summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorCraig Topper <craig.topper@gmail.com>2016-12-27 03:46:16 +0000
committerCraig Topper <craig.topper@gmail.com>2016-12-27 03:46:16 +0000
commit6ad5bcc8ac63298ae79ee9bef22f33bbd28bc363 (patch)
tree6d2cc83ff1ad960de3687cf2d6e37c503d552494
parent39b9e324933813122cb119cf1d39d6a472acfcc9 (diff)
downloadbcm5719-llvm-6ad5bcc8ac63298ae79ee9bef22f33bbd28bc363.tar.gz
bcm5719-llvm-6ad5bcc8ac63298ae79ee9bef22f33bbd28bc363.zip
[AVX-512] Replace masked 512-bit pmuldq and pmuludq builtins with the newly added unmasked versions and selects.
llvm-svn: 290575
-rw-r--r--clang/lib/Headers/avx512fintrin.h8
-rw-r--r--clang/test/CodeGen/avx512f-builtins.c24
2 files changed, 24 insertions, 8 deletions
diff --git a/clang/lib/Headers/avx512fintrin.h b/clang/lib/Headers/avx512fintrin.h
index 30ebfc0430f..e6a7217c896 100644
--- a/clang/lib/Headers/avx512fintrin.h
+++ b/clang/lib/Headers/avx512fintrin.h
@@ -1422,7 +1422,7 @@ _mm512_mul_epi32(__m512i __X, __m512i __Y)
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_mask_mul_epi32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_mul_epi32(__X, __Y),
(__v8di)__W);
}
@@ -1430,7 +1430,7 @@ _mm512_mask_mul_epi32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_mul_epi32(__mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_mul_epi32(__X, __Y),
(__v8di)_mm512_setzero_si512 ());
}
@@ -1444,7 +1444,7 @@ _mm512_mul_epu32(__m512i __X, __m512i __Y)
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_mask_mul_epu32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_mul_epu32(__X, __Y),
(__v8di)__W);
}
@@ -1452,7 +1452,7 @@ _mm512_mask_mul_epu32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_mul_epu32(__mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_mul_epu32(__X, __Y),
(__v8di)_mm512_setzero_si512 ());
}
diff --git a/clang/test/CodeGen/avx512f-builtins.c b/clang/test/CodeGen/avx512f-builtins.c
index ebd162835fc..760783af1ce 100644
--- a/clang/test/CodeGen/avx512f-builtins.c
+++ b/clang/test/CodeGen/avx512f-builtins.c
@@ -1867,29 +1867,45 @@ __m512i test_mm512_add_epi64(__m512i __A, __m512i __B) {
return _mm512_add_epi64(__A,__B);
}
+__m512i test_mm512_mul_epi32(__m512i __A, __m512i __B) {
+ //CHECK-LABEL: @test_mm512_mul_epi32
+ //CHECK: @llvm.x86.avx512.pmul.dq.512
+ return _mm512_mul_epi32(__A,__B);
+}
+
__m512i test_mm512_maskz_mul_epi32 (__mmask16 __k,__m512i __A, __m512i __B) {
//CHECK-LABEL: @test_mm512_maskz_mul_epi32
- //CHECK: @llvm.x86.avx512.mask.pmul.dq.512
+ //CHECK: @llvm.x86.avx512.pmul.dq.512
+ //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_maskz_mul_epi32(__k,__A,__B);
}
__m512i test_mm512_mask_mul_epi32 (__mmask16 __k,__m512i __A, __m512i __B,
__m512i __src) {
//CHECK-LABEL: @test_mm512_mask_mul_epi32
- //CHECK: @llvm.x86.avx512.mask.pmul.dq.512
+ //CHECK: @llvm.x86.avx512.pmul.dq.512
+ //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_mask_mul_epi32(__src,__k,__A,__B);
}
+__m512i test_mm512_mul_epu32 (__m512i __A, __m512i __B) {
+ //CHECK-LABEL: @test_mm512_mul_epu32
+ //CHECK: @llvm.x86.avx512.pmulu.dq.512
+ return _mm512_mul_epu32(__A,__B);
+}
+
__m512i test_mm512_maskz_mul_epu32 (__mmask16 __k,__m512i __A, __m512i __B) {
//CHECK-LABEL: @test_mm512_maskz_mul_epu32
- //CHECK: @llvm.x86.avx512.mask.pmulu.dq.512
+ //CHECK: @llvm.x86.avx512.pmulu.dq.512
+ //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_maskz_mul_epu32(__k,__A,__B);
}
__m512i test_mm512_mask_mul_epu32 (__mmask16 __k,__m512i __A, __m512i __B,
__m512i __src) {
//CHECK-LABEL: @test_mm512_mask_mul_epu32
- //CHECK: @llvm.x86.avx512.mask.pmulu.dq.512
+ //CHECK: @llvm.x86.avx512.pmulu.dq.512
+ //CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
return _mm512_mask_mul_epu32(__src,__k,__A,__B);
}
OpenPOWER on IntegriCloud