[CLANG] [AVX512] [BUILTIN] Adding PSRA{Q|D|QI|DI}{128|256|512} builtin

Differential Revision: http://reviews.llvm.org/D17693

llvm-svn: 262321
This commit is contained in:
Michael Zuckerman 2016-03-01 11:38:16 +00:00
parent 7878888690
commit 1ac360cca4
4 changed files with 219 additions and 1 deletions

View File

@ -1651,7 +1651,16 @@ TARGET_BUILTIN(__builtin_ia32_psllv2di_mask, "V2LLiV2LLiV2LLiV2LLiUc","","avx512
TARGET_BUILTIN(__builtin_ia32_psllv4di_mask, "V4LLiV4LLiV4LLiV4LLiUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllv4si_mask, "V4iV4iV4iV4iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllv8si_mask, "V8iV8iV8iV8iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_pslldi512_mask, "V16iV16iiV16iUs","","avx512f")
TARGET_BUILTIN(__builtin_ia32_psllqi512_mask, "V8LLiV8LLiiV8LLiUc","","avx512f")
TARGET_BUILTIN(__builtin_ia32_pslld128_mask, "V4iV4iV4iV4iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_pslld256_mask, "V8iV8iV4iV8iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_pslldi128_mask, "V4iV4iiV4iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_pslldi256_mask, "V8iV8iiV8iUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllq128_mask, "V2LLiV2LLiV2LLiV2LLiUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllq256_mask, "V4LLiV4LLiV2LLiV4LLiUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllqi128_mask, "V2LLiV2LLiiV2LLiUc","","avx512vl")
TARGET_BUILTIN(__builtin_ia32_psllqi256_mask, "V4LLiV4LLiiV4LLiUc","","avx512vl")
#undef BUILTIN
#undef TARGET_BUILTIN

View File

@ -3520,6 +3520,46 @@ __builtin_ia32_prorq512_mask ((__v8di)( __A),( __B),\
(__mmask8)( __U));\
})
#define _mm512_slli_epi32( __A, __B) __extension__ ({ \
__builtin_ia32_pslldi512_mask ((__v16si)( __A),( __B),\
(__v16si)\
_mm512_setzero_si512 (),\
(__mmask16) -1);\
})
#define _mm512_mask_slli_epi32( __W, __U, __A ,__B) __extension__ ({ \
__builtin_ia32_pslldi512_mask ((__v16si) (__A), (__B),\
(__v16si)( __W),\
(__mmask16)( __U));\
})
#define _mm512_maskz_slli_epi32( __U, __A, __B) __extension__ ({ \
__builtin_ia32_pslldi512_mask ((__v16si)( __A),( __B),\
(__v16si)\
_mm512_setzero_si512 (),\
(__mmask16)( __U));\
})
#define _mm512_slli_epi64( __A, __B) __extension__ ({ \
__builtin_ia32_psllqi512_mask ((__v8di)( __A),( __B),\
(__v8di)\
_mm512_setzero_si512 (),\
(__mmask8) -1);\
})
#define _mm512_mask_slli_epi64( __W, __U, __A ,__B) __extension__ ({ \
__builtin_ia32_psllqi512_mask ((__v8di) (__A), (__B),\
(__v8di)( __W),\
(__mmask8)( __U));\
})
#define _mm512_maskz_slli_epi64( __U, __A, __B) __extension__ ({ \
__builtin_ia32_psllqi512_mask ((__v8di)( __A),( __B),\
(__v8di)\
_mm512_setzero_si512 (),\
(__mmask8)( __U));\
})
#undef __DEFAULT_FN_ATTRS

View File

@ -5215,6 +5215,138 @@ __builtin_ia32_prorq256_mask ((__v4di)( __A),( __B),\
(__mmask8)( __U));\
})
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_sll_epi32 (__m128i __W, __mmask8 __U, __m128i __A,
__m128i __B)
{
return (__m128i) __builtin_ia32_pslld128_mask ((__v4si) __A,
(__v4si) __B,
(__v4si) __W,
(__mmask8) __U);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_sll_epi32 (__mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i) __builtin_ia32_pslld128_mask ((__v4si) __A,
(__v4si) __B,
(__v4si)
_mm_setzero_si128 (),
(__mmask8) __U);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_mask_sll_epi32 (__m256i __W, __mmask8 __U, __m256i __A,
__m128i __B)
{
return (__m256i) __builtin_ia32_pslld256_mask ((__v8si) __A,
(__v4si) __B,
(__v8si) __W,
(__mmask8) __U);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_sll_epi32 (__mmask8 __U, __m256i __A, __m128i __B)
{
return (__m256i) __builtin_ia32_pslld256_mask ((__v8si) __A,
(__v4si) __B,
(__v8si)
_mm256_setzero_si256 (),
(__mmask8) __U);
}
#define _mm_mask_slli_epi32( __W, __U, __A, __B) __extension__ ({ \
__builtin_ia32_pslldi128_mask ((__v4si)( __A),( __B),\
(__v4si)( __W),\
(__mmask8)( __U));\
})
#define _mm_maskz_slli_epi32( __U, __A, __B) __extension__ ({ \
__builtin_ia32_pslldi128_mask ((__v4si)( __A),( __B),\
(__v4si)\
_mm_setzero_si128 (),\
(__mmask8)( __U));\
})
#define _mm256_mask_slli_epi32( __W, __U, __A ,__B) __extension__ ({ \
__builtin_ia32_pslldi256_mask ((__v8si) (__A), (__B),\
(__v8si)( __W),\
(__mmask8)( __U));\
})
#define _mm256_maskz_slli_epi32( __U, __A, __B) __extension__ ({ \
__builtin_ia32_pslldi256_mask ((__v8si)( __A),( __B),\
(__v8si)\
_mm256_setzero_si256 (),\
(__mmask8)( __U));\
})
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_sll_epi64 (__m128i __W, __mmask8 __U, __m128i __A,
__m128i __B)
{
return (__m128i) __builtin_ia32_psllq128_mask ((__v2di) __A,
(__v2di) __B,
(__v2di) __W,
(__mmask8) __U);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_sll_epi64 (__mmask8 __U, __m128i __A, __m128i __B)
{
return (__m128i) __builtin_ia32_psllq128_mask ((__v2di) __A,
(__v2di) __B,
(__v2di)
_mm_setzero_di (),
(__mmask8) __U);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_mask_sll_epi64 (__m256i __W, __mmask8 __U, __m256i __A,
__m128i __B)
{
return (__m256i) __builtin_ia32_psllq256_mask ((__v4di) __A,
(__v2di) __B,
(__v4di) __W,
(__mmask8) __U);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_sll_epi64 (__mmask8 __U, __m256i __A, __m128i __B)
{
return (__m256i) __builtin_ia32_psllq256_mask ((__v4di) __A,
(__v2di) __B,
(__v4di)
_mm256_setzero_si256 (),
(__mmask8) __U);
}
#define _mm_mask_slli_epi64( __W, __U, __A, __B) __extension__ ({ \
__builtin_ia32_psllqi128_mask ((__v2di)( __A),( __B),\
(__v2di)( __W),\
(__mmask8)( __U));\
})
#define _mm_maskz_slli_epi64( __U, __A, __B) __extension__ ({ \
__builtin_ia32_psllqi128_mask ((__v2di)( __A),( __B),\
(__v2di)\
_mm_setzero_di (),\
(__mmask8)( __U));\
})
#define _mm256_mask_slli_epi64( __W, __U, __A ,__B) __extension__ ({ \
__builtin_ia32_psllqi256_mask ((__v4di) (__A), (__B),\
(__v4di)( __W),\
(__mmask8)( __U));\
})
#define _mm256_maskz_slli_epi64( __U, __A, __B) __extension__ ({ \
__builtin_ia32_psllqi256_mask ((__v4di)( __A),( __B),\
(__v4di)\
_mm256_setzero_si256 (),\
(__mmask8)( __U));\
})
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_rorv_epi32 (__m128i __A, __m128i __B)

View File

@ -2226,3 +2226,40 @@ __m512i test_mm512_maskz_rorv_epi64(__mmask8 __U, __m512i __A, __m512i __B) {
return _mm512_maskz_rorv_epi64(__U, __A, __B);
}
__m512i test_mm512_slli_epi32(__m512i __A) {
// CHECK-LABEL: @test_mm512_slli_epi32
// CHECK: @llvm.x86.avx512.mask.psll.di.512
return _mm512_slli_epi32(__A, 5);
}
__m512i test_mm512_mask_slli_epi32(__m512i __W, __mmask16 __U, __m512i __A) {
// CHECK-LABEL: @test_mm512_mask_slli_epi32
// CHECK: @llvm.x86.avx512.mask.psll.di.512
return _mm512_mask_slli_epi32(__W, __U, __A, 5);
}
__m512i test_mm512_maskz_slli_epi32(__mmask16 __U, __m512i __A) {
// CHECK-LABEL: @test_mm512_maskz_slli_epi32
// CHECK: @llvm.x86.avx512.mask.psll.di.512
return _mm512_maskz_slli_epi32(__U, __A, 5);
}
__m512i test_mm512_slli_epi64(__m512i __A) {
// CHECK-LABEL: @test_mm512_slli_epi64
// CHECK: @llvm.x86.avx512.mask.psll.qi.512
return _mm512_slli_epi64(__A, 5);
}
__m512i test_mm512_mask_slli_epi64(__m512i __W, __mmask8 __U, __m512i __A) {
// CHECK-LABEL: @test_mm512_mask_slli_epi64
// CHECK: @llvm.x86.avx512.mask.psll.qi.512
return _mm512_mask_slli_epi64(__W, __U, __A, 5);
}
__m512i test_mm512_maskz_slli_epi64(__mmask8 __U, __m512i __A) {
// CHECK-LABEL: @test_mm512_maskz_slli_epi64
// CHECK: @llvm.x86.avx512.mask.psll.qi.512
return _mm512_maskz_slli_epi64(__U, __A, 5);
}