · c
1// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -target-feature +avx -target-feature +avx2 -target-feature +avx512f -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefix=CHECK2// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -target-feature +avx -target-feature +avx2 -target-feature +avx512f -fno-signed-char -emit-llvm -o - -Wall -Werror | FileCheck %s --check-prefix=CHECK3 4// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -target-feature +avx -target-feature +avx2 -target-feature +avx512f -emit-llvm -o - -Wall -Werror -fmax-type-align=16 | FileCheck %s --check-prefix=CHECK5// RUN: %clang_cc1 -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +sse4.1 -target-feature +avx -target-feature +avx2 -target-feature +avx512f -fno-signed-char -emit-llvm -o - -Wall -Werror -fmax-type-align=16 | FileCheck %s --check-prefix=CHECK6 7#include <immintrin.h>8 9// (PR33830) Tests ensure the correct alignment of non-temporal load/stores on darwin targets where fmax-type-align is set to 16.10 11//12// 128-bit vectors13//14 15void test_mm_stream_pd(double* A, __m128d B) {16 // CHECK-LABEL: test_mm_stream_pd17 // CHECK: store <2 x double> %{{.*}}, ptr %{{.*}}, align 16, !nontemporal18 _mm_stream_pd(A, B);19}20 21void test_mm_stream_ps(float* A, __m128 B) {22 // CHECK16-LABEL: test_mm_stream_ps23 // CHECK16: store <4 x float> %{{.*}}, ptr %{{.*}}, align 16, !nontemporal24 _mm_stream_ps(A, B);25}26 27void test_mm_stream_si128(__m128i* A, __m128i B) {28 // CHECK-LABEL: test_mm_stream_si12829 // CHECK: store <2 x i64> %{{.*}}, ptr %{{.*}}, align 16, !nontemporal30 _mm_stream_si128(A, B);31}32 33__m128i test_mm_stream_load_si128(__m128i const *A) {34 // CHECK-LABEL: test_mm_stream_load_si12835 // CHECK: load <2 x i64>, ptr %{{.*}}, align 16, !nontemporal36 return _mm_stream_load_si128(A);37}38 39//40// 256-bit vectors41//42 43void test_mm256_stream_pd(double* A, __m256d B) {44 // CHECK-LABEL: test_mm256_stream_pd45 // CHECK: store <4 x double> %{{.*}}, ptr %{{.*}}, align 32, !nontemporal46 _mm256_stream_pd(A, B);47}48 49void test_mm256_stream_ps(float* A, __m256 B) {50 // CHECK-LABEL: test_mm256_stream_ps51 // CHECK: store <8 x float> %{{.*}}, ptr %{{.*}}, align 32, !nontemporal52 _mm256_stream_ps(A, B);53}54 55void test_mm256_stream_si256(__m256i* A, __m256i B) {56 // CHECK-LABEL: test_mm256_stream_si25657 // CHECK: store <4 x i64> %{{.*}}, ptr %{{.*}}, align 32, !nontemporal58 _mm256_stream_si256(A, B);59}60 61__m256i test_mm256_stream_load_si256(__m256i const *A) {62 // CHECK-LABEL: test_mm256_stream_load_si25663 // CHECK: load <4 x i64>, ptr %{{.*}}, align 32, !nontemporal64 return _mm256_stream_load_si256(A);65}66 67//68// 512-bit vectors69//70 71void test_mm512_stream_pd(double* A, __m512d B) {72 // CHECK-LABEL: test_mm512_stream_pd73 // CHECK: store <8 x double> %{{.*}}, ptr %{{.*}}, align 64, !nontemporal74 _mm512_stream_pd(A, B);75}76 77void test_mm512_stream_ps(float* A, __m512 B) {78 // CHECK-LABEL: test_mm512_stream_ps79 // CHECK: store <16 x float> %{{.*}}, ptr %{{.*}}, align 64, !nontemporal80 _mm512_stream_ps(A, B);81}82 83void test_mm512_stream_si512(__m512i* A, __m512i B) {84 // CHECK-LABEL: test_mm512_stream_si51285 // CHECK: store <8 x i64> %{{.*}}, ptr %{{.*}}, align 64, !nontemporal86 _mm512_stream_si512(A, B);87}88 89__m512i test_mm512_stream_load_si512(void *A) {90 // CHECK-LABEL: test_mm512_stream_load_si51291 // CHECK: load <8 x i64>, ptr %{{.*}}, align 64, !nontemporal92 return _mm512_stream_load_si512(A);93}94