84 lines · c
1// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror | FileCheck %s2// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror | FileCheck %s3// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror | FileCheck %s4// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror | FileCheck %s5 6// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s7// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s8// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s9// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=i386-apple-darwin -target-feature +avx512bitalg -emit-llvm -o - -Wall -Werror -fexperimental-new-constant-interpreter | FileCheck %s10 11 12#include <immintrin.h>13#include "builtin_test_helpers.h"14 15__m512i test_mm512_popcnt_epi16(__m512i __A) {16 // CHECK-LABEL: test_mm512_popcnt_epi1617 // CHECK: @llvm.ctpop.v32i1618 return _mm512_popcnt_epi16(__A);19}20TEST_CONSTEXPR(match_v32hi(_mm512_popcnt_epi16((__m512i)(__v32hi){+5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025, +5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025}), 2, 15, 14, 1, 0, 8, 1, 9, 2, 2, 4, 2, 6, 2, 9, 2, 2, 15, 14, 1, 0, 8, 1, 9, 2, 2, 4, 2, 6, 2, 9, 2));21 22__m512i test_mm512_mask_popcnt_epi16(__m512i __A, __mmask32 __U, __m512i __B) {23 // CHECK-LABEL: test_mm512_mask_popcnt_epi1624 // CHECK: @llvm.ctpop.v32i1625 // CHECK: select <32 x i1> %{{[0-9]+}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}26 return _mm512_mask_popcnt_epi16(__A, __U, __B);27}28TEST_CONSTEXPR(match_v32hi(_mm512_mask_popcnt_epi16(_mm512_set1_epi16(-1), 0xF0F0F0F0, (__m512i)(__v32hi){+5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025, +5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025}), -1, -1, -1, -1, 0, 8, 1, 9, -1, -1, -1, -1, 6, 2, 9, 2, -1, -1, -1, -1, 0, 8, 1, 9, -1, -1, -1, -1, 6, 2, 9, 2));29 30__m512i test_mm512_maskz_popcnt_epi16(__mmask32 __U, __m512i __B) {31 // CHECK-LABEL: test_mm512_maskz_popcnt_epi1632 // CHECK: @llvm.ctpop.v32i1633 // CHECK: select <32 x i1> %{{[0-9]+}}, <32 x i16> %{{.*}}, <32 x i16> %{{.*}}34 return _mm512_maskz_popcnt_epi16(__U, __B);35}36TEST_CONSTEXPR(match_v32hi(_mm512_maskz_popcnt_epi16(0x0F0F0F0F, (__m512i)(__v32hi){+5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025, +5, -3, -10, +8, 0, -256, +256, -128, +3, +9, +15, +33, +63, +129, +511, +1025}), 2, 15, 14, 1, 0, 0, 0, 0, 2, 2, 4, 2, 0, 0, 0, 0, 2, 15, 14, 1, 0, 0, 0, 0, 2, 2, 4, 2, 0, 0, 0, 0));37 38__m512i test_mm512_popcnt_epi8(__m512i __A) {39 // CHECK-LABEL: test_mm512_popcnt_epi840 // CHECK: @llvm.ctpop.v64i841 return _mm512_popcnt_epi8(__A);42}43TEST_CONSTEXPR(match_v64qi(_mm512_popcnt_epi8((__m512i)(__v64qi){+5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73}), 2, 7, 6, 1, 0, 4, 1, 4, 2, 2, 4, 2, 6, 2, 4, 3, 2, 7, 6, 1, 0, 4, 1, 4, 2, 2, 4, 2, 6, 2, 4, 3, 2, 7, 6, 1, 0, 4, 1, 4, 2, 2, 4, 2, 6, 2, 4, 3, 2, 7, 6, 1, 0, 4, 1, 4, 2, 2, 4, 2, 6, 2, 4, 3));44 45__m512i test_mm512_mask_popcnt_epi8(__m512i __A, __mmask64 __U, __m512i __B) {46 // CHECK-LABEL: test_mm512_mask_popcnt_epi847 // CHECK: @llvm.ctpop.v64i848 // CHECK: select <64 x i1> %{{[0-9]+}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}49 return _mm512_mask_popcnt_epi8(__A, __U, __B);50}51TEST_CONSTEXPR(match_v64qi(_mm512_mask_popcnt_epi8(_mm512_set1_epi8(-1), 0xF0F0F0F00F0F0F0FULL, (__m512i)(__v64qi){+5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73}), 2, 7, 6, 1, -1, -1, -1, -1, 2, 2, 4, 2, -1, -1, -1, -1, 2, 7, 6, 1, -1, -1, -1, -1, 2, 2, 4, 2, -1, -1, -1, -1, -1, -1, -1, -1, 0, 4, 1, 4, -1, -1, -1, -1, 6, 2, 4, 3, -1, -1, -1, -1, 0, 4, 1, 4, -1, -1, -1, -1, 6, 2, 4, 3));52 53__m512i test_mm512_maskz_popcnt_epi8(__mmask64 __U, __m512i __B) {54 // CHECK-LABEL: test_mm512_maskz_popcnt_epi855 // CHECK: @llvm.ctpop.v64i856 // CHECK: select <64 x i1> %{{[0-9]+}}, <64 x i8> %{{.*}}, <64 x i8> %{{.*}}57 return _mm512_maskz_popcnt_epi8(__U, __B);58}59TEST_CONSTEXPR(match_v64qi(_mm512_maskz_popcnt_epi8(0x0F0F0F0FF0F0F0F0ULL, (__m512i)(__v64qi){+5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73, +5, -3, -10, +8, 0, -16, +16, -16, +3, +9, +15, +33, +63, +33, +53, +73}), 0, 0, 0, 0, 0, 4, 1, 4, 0, 0, 0, 0, 6, 2, 4, 3, 0, 0, 0, 0, 0, 4, 1, 4, 0, 0, 0, 0, 6, 2, 4, 3, 2, 7, 6, 1, 0, 0, 0, 0, 2, 2, 4, 2, 0, 0, 0, 0, 2, 7, 6, 1, 0, 0, 0, 0, 2, 2, 4, 2, 0, 0, 0, 0));60 61__mmask64 test_mm512_mask_bitshuffle_epi64_mask(__mmask64 __U, __m512i __A, __m512i __B) {62 // CHECK-LABEL: test_mm512_mask_bitshuffle_epi64_mask63 // CHECK: @llvm.x86.avx512.vpshufbitqmb.51264 // CHECK: and <64 x i1> %{{.*}}, %{{.*}}65 return _mm512_mask_bitshuffle_epi64_mask(__U, __A, __B);66}67 68__mmask64 test_mm512_bitshuffle_epi64_mask(__m512i __A, __m512i __B) {69 // CHECK-LABEL: test_mm512_bitshuffle_epi64_mask70 // CHECK: @llvm.x86.avx512.vpshufbitqmb.51271 return _mm512_bitshuffle_epi64_mask(__A, __B);72}73TEST_CONSTEXPR(_mm512_bitshuffle_epi64_mask(74 (__m512i)(__v64qi){1,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,-128, -1,0,0,0,0,0,0,0, 85,85,85,85,85,85,85,85,75 1,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,-128, -1,0,0,0,0,0,0,0, 85,85,85,85,85,85,85,85},76 (__m512i)(__v64qi){0,1,2,3,4,5,6,7, 63,62,61,60,59,58,57,56, 0,1,2,3,4,5,6,7, 0,1,2,3,4,5,6,7,77 0,1,2,3,4,5,6,7, 63,62,61,60,59,58,57,56, 0,1,2,3,4,5,6,7, 0,1,2,3,4,5,6,7}) == 0x55ff010155ff0101ULL);78 79TEST_CONSTEXPR(_mm512_mask_bitshuffle_epi64_mask(0xFFFFFFFF00000000ULL,80 (__m512i)(__v64qi){1,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,-128, -1,0,0,0,0,0,0,0, 85,85,85,85,85,85,85,85,81 1,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,-128, -1,0,0,0,0,0,0,0, 85,85,85,85,85,85,85,85},82 (__m512i)(__v64qi){0,1,2,3,4,5,6,7, 63,62,61,60,59,58,57,56, 0,1,2,3,4,5,6,7, 0,1,2,3,4,5,6,7,83 0,1,2,3,4,5,6,7, 63,62,61,60,59,58,57,56, 0,1,2,3,4,5,6,7, 0,1,2,3,4,5,6,7}) == 0x55ff010100000000ULL);84