diff options
author | Craig Topper <craig.topper@intel.com> | 2018-08-16 07:28:06 +0000 |
---|---|---|
committer | Craig Topper <craig.topper@intel.com> | 2018-08-16 07:28:06 +0000 |
commit | ccacccb1ddeae075492b5f5af576f8ca0056b9d6 (patch) | |
tree | 08d60668bff7c6b9e71eddb6e5dfda518b3aac68 | |
parent | 347daedef2a65151b13d1e6c6feb813b6d9b0c93 (diff) |
[X86] Remove masking from the 512-bit paddus/psubus builtins. Use a select builtin instead.
git-svn-id: https://llvm.org/svn/llvm-project/cfe/trunk@339845 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r-- | include/clang/Basic/BuiltinsX86.def | 8 | ||||
-rw-r--r-- | lib/CodeGen/CGBuiltin.cpp | 14 | ||||
-rw-r--r-- | lib/Headers/avx512bwintrin.h | 88 |
3 files changed, 40 insertions, 70 deletions
diff --git a/include/clang/Basic/BuiltinsX86.def b/include/clang/Basic/BuiltinsX86.def index 2d0b96ea59..98de5fa0b8 100644 --- a/include/clang/Basic/BuiltinsX86.def +++ b/include/clang/Basic/BuiltinsX86.def @@ -1040,8 +1040,8 @@ TARGET_BUILTIN(__builtin_ia32_packusdw512, "V32sV16iV16i", "ncV:512:", "avx512bw TARGET_BUILTIN(__builtin_ia32_packuswb512, "V64cV32sV32s", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_paddsb512, "V64cV64cV64c", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_paddsw512, "V32sV32sV32s", "ncV:512:", "avx512bw") -TARGET_BUILTIN(__builtin_ia32_paddusb512_mask, "V64cV64cV64cV64cULLi", "ncV:512:", "avx512bw") -TARGET_BUILTIN(__builtin_ia32_paddusw512_mask, "V32sV32sV32sV32sUi", "ncV:512:", "avx512bw") +TARGET_BUILTIN(__builtin_ia32_paddusb512, "V64cV64cV64c", "ncV:512:", "avx512bw") +TARGET_BUILTIN(__builtin_ia32_paddusw512, "V32sV32sV32s", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_pmaxsb512, "V64cV64cV64c", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_pmaxsw512, "V32sV32sV32s", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_pmaxub512, "V64cV64cV64c", "ncV:512:", "avx512bw") @@ -1053,8 +1053,8 @@ TARGET_BUILTIN(__builtin_ia32_pminuw512, "V32sV32sV32s", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_pshufb512, "V64cV64cV64c", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_psubsb512, "V64cV64cV64c", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_psubsw512, "V32sV32sV32s", "ncV:512:", "avx512bw") -TARGET_BUILTIN(__builtin_ia32_psubusb512_mask, "V64cV64cV64cV64cULLi", "ncV:512:", "avx512bw") -TARGET_BUILTIN(__builtin_ia32_psubusw512_mask, "V32sV32sV32sV32sUi", "ncV:512:", "avx512bw") +TARGET_BUILTIN(__builtin_ia32_psubusb512, "V64cV64cV64c", "ncV:512:", "avx512bw") +TARGET_BUILTIN(__builtin_ia32_psubusw512, "V32sV32sV32s", "ncV:512:", "avx512bw") TARGET_BUILTIN(__builtin_ia32_vpconflictdi_128_mask, "V2LLiV2LLiV2LLiUc", "ncV:128:", "avx512cd,avx512vl") TARGET_BUILTIN(__builtin_ia32_vpconflictdi_256_mask, "V4LLiV4LLiV4LLiUc", "ncV:256:", "avx512cd,avx512vl") diff --git a/lib/CodeGen/CGBuiltin.cpp b/lib/CodeGen/CGBuiltin.cpp index bbd2904e39..22a9c2f275 100644 --- a/lib/CodeGen/CGBuiltin.cpp +++ b/lib/CodeGen/CGBuiltin.cpp @@ -8931,12 +8931,6 @@ static Value *EmitX86AddSubSatExpr(CodeGenFunction &CGF, const CallExpr *E, Res = CGF.Builder.CreateSub(Select, Ops[1]); } - if (E->getNumArgs() == 4) { // For masked intrinsics. - Value *VecSRC = Ops[2]; - Value *Mask = Ops[3]; - return EmitX86Select(CGF, Mask, Res, VecSRC); - } - return Res; } @@ -10563,15 +10557,15 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID, Load->setVolatile(true); return Load; } - case X86::BI__builtin_ia32_paddusb512_mask: - case X86::BI__builtin_ia32_paddusw512_mask: + case X86::BI__builtin_ia32_paddusb512: + case X86::BI__builtin_ia32_paddusw512: case X86::BI__builtin_ia32_paddusb256: case X86::BI__builtin_ia32_paddusw256: case X86::BI__builtin_ia32_paddusb128: case X86::BI__builtin_ia32_paddusw128: return EmitX86AddSubSatExpr(*this, E, Ops, true /* IsAddition */); - case X86::BI__builtin_ia32_psubusb512_mask: - case X86::BI__builtin_ia32_psubusw512_mask: + case X86::BI__builtin_ia32_psubusb512: + case X86::BI__builtin_ia32_psubusw512: case X86::BI__builtin_ia32_psubusb256: case X86::BI__builtin_ia32_psubusw256: case X86::BI__builtin_ia32_psubusb128: diff --git a/lib/Headers/avx512bwintrin.h b/lib/Headers/avx512bwintrin.h index 7eda8a6dde..96ae19bf9c 100644 --- a/lib/Headers/avx512bwintrin.h +++ b/lib/Headers/avx512bwintrin.h @@ -466,57 +466,45 @@ _mm512_maskz_adds_epi16 (__mmask32 __U, __m512i __A, __m512i __B) static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_adds_epu8 (__m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) _mm512_setzero_si512(), - (__mmask64) -1); + return (__m512i)__builtin_ia32_paddusb512((__v64qi) __A, (__v64qi) __B); } static __inline__ __m512i __DEFAULT_FN_ATTRS -_mm512_mask_adds_epu8 (__m512i __W, __mmask64 __U, __m512i __A, - __m512i __B) +_mm512_mask_adds_epu8 (__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) __W, - (__mmask64) __U); + return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U, + (__v64qi)_mm512_adds_epu8(__A, __B), + (__v64qi)__W); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_maskz_adds_epu8 (__mmask64 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) _mm512_setzero_si512(), - (__mmask64) __U); + return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U, + (__v64qi)_mm512_adds_epu8(__A, __B), + (__v64qi)_mm512_setzero_si512()); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_adds_epu16 (__m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) _mm512_setzero_si512(), - (__mmask32) -1); + return (__m512i)__builtin_ia32_paddusw512((__v32hi) __A, (__v32hi) __B); } static __inline__ __m512i __DEFAULT_FN_ATTRS -_mm512_mask_adds_epu16 (__m512i __W, __mmask32 __U, __m512i __A, - __m512i __B) +_mm512_mask_adds_epu16 (__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) __W, - (__mmask32) __U); + return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U, + (__v32hi)_mm512_adds_epu16(__A, __B), + (__v32hi)__W); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_maskz_adds_epu16 (__mmask32 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_paddusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) _mm512_setzero_si512(), - (__mmask32) __U); + return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U, + (__v32hi)_mm512_adds_epu16(__A, __B), + (__v32hi)_mm512_setzero_si512()); } static __inline__ __m512i __DEFAULT_FN_ATTRS @@ -819,57 +807,45 @@ _mm512_maskz_subs_epi16 (__mmask32 __U, __m512i __A, __m512i __B) static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_subs_epu8 (__m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) _mm512_setzero_si512(), - (__mmask64) -1); + return (__m512i)__builtin_ia32_psubusb512((__v64qi) __A, (__v64qi) __B); } static __inline__ __m512i __DEFAULT_FN_ATTRS -_mm512_mask_subs_epu8 (__m512i __W, __mmask64 __U, __m512i __A, - __m512i __B) +_mm512_mask_subs_epu8 (__m512i __W, __mmask64 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) __W, - (__mmask64) __U); + return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U, + (__v64qi)_mm512_subs_epu8(__A, __B), + (__v64qi)__W); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_maskz_subs_epu8 (__mmask64 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusb512_mask ((__v64qi) __A, - (__v64qi) __B, - (__v64qi) _mm512_setzero_si512(), - (__mmask64) __U); + return (__m512i)__builtin_ia32_selectb_512((__mmask64)__U, + (__v64qi)_mm512_subs_epu8(__A, __B), + (__v64qi)_mm512_setzero_si512()); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_subs_epu16 (__m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) _mm512_setzero_si512(), - (__mmask32) -1); + return (__m512i)__builtin_ia32_psubusw512((__v32hi) __A, (__v32hi) __B); } static __inline__ __m512i __DEFAULT_FN_ATTRS -_mm512_mask_subs_epu16 (__m512i __W, __mmask32 __U, __m512i __A, - __m512i __B) +_mm512_mask_subs_epu16 (__m512i __W, __mmask32 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) __W, - (__mmask32) __U); + return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U, + (__v32hi)_mm512_subs_epu16(__A, __B), + (__v32hi)__W); } static __inline__ __m512i __DEFAULT_FN_ATTRS _mm512_maskz_subs_epu16 (__mmask32 __U, __m512i __A, __m512i __B) { - return (__m512i) __builtin_ia32_psubusw512_mask ((__v32hi) __A, - (__v32hi) __B, - (__v32hi) _mm512_setzero_si512(), - (__mmask32) __U); + return (__m512i)__builtin_ia32_selectw_512((__mmask32)__U, + (__v32hi)_mm512_subs_epu16(__A, __B), + (__v32hi)_mm512_setzero_si512()); } static __inline__ __m512i __DEFAULT_FN_ATTRS |