maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[X86] Remove mask argument from some builtins that are handled completely in CGBuiltin.cpp. Just wrap a select builtin around them in the header file instead. 

llvm-svn: 333027
This commit is contained in:
Craig Topper 2018-05-22 20:48:24 +00:00
parent 99be40c363
commit 8e3689c066
6 changed files with 69 additions and 101 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
clang/include/clang/Basic/BuiltinsX86.def
View File

@ -1102,8 +1102,8 @@ TARGET_BUILTIN(__builtin_ia32_vpconflictsi_128_mask, "V4iV4iV4iUc", "nc", "avx51
TARGET_BUILTIN(__builtin_ia32_vpconflictsi_256_mask, "V8iV8iV8iUc", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vpconflictdi_512_mask, "V8LLiV8LLiV8LLiUc", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vpconflictsi_512_mask, "V16iV16iV16iUs", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_512_mask, "V16iV16iV16iUs", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_512_mask, "V8LLiV8LLiV8LLiUc", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_512, "V16iV16i", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_512, "V8LLiV8LLi", "nc", "avx512cd")
TARGET_BUILTIN(__builtin_ia32_vpopcntd_128, "V4iV4i", "nc", "avx512vpopcntdq,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vpopcntq_128, "V2LLiV2LLi", "nc", "avx512vpopcntdq,avx512vl")
@ -1550,10 +1550,10 @@ TARGET_BUILTIN(__builtin_ia32_rcp14pd128_mask, "V2dV2dV2dUc", "nc", "avx512vl")
TARGET_BUILTIN(__builtin_ia32_rcp14pd256_mask, "V4dV4dV4dUc", "nc", "avx512vl")
TARGET_BUILTIN(__builtin_ia32_rcp14ps128_mask, "V4fV4fV4fUc", "nc", "avx512vl")
TARGET_BUILTIN(__builtin_ia32_rcp14ps256_mask, "V8fV8fV8fUc", "nc", "avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_128_mask, "V4iV4iV4iUc", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_256_mask, "V8iV8iV8iUc", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_128_mask, "V2LLiV2LLiV2LLiUc", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_256_mask, "V4LLiV4LLiV4LLiUc", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_128, "V4iV4i", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntd_256, "V8iV8i", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_128, "V2LLiV2LLi", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vplzcntq_256, "V4LLiV4LLi", "nc", "avx512cd,avx512vl")
TARGET_BUILTIN(__builtin_ia32_vcvtsd2si32, "iV2dIi", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_vcvtsd2usi32, "UiV2dIi", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_vcvtss2si32, "iV4fIi", "nc", "avx512f")
@ -1778,7 +1778,7 @@ TARGET_BUILTIN(__builtin_ia32_kortestzhi, "iUsUs", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_kunpckhi, "UsUsUs", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_kxnorhi, "UsUsUs", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_kxorhi, "UsUsUs", "nc", "avx512f")
TARGET_BUILTIN(__builtin_ia32_palignr512_mask, "V64cV64cV64cIiV64cULLi", "nc", "avx512bw")
TARGET_BUILTIN(__builtin_ia32_palignr512, "V64cV64cV64cIi", "nc", "avx512bw")
TARGET_BUILTIN(__builtin_ia32_dbpsadbw128_mask, "V8sV16cV16cIiV8sUc", "nc", "avx512bw,avx512vl")
TARGET_BUILTIN(__builtin_ia32_dbpsadbw256_mask, "V16sV32cV32cIiV16sUs", "nc", "avx512bw,avx512vl")
TARGET_BUILTIN(__builtin_ia32_dbpsadbw512_mask, "V32sV64cV64cIiV32sUi", "nc", "avx512bw")

30
clang/lib/CodeGen/CGBuiltin.cpp
View File

@ -8890,7 +8890,7 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
}
case X86::BI__builtin_ia32_palignr128:
case X86::BI__builtin_ia32_palignr256:
case X86::BI__builtin_ia32_palignr512_mask: {
case X86::BI__builtin_ia32_palignr512: {
unsigned ShiftVal = cast<llvm::ConstantInt>(Ops[2])->getZExtValue();
unsigned NumElts = Ops[0]->getType()->getVectorNumElements();
@ -8920,15 +8920,9 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
}
}
Value *Align = Builder.CreateShuffleVector(Ops[1], Ops[0],
makeArrayRef(Indices, NumElts),
"palignr");
// If this isn't a masked builtin, just return the align operation.
if (Ops.size() == 3)
return Align;
return EmitX86Select(*this, Ops[4], Align, Ops[3]);
return Builder.CreateShuffleVector(Ops[1], Ops[0],
makeArrayRef(Indices, NumElts),
"palignr");
}
case X86::BI__builtin_ia32_vperm2f128_pd256:
@ -9097,16 +9091,14 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
return Builder.CreateBitCast(Res, Ops[0]->getType());
}
case X86::BI__builtin_ia32_vplzcntd_128_mask:
case X86::BI__builtin_ia32_vplzcntd_256_mask:
case X86::BI__builtin_ia32_vplzcntd_512_mask:
case X86::BI__builtin_ia32_vplzcntq_128_mask:
case X86::BI__builtin_ia32_vplzcntq_256_mask:
case X86::BI__builtin_ia32_vplzcntq_512_mask: {
case X86::BI__builtin_ia32_vplzcntd_128:
case X86::BI__builtin_ia32_vplzcntd_256:
case X86::BI__builtin_ia32_vplzcntd_512:
case X86::BI__builtin_ia32_vplzcntq_128:
case X86::BI__builtin_ia32_vplzcntq_256:
case X86::BI__builtin_ia32_vplzcntq_512: {
Function *F = CGM.getIntrinsic(Intrinsic::ctlz, Ops[0]->getType());
return EmitX86Select(*this, Ops[2],
Builder.CreateCall(F, {Ops[0],Builder.getInt1(false)}),
Ops[1]);
return Builder.CreateCall(F, {Ops[0],Builder.getInt1(false)});
}
case X86::BI__builtin_ia32_pabsb128:

20
clang/lib/Headers/avx512bwintrin.h
View File

@ -2056,22 +2056,18 @@ _mm512_mask_permutexvar_epi16 (__m512i __W, __mmask32 __M, __m512i __A,
}
#define _mm512_alignr_epi8(A, B, N) __extension__ ({\
(__m512i)__builtin_ia32_palignr512_mask((__v64qi)(__m512i)(A), \
(__v64qi)(__m512i)(B), (int)(N), \
(__v64qi)_mm512_undefined_pd(), \
(__mmask64)-1); })
(__m512i)__builtin_ia32_palignr512((__v64qi)(__m512i)(A), \
(__v64qi)(__m512i)(B), (int)(N)); })
#define _mm512_mask_alignr_epi8(W, U, A, B, N) __extension__({\
(__m512i)__builtin_ia32_palignr512_mask((__v64qi)(__m512i)(A), \
(__v64qi)(__m512i)(B), (int)(N), \
(__v64qi)(__m512i)(W), \
(__mmask64)(U)); })
(__m512i)__builtin_ia32_selectb_512((__mmask64)(U), \
(__v64qi)_mm512_alignr_epi8((A), (B), (int)(N)), \
(__v64qi)(__m512i)(W)); })
#define _mm512_maskz_alignr_epi8(U, A, B, N) __extension__({\
(__m512i)__builtin_ia32_palignr512_mask((__v64qi)(__m512i)(A), \
(__v64qi)(__m512i)(B), (int)(N), \
(__v64qi)_mm512_setzero_si512(), \
(__mmask64)(U)); })
(__m512i)__builtin_ia32_selectb_512((__mmask64)(U), \
(__v64qi)_mm512_alignr_epi8((A), (B), (int)(N)), \
(__v64qi)(__m512i)_mm512_setzero_si512()); })
#define _mm512_dbsad_epu8(A, B, imm) __extension__ ({\
(__m512i)__builtin_ia32_dbpsadbw512_mask((__v64qi)(__m512i)(A), \

32
clang/lib/Headers/avx512cdintrin.h
View File

@ -82,49 +82,45 @@ _mm512_maskz_conflict_epi32 (__mmask16 __U, __m512i __A)
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_lzcnt_epi32 (__m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntd_512_mask ((__v16si) __A,
(__v16si) _mm512_setzero_si512 (),
(__mmask16) -1);
return (__m512i) __builtin_ia32_vplzcntd_512 ((__v16si) __A);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_lzcnt_epi32 (__m512i __W, __mmask16 __U, __m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntd_512_mask ((__v16si) __A,
(__v16si) __W,
(__mmask16) __U);
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
(__v16si)_mm512_lzcnt_epi32(__A),
(__v16si)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_lzcnt_epi32 (__mmask16 __U, __m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntd_512_mask ((__v16si) __A,
(__v16si) _mm512_setzero_si512 (),
(__mmask16) __U);
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__U,
(__v16si)_mm512_lzcnt_epi32(__A),
(__v16si)_mm512_setzero_si512());
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_lzcnt_epi64 (__m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntq_512_mask ((__v8di) __A,
(__v8di) _mm512_setzero_si512 (),
(__mmask8) -1);
return (__m512i) __builtin_ia32_vplzcntq_512 ((__v8di) __A);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_lzcnt_epi64 (__m512i __W, __mmask8 __U, __m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntq_512_mask ((__v8di) __A,
(__v8di) __W,
(__mmask8) __U);
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
(__v8di)_mm512_lzcnt_epi64(__A),
(__v8di)__W);
}
static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_maskz_lzcnt_epi64 (__mmask8 __U, __m512i __A)
{
return (__m512i) __builtin_ia32_vplzcntq_512_mask ((__v8di) __A,
(__v8di) _mm512_setzero_si512 (),
(__mmask8) __U);
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__U,
(__v8di)_mm512_lzcnt_epi64(__A),
(__v8di)_mm512_setzero_si512());
}
static __inline__ __m512i __DEFAULT_FN_ATTRS

72
clang/lib/Headers/avx512vlcdintrin.h
View File

@ -157,105 +157,89 @@ _mm256_maskz_conflict_epi32 (__mmask8 __U, __m256i __A)
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_lzcnt_epi32 (__m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntd_128_mask ((__v4si) __A,
(__v4si)
_mm_setzero_si128 (),
(__mmask8) -1);
return (__m128i) __builtin_ia32_vplzcntd_128 ((__v4si) __A);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_lzcnt_epi32 (__m128i __W, __mmask8 __U, __m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntd_128_mask ((__v4si) __A,
(__v4si) __W,
(__mmask8) __U);
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_lzcnt_epi32(__A),
(__v4si)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_lzcnt_epi32 (__mmask8 __U, __m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntd_128_mask ((__v4si) __A,
(__v4si)
_mm_setzero_si128 (),
(__mmask8) __U);
return (__m128i)__builtin_ia32_selectd_128((__mmask8)__U,
(__v4si)_mm_lzcnt_epi32(__A),
(__v4si)_mm_setzero_si128());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_lzcnt_epi32 (__m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntd_256_mask ((__v8si) __A,
(__v8si)
_mm256_setzero_si256 (),
(__mmask8) -1);
return (__m256i) __builtin_ia32_vplzcntd_256 ((__v8si) __A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_mask_lzcnt_epi32 (__m256i __W, __mmask8 __U, __m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntd_256_mask ((__v8si) __A,
(__v8si) __W,
(__mmask8) __U);
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_lzcnt_epi32(__A),
(__v8si)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_lzcnt_epi32 (__mmask8 __U, __m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntd_256_mask ((__v8si) __A,
(__v8si)
_mm256_setzero_si256 (),
(__mmask8) __U);
return (__m256i)__builtin_ia32_selectd_256((__mmask8)__U,
(__v8si)_mm256_lzcnt_epi32(__A),
(__v8si)_mm256_setzero_si256());
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_lzcnt_epi64 (__m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntq_128_mask ((__v2di) __A,
(__v2di)
_mm_setzero_di (),
(__mmask8) -1);
return (__m128i) __builtin_ia32_vplzcntq_128 ((__v2di) __A);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_mask_lzcnt_epi64 (__m128i __W, __mmask8 __U, __m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntq_128_mask ((__v2di) __A,
(__v2di) __W,
(__mmask8) __U);
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_lzcnt_epi64(__A),
(__v2di)__W);
}
static __inline__ __m128i __DEFAULT_FN_ATTRS
_mm_maskz_lzcnt_epi64 (__mmask8 __U, __m128i __A)
{
return (__m128i) __builtin_ia32_vplzcntq_128_mask ((__v2di) __A,
(__v2di)
_mm_setzero_di (),
(__mmask8) __U);
return (__m128i)__builtin_ia32_selectq_128((__mmask8)__U,
(__v2di)_mm_lzcnt_epi64(__A),
(__v2di)_mm_setzero_di());
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_lzcnt_epi64 (__m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntq_256_mask ((__v4di) __A,
(__v4di)
_mm256_setzero_si256 (),
(__mmask8) -1);
return (__m256i) __builtin_ia32_vplzcntq_256 ((__v4di) __A);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_mask_lzcnt_epi64 (__m256i __W, __mmask8 __U, __m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntq_256_mask ((__v4di) __A,
(__v4di) __W,
(__mmask8) __U);
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_lzcnt_epi64(__A),
(__v4di)__W);
}
static __inline__ __m256i __DEFAULT_FN_ATTRS
_mm256_maskz_lzcnt_epi64 (__mmask8 __U, __m256i __A)
{
return (__m256i) __builtin_ia32_vplzcntq_256_mask ((__v4di) __A,
(__v4di)
_mm256_setzero_si256 (),
(__mmask8) __U);
return (__m256i)__builtin_ia32_selectq_256((__mmask8)__U,
(__v4di)_mm256_lzcnt_epi64(__A),
(__v4di)_mm256_setzero_si256());
}
#undef __DEFAULT_FN_ATTRS

2
clang/lib/Sema/SemaChecking.cpp
View File

@ -2689,7 +2689,7 @@ bool Sema::CheckX86BuiltinFunctionCall(unsigned BuiltinID, CallExpr *TheCall) {
break;
case X86::BI__builtin_ia32_palignr128:
case X86::BI__builtin_ia32_palignr256:
case X86::BI__builtin_ia32_palignr512_mask:
case X86::BI__builtin_ia32_palignr512:
case X86::BI__builtin_ia32_vcomisd:
case X86::BI__builtin_ia32_vcomiss:
case X86::BI__builtin_ia32_dbpsadbw128_mask: