Author: ctopper
Date: Mon Dec 26 22:04:57 2016
New Revision: 290580
URL: http://llvm.org/viewvc/llvm-project?rev=290580&view=rev
Log:
[AVX-512] Replace masked 512-bit pmuldq and pmuludq builtins with the newly
added unmasked versions and selects.
Modified:
cfe/trunk/include/clang/Basic/BuiltinsX86.def
cfe/trunk/lib/Headers/avx512fintrin.h
cfe/trunk/test/CodeGen/avx512f-builtins.c
Modified: cfe/trunk/include/clang/Basic/BuiltinsX86.def
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/include/clang/Basic/BuiltinsX86.def?rev=290580&r1=290579&r2=290580&view=diff
==============================================================================
--- cfe/trunk/include/clang/Basic/BuiltinsX86.def (original)
+++ cfe/trunk/include/clang/Basic/BuiltinsX86.def Mon Dec 26 22:04:57 2016
@@ -974,8 +974,8 @@ TARGET_BUILTIN(__builtin_ia32_pminsd512_
TARGET_BUILTIN(__builtin_ia32_pminsq512_mask, "V8LLiV8LLiV8LLiV8LLiUc", "",
"avx512f")
TARGET_BUILTIN(__builtin_ia32_pminud512_mask, "V16iV16iV16iV16iUs", "",
"avx512f")
TARGET_BUILTIN(__builtin_ia32_pminuq512_mask, "V8LLiV8LLiV8LLiV8LLiUc", "",
"avx512f")
-TARGET_BUILTIN(__builtin_ia32_pmuldq512_mask, "V8LLiV16iV16iV8LLiUc", "",
"avx512f")
-TARGET_BUILTIN(__builtin_ia32_pmuludq512_mask, "V8LLiV16iV16iV8LLiUc", "",
"avx512f")
+TARGET_BUILTIN(__builtin_ia32_pmuldq512, "V8LLiV16iV16i", "", "avx512f")
+TARGET_BUILTIN(__builtin_ia32_pmuludq512, "V8LLiV16iV16i", "", "avx512f")
TARGET_BUILTIN(__builtin_ia32_ptestmd512, "UsV16iV16iUs", "", "avx512f")
TARGET_BUILTIN(__builtin_ia32_ptestmq512, "UcV8LLiV8LLiUc", "", "avx512f")
TARGET_BUILTIN(__builtin_ia32_pbroadcastd512_gpr_mask, "V16iiV16iUs", "",
"avx512f")
Modified: cfe/trunk/lib/Headers/avx512fintrin.h
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Headers/avx512fintrin.h?rev=290580&r1=290579&r2=290580&view=diff
==============================================================================
--- cfe/trunk/lib/Headers/avx512fintrin.h (original)
+++ cfe/trunk/lib/Headers/avx512fintrin.h Mon Dec 26 22:04:57 2016
@@ -1416,57 +1416,45 @@ _mm512_maskz_min_epu64 (__mmask8 __M, __
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_mul_epi32(__m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuldq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di)
- _mm512_setzero_si512 (),
- (__mmask8) -1);
+ return (__m512i)__builtin_ia32_pmuldq512((__v16si)__X, (__v16si) __Y);
}
static __inline __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_mul_epi32 (__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
+_mm512_mask_mul_epi32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuldq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di) __W, __M);
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
+ (__v8di)_mm512_mul_epi32(__X,
__Y),
+ (__v8di)__W);
}
static __inline __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_mul_epi32 (__mmask8 __M, __m512i __X, __m512i __Y)
+_mm512_maskz_mul_epi32(__mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuldq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di)
- _mm512_setzero_si512 (),
- __M);
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
+ (__v8di)_mm512_mul_epi32(__X,
__Y),
+ (__v8di)_mm512_setzero_si512 ());
}
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_mul_epu32(__m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuludq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di)
- _mm512_setzero_si512 (),
- (__mmask8) -1);
+ return (__m512i)__builtin_ia32_pmuludq512((__v16si)__X, (__v16si)__Y);
}
static __inline __m512i __DEFAULT_FN_ATTRS
-_mm512_mask_mul_epu32 (__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
+_mm512_mask_mul_epu32(__m512i __W, __mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuludq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di) __W, __M);
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
+ (__v8di)_mm512_mul_epu32(__X,
__Y),
+ (__v8di)__W);
}
static __inline __m512i __DEFAULT_FN_ATTRS
-_mm512_maskz_mul_epu32 (__mmask8 __M, __m512i __X, __m512i __Y)
+_mm512_maskz_mul_epu32(__mmask8 __M, __m512i __X, __m512i __Y)
{
- return (__m512i) __builtin_ia32_pmuludq512_mask ((__v16si) __X,
- (__v16si) __Y,
- (__v8di)
- _mm512_setzero_si512 (),
- __M);
+ return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
+ (__v8di)_mm512_mul_epu32(__X,
__Y),
+ (__v8di)_mm512_setzero_si512 ());
}
static __inline __m512i __DEFAULT_FN_ATTRS
Modified: cfe/trunk/test/CodeGen/avx512f-builtins.c
URL:
http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/avx512f-builtins.c?rev=290580&r1=290579&r2=290580&view=diff
==============================================================================
--- cfe/trunk/test/CodeGen/avx512f-builtins.c (original)
+++ cfe/trunk/test/CodeGen/avx512f-builtins.c Mon Dec 26 22:04:57 2016
@@ -1867,29 +1867,45 @@ __m512i test_mm512_add_epi64(__m512i __A
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);
}
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
