mirror of
https://github.com/amd/blis.git
synced 2026-04-30 20:41:13 +00:00
5fec95b99f
Details: - Implemented support for gemm where A, B, and C may have different storage datatypes, as well as a computational precision (and implied computation domain) that may be different from the storage precision of either A or B. This results in 128 different combinations, all which are implemented within this commit. (For now, the mixed-datatype functionality is only supported via the object API.) If desired, the mixed-datatype support may be disabled at configure-time. - Added a memory-intensive optimization to certain mixed-datatype cases that requires a single m-by-n matrix be allocated (temporarily) per call to gemm. This optimization aims to avoid the overhead involved in repeatedly updating C with general stride, or updating C after a typecast from the computation precision. This memory optimization may be disabled at configure-time (provided that the mixed-datatype support is enabled in the first place). - Added support for testing mixed-datatype combinations to testsuite. The user may test gemm with mixed domains, precisions, both, or neither. - Added a standalone test driver directory for building and running mixed-datatype performance experiments. - Defined a new variation of castm, castnzm, which operates like castm except that imaginary values are not touched when casting a real operand to a complex operand. (By contrast, in these situations castm sets the imaginary components of the destination matrix to zero.) - Defined bli_obj_imag_is_zero() and substituted calls in lieu of all usages of bli_obj_imag_equals() that tested against BLIS_ZERO, and also simplified the implementation of bli_obj_imag_equals(). - Fixed bad behavior from bli_obj_is_real() and bli_obj_is_complex() when given BLIS_CONSTANT objects. - Disabled dt_on_output field in auxinfo_t structure as well as all accessor functions. Also commented out all usage of accessor functions within macrokernels. (Typecasting in the microkernel is still feasible, though probably unrealistic for now given the additional complexity required.) - Use void function pointer type (instead of void*) for storing function pointers in bli_l0_fpa.c. - Added documentation for using gemm with mixed datatypes in docs/MixedDatatypes.md and example code in examples/oapi/11gemm_md.c. - Defined level-1d operation xpbyd and level-1m operation xpbym. - Added xpbym test module to testsuite. - Updated frame/include/bli_x86_asm_macros.h with additional macros (courtsey of Devin Matthews).
328 lines
10 KiB
C
328 lines
10 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 of The University of Texas at Austin 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 "bli_gemm_md_c2r_ref.h"
|
|
|
|
// Define a local struct type that makes returning two values easier.
|
|
typedef struct mddm_s
|
|
{
|
|
dom_t comp;
|
|
dom_t exec;
|
|
} mddm_t;
|
|
|
|
void bli_gemm_md
|
|
(
|
|
obj_t* a,
|
|
obj_t* b,
|
|
obj_t* beta,
|
|
obj_t* c,
|
|
cntx_t* cntx_local,
|
|
cntx_t** cntx
|
|
);
|
|
mddm_t bli_gemm_md_ccc( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_ccr( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_crc( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_rcc( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_rrc( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_rcr( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_crr( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
mddm_t bli_gemm_md_rrr( obj_t* a, obj_t* b, obj_t* beta, obj_t* c, cntx_t* cntx_l, cntx_t** cntx );
|
|
|
|
// -----------------------------------------------------------------------------
|
|
|
|
void bli_gemm_md_front
|
|
(
|
|
obj_t* alpha,
|
|
obj_t* a,
|
|
obj_t* b,
|
|
obj_t* beta,
|
|
obj_t* c,
|
|
cntx_t* cntx,
|
|
rntm_t* rntm,
|
|
cntl_t* cntl
|
|
);
|
|
|
|
void bli_gemm_md_zgemm
|
|
(
|
|
obj_t* alpha,
|
|
obj_t* a,
|
|
obj_t* b,
|
|
obj_t* beta,
|
|
obj_t* c,
|
|
cntx_t* cntx,
|
|
rntm_t* rntm,
|
|
cntl_t* cntl
|
|
);
|
|
|
|
// -----------------------------------------------------------------------------
|
|
|
|
static bool_t bli_gemm_md_is_crr( obj_t* a, obj_t* b, obj_t* c )
|
|
{
|
|
bool_t r_val = FALSE;
|
|
|
|
// NOTE: The last conditional subexpression is necessary if/when we
|
|
// allow the user to specify the computation domain. (The computation
|
|
// domain is currently ignored, but once it is honored as a user-
|
|
// settable value, it will affect the execution domain, which is what
|
|
// is checked below. Until then, the last expression is not actually
|
|
// necessary since crr is already unconditionally associated with an
|
|
// execution domain of BLIS_REAL.)
|
|
if ( bli_obj_is_complex( c ) &&
|
|
bli_obj_is_real( a ) &&
|
|
bli_obj_is_real( b ) &&
|
|
bli_obj_exec_domain( c ) == BLIS_REAL )
|
|
r_val = TRUE;
|
|
|
|
return r_val;
|
|
}
|
|
|
|
static bool_t bli_gemm_md_is_ccr( obj_t* a, obj_t* b, obj_t* c )
|
|
{
|
|
bool_t r_val = FALSE;
|
|
|
|
// NOTE: The last conditional subexpression is necessary if/when we
|
|
// allow the user to specify the computation domain. (The computation
|
|
// domain is currently ignored, but once it is honored as a user-
|
|
// settable value, it will affect the execution domain, which is what
|
|
// is checked below. Until then, the last expression is not actually
|
|
// necessary since ccr is already unconditionally associated with an
|
|
// execution domain of BLIS_COMPLEX.)
|
|
if ( bli_obj_is_complex( c ) &&
|
|
bli_obj_is_complex( a ) &&
|
|
bli_obj_is_real( b ) &&
|
|
bli_obj_exec_domain( c ) == BLIS_COMPLEX )
|
|
r_val = TRUE;
|
|
|
|
return r_val;
|
|
}
|
|
|
|
static bool_t bli_gemm_md_is_crc( obj_t* a, obj_t* b, obj_t* c )
|
|
{
|
|
bool_t r_val = FALSE;
|
|
|
|
// NOTE: The last conditional subexpression is necessary if/when we
|
|
// allow the user to specify the computation domain. (The computation
|
|
// domain is currently ignored, but once it is honored as a user-
|
|
// settable value, it will affect the execution domain, which is what
|
|
// is checked below. Until then, the last expression is not actually
|
|
// necessary since crc is already unconditionally associated with an
|
|
// execution domain of BLIS_COMPLEX.)
|
|
if ( bli_obj_is_complex( c ) &&
|
|
bli_obj_is_real( a ) &&
|
|
bli_obj_is_complex( b ) &&
|
|
bli_obj_exec_domain( c ) == BLIS_COMPLEX )
|
|
r_val = TRUE;
|
|
|
|
return r_val;
|
|
}
|
|
|
|
// -----------------------------------------------------------------------------
|
|
|
|
static void bli_gemm_md_ker_var2_recast
|
|
(
|
|
num_t* dt_comp,
|
|
num_t dt_a,
|
|
num_t dt_b,
|
|
num_t dt_c,
|
|
dim_t* m,
|
|
dim_t* n,
|
|
dim_t* k,
|
|
inc_t* pd_a, inc_t* ps_a,
|
|
inc_t* pd_b, inc_t* ps_b,
|
|
obj_t* c,
|
|
inc_t* rs_c, inc_t* cs_c
|
|
)
|
|
{
|
|
if ( bli_is_real( dt_c ) &&
|
|
bli_is_complex( dt_a ) &&
|
|
bli_is_complex( dt_b ) )
|
|
{
|
|
// The rcc case is executed with a real macrokernel, so we need to
|
|
// double the k dimension (because both A and B are packed to the 1r
|
|
// schema), and also the panel strides of A and B since they were
|
|
// packed as complex matrices and we now need to convert them to
|
|
// units of real elements.
|
|
*k *= 2;
|
|
*ps_a *= 2;
|
|
*ps_b *= 2;
|
|
}
|
|
else if ( bli_is_complex( dt_c ) &&
|
|
bli_is_real( dt_a ) &&
|
|
bli_is_complex( dt_b ) )
|
|
{
|
|
#if 1
|
|
obj_t beta;
|
|
|
|
bli_obj_scalar_detach( c, &beta );
|
|
|
|
if ( //bli_obj_imag_equals( &beta, &BLIS_ZERO ) &&
|
|
bli_obj_imag_is_zero( &beta ) &&
|
|
bli_is_row_stored( *rs_c, *cs_c ) &&
|
|
bli_obj_prec( c ) == bli_obj_comp_prec( c ) )
|
|
{
|
|
// If beta is real, and C is not general-stored, and the computation
|
|
// precision is equal to the storage precision of C, we can use the
|
|
// real macrokernel (and real microkernel, which is already stored
|
|
// to the real virtual microkernel slots of the context) instead of
|
|
// the complex macrokernel and c2r virtual microkernel.
|
|
*dt_comp = bli_dt_proj_to_real( *dt_comp );
|
|
*n *= 2;
|
|
*pd_b *= 2; *ps_b *= 2;
|
|
*rs_c *= 2;
|
|
}
|
|
else
|
|
#endif
|
|
{
|
|
// Generally speaking, the crc case is executed with a complex
|
|
// macrokernel, so we need to halve the panel stride of A (which
|
|
// is real) since the macrokernel will perform the pointer
|
|
// arithmetic in units of complex elements.
|
|
*ps_a /= 2;
|
|
}
|
|
}
|
|
else if ( bli_is_complex( dt_c ) &&
|
|
bli_is_complex( dt_a ) &&
|
|
bli_is_real( dt_b ) )
|
|
{
|
|
#if 1
|
|
obj_t beta;
|
|
|
|
bli_obj_scalar_detach( c, &beta );
|
|
|
|
if ( //bli_obj_imag_equals( &beta, &BLIS_ZERO ) &&
|
|
bli_obj_imag_is_zero( &beta ) &&
|
|
bli_is_col_stored( *rs_c, *cs_c ) &&
|
|
bli_obj_prec( c ) == bli_obj_comp_prec( c ) )
|
|
{
|
|
// If beta is real, and C is not general-stored, and the computation
|
|
// precision is equal to the storage precision of C, we can use the
|
|
// real macrokernel (and real microkernel, which is already stored
|
|
// to the real virtual microkernel slots of the context) instead of
|
|
// the complex macrokernel and c2r virtual microkernel.
|
|
*dt_comp = bli_dt_proj_to_real( *dt_comp );
|
|
*m *= 2;
|
|
*pd_a *= 2; *ps_a *= 2;
|
|
*cs_c *= 2;
|
|
}
|
|
else
|
|
#endif
|
|
{
|
|
// Generally speaking, the ccr case is executed with a complex
|
|
// macrokernel, so we need to halve the panel stride of B (which
|
|
// is real) since the macrokernel will perform the pointer
|
|
// arithmetic in units of complex elements.
|
|
*ps_b /= 2;
|
|
}
|
|
}
|
|
#if 0
|
|
else if ( bli_is_real( dt_c ) &&
|
|
bli_is_real( dt_a ) &&
|
|
bli_is_real( dt_b ) )
|
|
{
|
|
// No action needed.
|
|
//printf( "gemm_md.h: rrr: m n k are now %d %d %d\n", (int)*m, (int)*n, (int)*k );
|
|
}
|
|
else if ( bli_is_complex( dt_c ) &&
|
|
bli_is_real( dt_a ) &&
|
|
bli_is_real( dt_b ) )
|
|
{
|
|
// No action needed.
|
|
}
|
|
else if ( bli_is_real( dt_c ) &&
|
|
bli_is_complex( dt_a ) &&
|
|
bli_is_real( dt_b ) )
|
|
{
|
|
// No action needed.
|
|
}
|
|
else if ( bli_is_real( dt_c ) &&
|
|
bli_is_real( dt_a ) &&
|
|
bli_is_complex( dt_b ) )
|
|
{
|
|
// No action needed.
|
|
}
|
|
#endif
|
|
}
|
|
|
|
// -----------------------------------------------------------------------------
|
|
|
|
//
|
|
// Prototype object-based interfaces.
|
|
//
|
|
|
|
#undef GENPROT
|
|
#define GENPROT( opname ) \
|
|
\
|
|
void PASTEMAC0(opname) \
|
|
( \
|
|
obj_t* a, \
|
|
obj_t* b, \
|
|
obj_t* c, \
|
|
cntx_t* cntx, \
|
|
rntm_t* rntm, \
|
|
cntl_t* cntl, \
|
|
thrinfo_t* thread \
|
|
);
|
|
|
|
GENPROT( gemm_ker_var2_md )
|
|
|
|
//
|
|
// Prototype BLAS-like interfaces with void pointer operands.
|
|
//
|
|
|
|
#undef GENTPROT2
|
|
#define GENTPROT2( ctype_c, ctype_e, chc, che, varname ) \
|
|
\
|
|
void PASTEMAC2(chc,che,varname) \
|
|
( \
|
|
pack_t schema_a, \
|
|
pack_t schema_b, \
|
|
dim_t m, \
|
|
dim_t n, \
|
|
dim_t k, \
|
|
void* alpha, \
|
|
void* a, inc_t cs_a, inc_t is_a, \
|
|
dim_t pd_a, inc_t ps_a, \
|
|
void* b, inc_t rs_b, inc_t is_b, \
|
|
dim_t pd_b, inc_t ps_b, \
|
|
void* beta, \
|
|
void* c, inc_t rs_c, inc_t cs_c, \
|
|
cntx_t* cntx, \
|
|
rntm_t* rntm, \
|
|
thrinfo_t* thread \
|
|
);
|
|
|
|
INSERT_GENTPROT2_BASIC0( gemm_ker_var2_md )
|
|
INSERT_GENTPROT2_MIXDP0( gemm_ker_var2_md )
|
|
|