mirror of
git://gcc.gnu.org/git/gcc.git
synced 2024-12-29 08:54:42 +08:00
644cb69f80
PR libfortran/19308 PR fortran/20120 PR libfortran/22437 * Makefile.am: Add generated files for large real and integers kinds. Add a rule to create the kinds.inc c99_protos.inc files. Use kinds.inc to preprocess Fortran generated files. * libgfortran.h: Add macro definitions for GFC_INTEGER_16_HUGE, GFC_REAL_10_HUGE and GFC_REAL_16_HUGE. Add types gfc_array_i16, gfc_array_r10, gfc_array_r16, gfc_array_c10, gfc_array_c16, gfc_array_l16. * mk-kinds-h.sh: Define macros HAVE_GFC_LOGICAL_* and HAVE_GFC_COMPLEX_* when these types are available. * intrinsics/ishftc.c (ishftc16): New function for GFC_INTEGER_16. * m4/all.m4, m4/any.m4, m4/count.m4, m4/cshift1.m4, m4/dotprod.m4, m4/dotprodc.m4, m4/dotprodl.m4, m4/eoshift1.m4, m4/eoshift3.m4, m4/exponent.m4, m4/fraction.m4, m4/in_pack.m4, m4/in_unpack.m4, m4/matmul.m4, m4/matmull.m4, m4/maxloc0.m4, m4/maxloc1.m4, m4/maxval.m4, m4/minloc0.m4, m4/minloc1.m4, m4/minval.m4, m4/mtype.m4, m4/nearest.m4, m4/pow.m4, m4/product.m4, m4/reshape.m4, m4/set_exponent.m4, m4/shape.m4, m4/specific.m4, m4/specific2.m4, m4/sum.m4, m4/transpose.m4: Protect generated functions with appropriate "#if defined (HAVE_GFC_type_kind)" preprocessor directives. * Makefile.in: Regenerate. * all files in generated/: Regenerate. * f95-lang.c (DO_DEFINE_MATH_BUILTIN): Add support for long double builtin function. (gfc_init_builtin_functions): Add mfunc_longdouble, mfunc_clongdouble and func_clongdouble_longdouble trees. Build them for round, trunc, cabs, copysign and pow functions. * iresolve.c (gfc_resolve_reshape, gfc_resolve_transpose): Add case for kind 10 and 16. * trans-decl.c: Add trees for cpowl10, cpowl16, ishftc16, exponent10 and exponent16. (gfc_build_intrinsic_function_decls): Build nodes for int16, real10, real16, complex10 and complex16 types. Build all possible combinations for function _gfortran_pow_?n_?n. Build function calls cpowl10, cpowl16, ishftc16, exponent10 and exponent16. * trans-expr.c (gfc_conv_power_op): Add case for integer(16), real(10) and real(16). * trans-intrinsic.c: Add suppport for long double builtin functions in BUILT_IN_FUNCTION, LIBM_FUNCTION and LIBF_FUNCTION macros. (gfc_conv_intrinsic_aint): Add case for integer(16), real(10) and real(16) kinds. (gfc_build_intrinsic_lib_fndecls): Add support for real10_decl and real16_decl in library functions. (gfc_get_intrinsic_lib_fndecl): Add cases for real and complex kinds 10 and 16. (gfc_conv_intrinsic_exponent): Add cases for real(10) and real(16) kinds. (gfc_conv_intrinsic_sign): Likewise. (gfc_conv_intrinsic_ishftc): Add case for integer(16) kind. * trans-types.c (gfc_get_int_type, gfc_get_real_type, gfc_get_complex_type, gfc_get_logical_type): Doesn't error out in the case of kinds not available. * trans.h: Declare trees for cpowl10, cpowl16, ishftc16, exponent10 and exponent16. * gfortran.dg/large_real_kind_2.F90: New test. * gfortran.dg/large_integer_kind_2.f90: New test. From-SVN: r104889
199 lines
5.6 KiB
Plaintext
199 lines
5.6 KiB
Plaintext
`/* Implementation of the MATMUL intrinsic
|
|
Copyright 2002, 2005 Free Software Foundation, Inc.
|
|
Contributed by Paul Brook <paul@nowt.org>
|
|
|
|
This file is part of the GNU Fortran 95 runtime library (libgfortran).
|
|
|
|
Libgfortran is free software; you can redistribute it and/or
|
|
modify it under the terms of the GNU General Public
|
|
License as published by the Free Software Foundation; either
|
|
version 2 of the License, or (at your option) any later version.
|
|
|
|
In addition to the permissions in the GNU General Public License, the
|
|
Free Software Foundation gives you unlimited permission to link the
|
|
compiled version of this file into combinations with other programs,
|
|
and to distribute those combinations without any restriction coming
|
|
from the use of this file. (The General Public License restrictions
|
|
do apply in other respects; for example, they cover modification of
|
|
the file, and distribution when not linked into a combine
|
|
executable.)
|
|
|
|
Libgfortran is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public
|
|
License along with libgfortran; see the file COPYING. If not,
|
|
write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
|
|
Boston, MA 02110-1301, USA. */
|
|
|
|
#include "config.h"
|
|
#include <stdlib.h>
|
|
#include <assert.h>
|
|
#include "libgfortran.h"'
|
|
include(iparm.m4)dnl
|
|
|
|
`#if defined (HAVE_'rtype_name`)'
|
|
|
|
/* Dimensions: retarray(x,y) a(x, count) b(count,y).
|
|
Either a or b can be rank 1. In this case x or y is 1. */
|
|
|
|
extern void matmul_`'rtype_code (rtype *, gfc_array_l4 *, gfc_array_l4 *);
|
|
export_proto(matmul_`'rtype_code);
|
|
|
|
void
|
|
matmul_`'rtype_code (rtype * retarray, gfc_array_l4 * a, gfc_array_l4 * b)
|
|
{
|
|
GFC_INTEGER_4 *abase;
|
|
GFC_INTEGER_4 *bbase;
|
|
rtype_name *dest;
|
|
index_type rxstride;
|
|
index_type rystride;
|
|
index_type xcount;
|
|
index_type ycount;
|
|
index_type xstride;
|
|
index_type ystride;
|
|
index_type x;
|
|
index_type y;
|
|
|
|
GFC_INTEGER_4 *pa;
|
|
GFC_INTEGER_4 *pb;
|
|
index_type astride;
|
|
index_type bstride;
|
|
index_type count;
|
|
index_type n;
|
|
|
|
assert (GFC_DESCRIPTOR_RANK (a) == 2
|
|
|| GFC_DESCRIPTOR_RANK (b) == 2);
|
|
|
|
if (retarray->data == NULL)
|
|
{
|
|
if (GFC_DESCRIPTOR_RANK (a) == 1)
|
|
{
|
|
retarray->dim[0].lbound = 0;
|
|
retarray->dim[0].ubound = b->dim[1].ubound - b->dim[1].lbound;
|
|
retarray->dim[0].stride = 1;
|
|
}
|
|
else if (GFC_DESCRIPTOR_RANK (b) == 1)
|
|
{
|
|
retarray->dim[0].lbound = 0;
|
|
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
|
|
retarray->dim[0].stride = 1;
|
|
}
|
|
else
|
|
{
|
|
retarray->dim[0].lbound = 0;
|
|
retarray->dim[0].ubound = a->dim[0].ubound - a->dim[0].lbound;
|
|
retarray->dim[0].stride = 1;
|
|
|
|
retarray->dim[1].lbound = 0;
|
|
retarray->dim[1].ubound = b->dim[1].ubound - b->dim[1].lbound;
|
|
retarray->dim[1].stride = retarray->dim[0].ubound+1;
|
|
}
|
|
|
|
retarray->data
|
|
= internal_malloc_size (sizeof (rtype_name) * size0 ((array_t *) retarray));
|
|
retarray->offset = 0;
|
|
}
|
|
|
|
abase = a->data;
|
|
if (GFC_DESCRIPTOR_SIZE (a) != 4)
|
|
{
|
|
assert (GFC_DESCRIPTOR_SIZE (a) == 8);
|
|
abase = GFOR_POINTER_L8_TO_L4 (abase);
|
|
}
|
|
bbase = b->data;
|
|
if (GFC_DESCRIPTOR_SIZE (b) != 4)
|
|
{
|
|
assert (GFC_DESCRIPTOR_SIZE (b) == 8);
|
|
bbase = GFOR_POINTER_L8_TO_L4 (bbase);
|
|
}
|
|
dest = retarray->data;
|
|
|
|
if (retarray->dim[0].stride == 0)
|
|
retarray->dim[0].stride = 1;
|
|
if (a->dim[0].stride == 0)
|
|
a->dim[0].stride = 1;
|
|
if (b->dim[0].stride == 0)
|
|
b->dim[0].stride = 1;
|
|
|
|
sinclude(`matmul_asm_'rtype_code`.m4')dnl
|
|
|
|
if (GFC_DESCRIPTOR_RANK (retarray) == 1)
|
|
{
|
|
rxstride = retarray->dim[0].stride;
|
|
rystride = rxstride;
|
|
}
|
|
else
|
|
{
|
|
rxstride = retarray->dim[0].stride;
|
|
rystride = retarray->dim[1].stride;
|
|
}
|
|
|
|
/* If we have rank 1 parameters, zero the absent stride, and set the size to
|
|
one. */
|
|
if (GFC_DESCRIPTOR_RANK (a) == 1)
|
|
{
|
|
astride = a->dim[0].stride;
|
|
count = a->dim[0].ubound + 1 - a->dim[0].lbound;
|
|
xstride = 0;
|
|
rxstride = 0;
|
|
xcount = 1;
|
|
}
|
|
else
|
|
{
|
|
astride = a->dim[1].stride;
|
|
count = a->dim[1].ubound + 1 - a->dim[1].lbound;
|
|
xstride = a->dim[0].stride;
|
|
xcount = a->dim[0].ubound + 1 - a->dim[0].lbound;
|
|
}
|
|
if (GFC_DESCRIPTOR_RANK (b) == 1)
|
|
{
|
|
bstride = b->dim[0].stride;
|
|
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
|
|
ystride = 0;
|
|
rystride = 0;
|
|
ycount = 1;
|
|
}
|
|
else
|
|
{
|
|
bstride = b->dim[0].stride;
|
|
assert(count == b->dim[0].ubound + 1 - b->dim[0].lbound);
|
|
ystride = b->dim[1].stride;
|
|
ycount = b->dim[1].ubound + 1 - b->dim[1].lbound;
|
|
}
|
|
|
|
for (y = 0; y < ycount; y++)
|
|
{
|
|
for (x = 0; x < xcount; x++)
|
|
{
|
|
/* Do the summation for this element. For real and integer types
|
|
this is the same as DOT_PRODUCT. For complex types we use do
|
|
a*b, not conjg(a)*b. */
|
|
pa = abase;
|
|
pb = bbase;
|
|
*dest = 0;
|
|
|
|
for (n = 0; n < count; n++)
|
|
{
|
|
if (*pa && *pb)
|
|
{
|
|
*dest = 1;
|
|
break;
|
|
}
|
|
pa += astride;
|
|
pb += bstride;
|
|
}
|
|
|
|
dest += rxstride;
|
|
abase += xstride;
|
|
}
|
|
abase -= xstride * xcount;
|
|
bbase += ystride;
|
|
dest += rystride - (rxstride * xcount);
|
|
}
|
|
}
|
|
|
|
#endif
|