This is an archive of the discontinued LLVM Phabricator instance.

Differential D20160

[Clang][AVX512][intrinsics] continue completing missing set intrinsics
ClosedPublic

Authored by m_zuckerman on May 11 2016, 5:34 AM.

Details

Reviewers
AsafBadouh
delena
igorb
Commits
rG178113e8cc70: [Clang][AVX512][intrinsics] continue completing missing set intrinsics
rC270047: [Clang][AVX512][intrinsics] continue completing missing set intrinsics
rL270047: [Clang][AVX512][intrinsics] continue completing missing set intrinsics

Diff Detail

Repository
rL LLVM

Event Timeline

m_zuckerman updated this revision to Diff 56885.May 11 2016, 5:34 AM
m_zuckerman retitled this revision from to [Clang][AVX512][intrinsics] continue completing missing set intrinsics .
m_zuckerman updated this object.
m_zuckerman added reviewers: AsafBadouh, igorb, delena.
m_zuckerman added a subscriber: llvm-commits.
AsafBadouh accepted this revision.May 19 2016, 12:45 AM
AsafBadouh edited edge metadata.

LGTM

This revision is now accepted and ready to land.May 19 2016, 12:45 AM
Closed by commit rL270047: [Clang][AVX512][intrinsics] continue completing missing set intrinsics (authored by mzuckerm). · Explain WhyMay 19 2016, 5:14 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
cfe/
trunk/
lib/
Headers/
24 lines
test/
CodeGen/
157 lines

Diff 57765

cfe/trunk/lib/Headers/avx512fintrin.h

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 8,977 Lines • ▼ Show 20 Lines
static __inline__ __m512i __DEFAULT_FN_ATTRS static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_set1_epi64 (__m512i __O, __mmask8 __M, long long __A) _mm512_mask_set1_epi64 (__m512i __O, __mmask8 __M, long long __A)
{ {
return (__m512i) __builtin_ia32_pbroadcastq512_gpr_mask (__A, (__v8di) __O, return (__m512i) __builtin_ia32_pbroadcastq512_gpr_mask (__A, (__v8di) __O,
__M); __M);
} }
static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_set_epi32 (int __A, int __B, int __C, int __D,
int __E, int __F, int __G, int __H,
int __I, int __J, int __K, int __L,
int __M, int __N, int __O, int __P)
{
return __extension__ (__m512i)(__v16si)
{ __P, __O, __N, __M, __L, __K, __J, __I,
__H, __G, __F, __E, __D, __C, __B, __A };
}
#define _mm512_setr_epi32(e0,e1,e2,e3,e4,e5,e6,e7, \
e8,e9,e10,e11,e12,e13,e14,e15) \
_mm512_set_epi32(e15,e14,e13,e12,e11,e10,e9,e8,e7,e6,e5,e4,e3,e2,e1,e0)
static __inline__ __m512i __DEFAULT_FN_ATTRS static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_set_epi64 (long long __A, long long __B, long long __C, _mm512_set_epi64 (long long __A, long long __B, long long __C,
long long __D, long long __E, long long __F, long long __D, long long __E, long long __F,
long long __G, long long __H) long long __G, long long __H)
{ {
return __extension__ (__m512i) (__v8di) return __extension__ (__m512i) (__v8di)
{ __H, __G, __F, __E, __D, __C, __B, __A }; { __H, __G, __F, __E, __D, __C, __B, __A };
} }
#define _mm512_setr_epi64(e0,e1,e2,e3,e4,e5,e6,e7) \
_mm512_set_epi64(e7,e6,e5,e4,e3,e2,e1,e0)
static __inline__ __m512d __DEFAULT_FN_ATTRS static __inline__ __m512d __DEFAULT_FN_ATTRS
_mm512_set_pd (double __A, double __B, double __C, double __D, _mm512_set_pd (double __A, double __B, double __C, double __D,
double __E, double __F, double __G, double __H) double __E, double __F, double __G, double __H)
{ {
return __extension__ (__m512d) return __extension__ (__m512d)
{ __H, __G, __F, __E, __D, __C, __B, __A }; { __H, __G, __F, __E, __D, __C, __B, __A };
} }
#define _mm512_setr_pd(e0,e1,e2,e3,e4,e5,e6,e7) \
_mm512_set_pd(e7,e6,e5,e4,e3,e2,e1,e0)
static __inline__ __m512 __DEFAULT_FN_ATTRS static __inline__ __m512 __DEFAULT_FN_ATTRS
_mm512_set_ps (float __A, float __B, float __C, float __D, _mm512_set_ps (float __A, float __B, float __C, float __D,
float __E, float __F, float __G, float __H, float __E, float __F, float __G, float __H,
float __I, float __J, float __K, float __L, float __I, float __J, float __K, float __L,
float __M, float __N, float __O, float __P) float __M, float __N, float __O, float __P)
{ {
return __extension__ (__m512) return __extension__ (__m512)
{ __P, __O, __N, __M, __L, __K, __J, __I, { __P, __O, __N, __M, __L, __K, __J, __I,
__H, __G, __F, __E, __D, __C, __B, __A }; __H, __G, __F, __E, __D, __C, __B, __A };
} }
#define _mm512_setr_ps(e0,e1,e2,e3,e4,e5,e6,e7,e8,e9,e10,e11,e12,e13,e14,e15) \
_mm512_set_ps(e15,e14,e13,e12,e11,e10,e9,e8,e7,e6,e5,e4,e3,e2,e1,e0)
#undef __DEFAULT_FN_ATTRS #undef __DEFAULT_FN_ATTRS
#endif // __AVX512FINTRIN_H #endif // __AVX512FINTRIN_H

cfe/trunk/test/CodeGen/avx512f-builtins.c

Show First 20 Lines • Show All 6,515 Lines • ▼ Show 20 Lines
__m512i test_mm512_mask_set1_epi32 (__m512i __O, __mmask16 __M, int __A) __m512i test_mm512_mask_set1_epi32 (__m512i __O, __mmask16 __M, int __A)
{ {
//CHECK-LABLE: @test_mm512_mask_set1_epi32 //CHECK-LABLE: @test_mm512_mask_set1_epi32
//CHECK: @llvm.x86.avx512.mask.pbroadcast.d.gpr.512 //CHECK: @llvm.x86.avx512.mask.pbroadcast.d.gpr.512
return _mm512_mask_set1_epi32 ( __O, __M, __A); return _mm512_mask_set1_epi32 ( __O, __M, __A);
} }
__m512i test_mm512_set_epi32 (int __A, int __B, int __C, int __D,
int __E, int __F, int __G, int __H,
int __I, int __J, int __K, int __L,
int __M, int __N, int __O, int __P)
{
//CHECK-LABLE: @test_mm512_set_epi32
//CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7
//CHECK: insertelement{{.*}}i32 8
//CHECK: insertelement{{.*}}i32 9
//CHECK: insertelement{{.*}}i32 10
//CHECK: insertelement{{.*}}i32 11
//CHECK: insertelement{{.*}}i32 12
//CHECK: insertelement{{.*}}i32 13
//CHECK: insertelement{{.*}}i32 14
//CHECK: insertelement{{.*}}i32 15
return _mm512_set_epi32( __A, __B, __C, __D,__E, __F, __G, __H,
__I, __J, __K, __L,__M, __N, __O, __P);
}
__m512i test_mm512_setr_epi32 (int __A, int __B, int __C, int __D,
int __E, int __F, int __G, int __H,
int __I, int __J, int __K, int __L,
int __M, int __N, int __O, int __P)
{
//CHECK-LABLE: @test_mm512_setr_epi32
//CHECK: %0 = load{{.*}}%__P.addr, align 4
//CHECK: %1 = load{{.*}}%__O.addr, align 4
//CHECK: %2 = load{{.*}}%__N.addr, align 4
//CHECK: %3 = load{{.*}}%__M.addr, align 4
//CHECK: %4 = load{{.*}}%__L.addr, align 4
//CHECK: %5 = load{{.*}}%__K.addr, align 4
//CHECK: %6 = load{{.*}}%__J.addr, align 4
//CHECK: %7 = load{{.*}}%__I.addr, align 4
//CHECK: %8 = load{{.*}}%__H.addr, align 4
//CHECK: %9 = load{{.*}}%__G.addr, align 4
//CHECK: %10 = load{{.*}}%__F.addr, align 4
//CHECK: %11 = load{{.*}}%__E.addr, align 4
//CHECK: %12 = load{{.*}}%__D.addr, align 4
//CHECK: %13 = load{{.*}}%__C.addr, align 4
//CHECK: %14 = load{{.*}}%__B.addr, align 4
//CHECK: %15 = load{{.*}}%__A.addr, align 4
//CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7
//CHECK: insertelement{{.*}}i32 8
//CHECK: insertelement{{.*}}i32 9
//CHECK: insertelement{{.*}}i32 10
//CHECK: insertelement{{.*}}i32 11
//CHECK: insertelement{{.*}}i32 12
//CHECK: insertelement{{.*}}i32 13
//CHECK: insertelement{{.*}}i32 14
//CHECK: insertelement{{.*}}i32 15
return _mm512_setr_epi32( __A, __B, __C, __D,__E, __F, __G, __H,
__I, __J, __K, __L,__M, __N, __O, __P);
}
__m512i test_mm512_mask_set1_epi64 (__m512i __O, __mmask8 __M, long long __A) __m512i test_mm512_mask_set1_epi64 (__m512i __O, __mmask8 __M, long long __A)
{ {
//CHECK-LABLE: @test_mm512_mask_set1_epi64 //CHECK-LABLE: @test_mm512_mask_set1_epi64
//CHECK: @llvm.x86.avx512.mask.pbroadcast.q.gpr.512 //CHECK: @llvm.x86.avx512.mask.pbroadcast.q.gpr.512
return _mm512_mask_set1_epi64 (__O, __M, __A); return _mm512_mask_set1_epi64 (__O, __M, __A);
} }
__m512i test_mm512_set_epi64 (long long __A, long long __B, long long __C, __m512i test_mm512_set_epi64 (long long __A, long long __B, long long __C,
long long __D, long long __E, long long __F, long long __D, long long __E, long long __F,
long long __G, long long __H) long long __G, long long __H)
{ {
//CHECK-LABLE: @test_mm512_set_epi64 //CHECK-LABLE: @test_mm512_set_epi64
//CHECK: insertelement{{.*}}i32 0 //CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1 //CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2 //CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3 //CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4 //CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5 //CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6 //CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7 //CHECK: insertelement{{.*}}i32 7
return _mm512_set_epi64(__A, __B, __C, __D, __E, __F, __G, __H ); return _mm512_set_epi64(__A, __B, __C, __D, __E, __F, __G, __H );
} }
__m512i test_mm512_setr_epi64 (long long __A, long long __B, long long __C,
long long __D, long long __E, long long __F,
long long __G, long long __H)
{
//CHECK-LABLE: @test_mm512_setr_epi64
//CHECK: %0 = load{{.*}}%__H.addr, align 8
//CHECK: %1 = load{{.*}}%__G.addr, align 8
//CHECK: %2 = load{{.*}}%__F.addr, align 8
//CHECK: %3 = load{{.*}}%__E.addr, align 8
//CHECK: %4 = load{{.*}}%__D.addr, align 8
//CHECK: %5 = load{{.*}}%__C.addr, align 8
//CHECK: %6 = load{{.*}}%__B.addr, align 8
//CHECK: %7 = load{{.*}}%__A.addr, align 8
//CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7
return _mm512_setr_epi64(__A, __B, __C, __D, __E, __F, __G, __H );
}
__m512d test_mm512_set_pd (double __A, double __B, double __C, double __D, __m512d test_mm512_set_pd (double __A, double __B, double __C, double __D,
double __E, double __F, double __G, double __H) double __E, double __F, double __G, double __H)
{ {
//CHECK-LABLE: @test_mm512_set_pd //CHECK-LABLE: @test_mm512_set_pd
//CHECK: insertelement{{.*}}i32 0 //CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1 //CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2 //CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3 //CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4 //CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5 //CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6 //CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7 //CHECK: insertelement{{.*}}i32 7
return _mm512_set_pd( __A, __B, __C, __D, __E, __F, __G, __H); return _mm512_set_pd( __A, __B, __C, __D, __E, __F, __G, __H);
} }
__m512d test_mm512_setr_pd (double __A, double __B, double __C, double __D,
double __E, double __F, double __G, double __H)
{
//CHECK-LABLE: @test_mm512_setr_pd
//CHECK: %0 = load{{.*}}%__H.addr, align 8
//CHECK: %1 = load{{.*}}%__G.addr, align 8
//CHECK: %2 = load{{.*}}%__F.addr, align 8
//CHECK: %3 = load{{.*}}%__E.addr, align 8
//CHECK: %4 = load{{.*}}%__D.addr, align 8
//CHECK: %5 = load{{.*}}%__C.addr, align 8
//CHECK: %6 = load{{.*}}%__B.addr, align 8
//CHECK: %7 = load{{.*}}%__A.addr, align 8
//CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7
return _mm512_setr_pd( __A, __B, __C, __D, __E, __F, __G, __H);
}
__m512 test_mm512_set_ps (float __A, float __B, float __C, float __D, __m512 test_mm512_set_ps (float __A, float __B, float __C, float __D,
float __E, float __F, float __G, float __H, float __E, float __F, float __G, float __H,
float __I, float __J, float __K, float __L, float __I, float __J, float __K, float __L,
float __M, float __N, float __O, float __P) float __M, float __N, float __O, float __P)
{ {
//CHECK-LABLE: @test_mm512_set_ps //CHECK-LABLE: @test_mm512_set_ps
//CHECK: insertelement{{.*}}i32 0 //CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1 //CHECK: insertelement{{.*}}i32 1
Show All 37 Lines
} }
__m512i test_mm512_maskz_abs_epi32 (__mmask16 __U, __m512i __A) __m512i test_mm512_maskz_abs_epi32 (__mmask16 __U, __m512i __A)
{ {
// CHECK-LABEL: @test_mm512_maskz_abs_epi32 // CHECK-LABEL: @test_mm512_maskz_abs_epi32
// CHECK: @llvm.x86.avx512.mask.pabs.d.512 // CHECK: @llvm.x86.avx512.mask.pabs.d.512
return _mm512_maskz_abs_epi32 (__U,__A); return _mm512_maskz_abs_epi32 (__U,__A);
} }
__m512 test_mm512_setr_ps (float __A, float __B, float __C, float __D,
float __E, float __F, float __G, float __H,
float __I, float __J, float __K, float __L,
float __M, float __N, float __O, float __P)
{
//CHECK-LABLE: @test_mm512_setr_ps
//CHECK: %0 = load{{.*}}%__P.addr, align 4
//CHECK: %1 = load{{.*}}%__O.addr, align 4
//CHECK: %2 = load{{.*}}%__N.addr, align 4
//CHECK: %3 = load{{.*}}%__M.addr, align 4
//CHECK: %4 = load{{.*}}%__L.addr, align 4
//CHECK: %5 = load{{.*}}%__K.addr, align 4
//CHECK: %6 = load{{.*}}%__J.addr, align 4
//CHECK: %7 = load{{.*}}%__I.addr, align 4
//CHECK: %8 = load{{.*}}%__H.addr, align 4
//CHECK: %9 = load{{.*}}%__G.addr, align 4
//CHECK: %10 = load{{.*}}%__F.addr, align 4
//CHECK: %11 = load{{.*}}%__E.addr, align 4
//CHECK: %12 = load{{.*}}%__D.addr, align 4
//CHECK: %13 = load{{.*}}%__C.addr, align 4
//CHECK: %14 = load{{.*}}%__B.addr, align 4
//CHECK: %15 = load{{.*}}%__A.addr, align 4
//CHECK: insertelement{{.*}}i32 0
//CHECK: insertelement{{.*}}i32 1
//CHECK: insertelement{{.*}}i32 2
//CHECK: insertelement{{.*}}i32 3
//CHECK: insertelement{{.*}}i32 4
//CHECK: insertelement{{.*}}i32 5
//CHECK: insertelement{{.*}}i32 6
//CHECK: insertelement{{.*}}i32 7
//CHECK: insertelement{{.*}}i32 8
//CHECK: insertelement{{.*}}i32 9
//CHECK: insertelement{{.*}}i32 10
//CHECK: insertelement{{.*}}i32 11
//CHECK: insertelement{{.*}}i32 12
//CHECK: insertelement{{.*}}i32 13
//CHECK: insertelement{{.*}}i32 14
//CHECK: insertelement{{.*}}i32 15
return _mm512_setr_ps( __A, __B, __C, __D, __E, __F, __G, __H,
__I, __J, __K, __L, __M, __N, __O, __P);
}