Files
blis/ref_kernels/3/bli_trsm_ref.c
Field G. Van Zee bdd46f9ee8 Rewrote reference kernels to use #pragma omp simd.
Details:
- Rewrote level-1v, -1f, and -3 reference kernels in terms of simplified
  indexing annotated by the #pragma omp simd directive, which a compiler
  can use to vectorize certain constant-bounded loops. (The new kernels
  actually use _Pragma("omp simd") since the kernels are defined via
  templatizing macros.) Modest speedup was observed in most cases using
  gcc 5.4.0, which may improve with newer versions. Thanks to Devin
  Matthews for suggesting this via issue #286 and #259.
- Updated default blocksizes defined in ref_kernels/bli_cntx_ref.c to
  be 4x16, 4x8, 4x8, and 4x4 for single, double, scomplex and dcomplex,
  respectively, with a default row preference for the gemm ukernel. Also
  updated axpyf, dotxf, and dotxaxpyf fusing factors to 8, 6, and 4,
  respectively, for all datatypes.
- Modified configure to verify that -fopenmp-simd is a valid compiler
  option (via a new detect/omp_simd/omp_simd_detect.c file).
- Added a new header in which prefetch macros are defined according to
  which compiler is detected (via macros such as __GNUC__). These
  prefetch macros are not yet employed anywhere, though.
- Updated the year in copyrights of template license headers in
  build/templates and removed AMD as a default copyright holder.
2019-01-24 17:23:18 -06:00

332 lines
11 KiB
C

/*
BLIS
An object-based framework for developing high-performance BLAS-like
libraries.
Copyright (C) 2014, The University of Texas at Austin
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
- Neither the name(s) of the copyright holder(s) nor the names of its
contributors may be used to endorse or promote products derived
from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "blis.h"
#if 1
// An implementation that attempts to facilitate emission of vectorized
// instructions via constant loop bounds + #pragma omp simd directives.
#undef GENTFUNC
#define GENTFUNC( ctype, ch, opname, arch, suf, mr, nr ) \
\
void PASTEMAC3(ch,opname,arch,suf) \
( \
ctype* restrict a, \
ctype* restrict b, \
ctype* restrict c, inc_t rs_c, inc_t cs_c, \
auxinfo_t* restrict data, \
cntx_t* restrict cntx \
) \
{ \
const inc_t rs_a = 1; \
const inc_t cs_a = mr; \
\
const inc_t rs_b = nr; \
const inc_t cs_b = 1; \
\
_Pragma( "omp simd" ) \
for ( dim_t i = 0; i < mr; ++i ) \
{ \
/* b1 = b1 - a10t * B0; */ \
/* b1 = b1 / alpha11; */ \
for ( dim_t j = 0; j < nr; ++j ) \
{ \
ctype beta11c = b[i*rs_b + j*cs_b]; \
ctype rho11; \
\
/* beta11 = beta11 - a10t * b01; */ \
PASTEMAC(ch,set0s)( rho11 ); \
for ( dim_t l = 0; l < i; ++l ) \
{ \
PASTEMAC(ch,axpys)( a[i*rs_a + l*cs_a], \
b[l*rs_b + j*cs_b], rho11 ); \
} \
PASTEMAC(ch,subs)( rho11, beta11c ); \
\
/* beta11 = beta11 / alpha11; */ \
/* NOTE: The INVERSE of alpha11 (1.0/alpha11) is stored instead
of alpha11, so we can multiply rather than divide. We store
the inverse of alpha11 intentionally to avoid expensive
division instructions within the micro-kernel. */ \
PASTEMAC(ch,scals)( a[i*rs_a + i*cs_a], beta11c ); \
\
/* Output final result to matrix c. */ \
PASTEMAC(ch,copys)( beta11c, c[i*rs_c + j*cs_c] ); \
\
/* Store the local value back to b11. */ \
PASTEMAC(ch,copys)( beta11c, b[i*rs_b + j*cs_b] ); \
} \
} \
}
//INSERT_GENTFUNC_BASIC2( trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX )
GENTFUNC( float, s, trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 16 )
GENTFUNC( double, d, trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 8 )
GENTFUNC( scomplex, c, trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 8 )
GENTFUNC( dcomplex, z, trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 4 )
#undef GENTFUNC
#define GENTFUNC( ctype, ch, opname, arch, suf, mr, nr ) \
\
void PASTEMAC3(ch,opname,arch,suf) \
( \
ctype* restrict a, \
ctype* restrict b, \
ctype* restrict c, inc_t rs_c, inc_t cs_c, \
auxinfo_t* restrict data, \
cntx_t* restrict cntx \
) \
{ \
const inc_t rs_a = 1; \
const inc_t cs_a = mr; \
\
const inc_t rs_b = nr; \
const inc_t cs_b = 1; \
\
_Pragma( "omp simd" ) \
for ( dim_t iter = 0; iter < mr; ++iter ) \
{ \
dim_t i = mr - iter - 1; \
\
/* b1 = b1 - a12t * B2; */ \
/* b1 = b1 / alpha11; */ \
for ( dim_t j = 0; j < nr; ++j ) \
{ \
ctype beta11c = b[i*rs_b + j*cs_b]; \
ctype rho11; \
\
/* beta11 = beta11 - a12t * b21; */ \
PASTEMAC(ch,set0s)( rho11 ); \
for ( dim_t l = 0; l < iter; ++l ) \
{ \
PASTEMAC(ch,axpys)( a[i*rs_a + (i+1+l)*cs_a], \
b[(i+1+l)*rs_b + j*cs_b], rho11 ); \
} \
PASTEMAC(ch,subs)( rho11, beta11c ); \
\
/* beta11 = beta11 / alpha11; */ \
/* NOTE: The INVERSE of alpha11 (1.0/alpha11) is stored instead
of alpha11, so we can multiply rather than divide. We store
the inverse of alpha11 intentionally to avoid expensive
division instructions within the micro-kernel. */ \
PASTEMAC(ch,scals)( a[i*rs_a + i*cs_a], beta11c ); \
\
/* Output final result to matrix c. */ \
PASTEMAC(ch,copys)( beta11c, c[i*rs_c + j*cs_c] ); \
\
/* Store the local value back to b11. */ \
PASTEMAC(ch,copys)( beta11c, b[i*rs_b + j*cs_b] ); \
} \
} \
}
//INSERT_GENTFUNC_BASIC2( trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX )
GENTFUNC( float, s, trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 16 )
GENTFUNC( double, d, trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 8 )
GENTFUNC( scomplex, c, trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 8 )
GENTFUNC( dcomplex, z, trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX, 4, 4 )
#else
// An implementation that uses variable loop bounds (queried from the context)
// and makes no use of #pragma omp simd.
#undef GENTFUNC
#define GENTFUNC( ctype, ch, opname, arch, suf ) \
\
void PASTEMAC3(ch,opname,arch,suf) \
( \
ctype* restrict a, \
ctype* restrict b, \
ctype* restrict c, inc_t rs_c, inc_t cs_c, \
auxinfo_t* restrict data, \
cntx_t* restrict cntx \
) \
{ \
const num_t dt = PASTEMAC(ch,type); \
\
const dim_t mr = bli_cntx_get_blksz_def_dt( dt, BLIS_MR, cntx ); \
const dim_t nr = bli_cntx_get_blksz_def_dt( dt, BLIS_NR, cntx ); \
\
const inc_t packmr = bli_cntx_get_blksz_max_dt( dt, BLIS_MR, cntx ); \
const inc_t packnr = bli_cntx_get_blksz_max_dt( dt, BLIS_NR, cntx ); \
\
const dim_t m = mr; \
const dim_t n = nr; \
\
const inc_t rs_a = 1; \
const inc_t cs_a = packmr; \
\
const inc_t rs_b = packnr; \
const inc_t cs_b = 1; \
\
dim_t iter, i, j, l; \
dim_t n_behind; \
\
for ( iter = 0; iter < m; ++iter ) \
{ \
i = iter; \
n_behind = i; \
\
ctype* restrict alpha11 = a + (i )*rs_a + (i )*cs_a; \
ctype* restrict a10t = a + (i )*rs_a + (0 )*cs_a; \
ctype* restrict B0 = b + (0 )*rs_b + (0 )*cs_b; \
ctype* restrict b1 = b + (i )*rs_b + (0 )*cs_b; \
\
/* b1 = b1 - a10t * B0; */ \
/* b1 = b1 / alpha11; */ \
for ( j = 0; j < n; ++j ) \
{ \
ctype* restrict b01 = B0 + (0 )*rs_b + (j )*cs_b; \
ctype* restrict beta11 = b1 + (0 )*rs_b + (j )*cs_b; \
ctype* restrict gamma11 = c + (i )*rs_c + (j )*cs_c; \
ctype beta11c = *beta11; \
ctype rho11; \
\
/* beta11 = beta11 - a10t * b01; */ \
PASTEMAC(ch,set0s)( rho11 ); \
for ( l = 0; l < n_behind; ++l ) \
{ \
ctype* restrict alpha10 = a10t + (l )*cs_a; \
ctype* restrict beta01 = b01 + (l )*rs_b; \
\
PASTEMAC(ch,axpys)( *alpha10, *beta01, rho11 ); \
} \
PASTEMAC(ch,subs)( rho11, beta11c ); \
\
/* beta11 = beta11 / alpha11; */ \
/* NOTE: The INVERSE of alpha11 (1.0/alpha11) is stored instead
of alpha11, so we can multiply rather than divide. We store
the inverse of alpha11 intentionally to avoid expensive
division instructions within the micro-kernel. */ \
PASTEMAC(ch,scals)( *alpha11, beta11c ); \
\
/* Output final result to matrix c. */ \
PASTEMAC(ch,copys)( beta11c, *gamma11 ); \
\
/* Store the local value back to b11. */ \
PASTEMAC(ch,copys)( beta11c, *beta11 ); \
} \
} \
}
INSERT_GENTFUNC_BASIC2( trsm_l, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX )
#undef GENTFUNC
#define GENTFUNC( ctype, ch, opname, arch, suf ) \
\
void PASTEMAC3(ch,opname,arch,suf) \
( \
ctype* restrict a, \
ctype* restrict b, \
ctype* restrict c, inc_t rs_c, inc_t cs_c, \
auxinfo_t* restrict data, \
cntx_t* restrict cntx \
) \
{ \
const num_t dt = PASTEMAC(ch,type); \
\
const dim_t mr = bli_cntx_get_blksz_def_dt( dt, BLIS_MR, cntx ); \
const dim_t nr = bli_cntx_get_blksz_def_dt( dt, BLIS_NR, cntx ); \
\
const inc_t packmr = bli_cntx_get_blksz_max_dt( dt, BLIS_MR, cntx ); \
const inc_t packnr = bli_cntx_get_blksz_max_dt( dt, BLIS_NR, cntx ); \
\
const dim_t m = mr; \
const dim_t n = nr; \
\
const inc_t rs_a = 1; \
const inc_t cs_a = packmr; \
\
const inc_t rs_b = packnr; \
const inc_t cs_b = 1; \
\
dim_t iter, i, j, l; \
dim_t n_behind; \
\
for ( iter = 0; iter < m; ++iter ) \
{ \
i = m - iter - 1; \
n_behind = iter; \
\
ctype* restrict alpha11 = a + (i )*rs_a + (i )*cs_a; \
ctype* restrict a12t = a + (i )*rs_a + (i+1)*cs_a; \
ctype* restrict b1 = b + (i )*rs_b + (0 )*cs_b; \
ctype* restrict B2 = b + (i+1)*rs_b + (0 )*cs_b; \
\
/* b1 = b1 - a12t * B2; */ \
/* b1 = b1 / alpha11; */ \
for ( j = 0; j < n; ++j ) \
{ \
ctype* restrict beta11 = b1 + (0 )*rs_b + (j )*cs_b; \
ctype* restrict b21 = B2 + (0 )*rs_b + (j )*cs_b; \
ctype* restrict gamma11 = c + (i )*rs_c + (j )*cs_c; \
ctype beta11c = *beta11; \
ctype rho11; \
\
/* beta11 = beta11 - a12t * b21; */ \
PASTEMAC(ch,set0s)( rho11 ); \
for ( l = 0; l < n_behind; ++l ) \
{ \
ctype* restrict alpha12 = a12t + (l )*cs_a; \
ctype* restrict beta21 = b21 + (l )*rs_b; \
\
PASTEMAC(ch,axpys)( *alpha12, *beta21, rho11 ); \
} \
PASTEMAC(ch,subs)( rho11, beta11c ); \
\
/* beta11 = beta11 / alpha11; */ \
/* NOTE: The INVERSE of alpha11 (1.0/alpha11) is stored instead
of alpha11, so we can multiply rather than divide. We store
the inverse of alpha11 intentionally to avoid expensive
division instructions within the micro-kernel. */ \
PASTEMAC(ch,scals)( *alpha11, beta11c ); \
\
/* Output final result to matrix c. */ \
PASTEMAC(ch,copys)( beta11c, *gamma11 ); \
\
/* Store the local value back to b11. */ \
PASTEMAC(ch,copys)( beta11c, *beta11 ); \
} \
} \
}
INSERT_GENTFUNC_BASIC2( trsm_u, BLIS_CNAME_INFIX, BLIS_REF_SUFFIX )
#endif