7a15726687
Using pointer sized variables (e.g. size_t / ptrdiff_t) when the variables are used as array indices allows accessing larger arrays, and can be a slight performance improvement due to no need for sign or zero extending, or masking. Regtested on x86_64-pc-linux-gnu. libgfortran/ChangeLog: 2018-01-31 Janne Blomqvist <jb@gcc.gnu.org> * generated/cshift1_16.c (cshift1): Regenerated. * generated/cshift1_4.c (cshift1): Regenerated. * generated/cshift1_8.c (cshift1): Regenerated. * generated/eoshift1_16.c (eoshift1): Regenerated. * generated/eoshift1_4.c (eoshift1): Regenerated. * generated/eoshift1_8.c (eoshift1): Regenerated. * generated/eoshift3_16.c (eoshift3): Regenerated. * generated/eoshift3_4.c (eoshift3): Regenerated. * generated/eoshift3_8.c (eoshift3): Regenerated. * generated/in_pack_c10.c (internal_pack_c10): Regenerated. * generated/in_pack_c16.c (internal_pack_c16): Regenerated. * generated/in_pack_c4.c (internal_pack_c4): Regenerated. * generated/in_pack_c8.c (internal_pack_c8): Regenerated. * generated/in_pack_i1.c (internal_pack_1): Regenerated. * generated/in_pack_i16.c (internal_pack_16): Regenerated. * generated/in_pack_i2.c (internal_pack_2): Regenerated. * generated/in_pack_i4.c (internal_pack_4): Regenerated. * generated/in_pack_i8.c (internal_pack_8): Regenerated. * generated/in_pack_r10.c (internal_pack_r10): Regenerated. * generated/in_pack_r16.c (internal_pack_r16): Regenerated. * generated/in_pack_r4.c (internal_pack_r4): Regenerated. * generated/in_pack_r8.c (internal_pack_r8): Regenerated. * generated/in_unpack_c10.c (internal_unpack_c10): Regenerated. * generated/in_unpack_c16.c (internal_unpack_c16): Regenerated. * generated/in_unpack_c4.c (internal_unpack_c4): Regenerated. * generated/in_unpack_c8.c (internal_unpack_c8): Regenerated. * generated/in_unpack_i1.c (internal_unpack_1): Regenerated. * generated/in_unpack_i16.c (internal_unpack_16): Regenerated. * generated/in_unpack_i2.c (internal_unpack_2): Regenerated. * generated/in_unpack_i4.c (internal_unpack_4): Regenerated. * generated/in_unpack_i8.c (internal_unpack_8): Regenerated. * generated/in_unpack_r10.c (internal_unpack_r10): Regenerated. * generated/in_unpack_r16.c (internal_unpack_r16): Regenerated. * generated/in_unpack_r4.c (internal_unpack_r4): Regenerated. * generated/in_unpack_r8.c (internal_unpack_r8): Regenerated. * generated/reshape_c10.c (reshape_c10): Regenerated. * generated/reshape_c16.c (reshape_c16): Regenerated. * generated/reshape_c4.c (reshape_c4): Regenerated. * generated/reshape_c8.c (reshape_c8): Regenerated. * generated/reshape_i16.c (reshape_16): Regenerated. * generated/reshape_i4.c (reshape_4): Regenerated. * generated/reshape_i8.c (reshape_8): Regenerated. * generated/reshape_r10.c (reshape_r10): Regenerated. * generated/reshape_r16.c (reshape_r16): Regenerated. * generated/reshape_r4.c (reshape_r4): Regenerated. * generated/reshape_r8.c (reshape_r8): Regenerated. * generated/shape_i1.c (shape_1): Regenerated. * generated/shape_i16.c (shape_16): Regenerated. * generated/shape_i2.c (shape_2): Regenerated. * generated/shape_i4.c (shape_4): Regenerated. * generated/shape_i8.c (shape_8): Regenerated. * generated/spread_c10.c (spread_scalar_c10): Regenerated. * generated/spread_c16.c (spread_scalar_c16): Regenerated. * generated/spread_c4.c (spread_scalar_c4): Regenerated. * generated/spread_c8.c (spread_scalar_c8): Regenerated. * generated/spread_i1.c (spread_scalar_i1): Regenerated. * generated/spread_i16.c (spread_scalar_i16): Regenerated. * generated/spread_i2.c (spread_scalar_i2): Regenerated. * generated/spread_i4.c (spread_scalar_i4): Regenerated. * generated/spread_i8.c (spread_scalar_i8): Regenerated. * generated/spread_r10.c (spread_scalar_r10): Regenerated. * generated/spread_r16.c (spread_scalar_r16): Regenerated. * generated/spread_r4.c (spread_scalar_r4): Regenerated. * generated/spread_r8.c (spread_scalar_r8): Regenerated. * intrinsics/random.c (jump): Use size_t for array index in loop. (getosrandom): Likewise. (arandom_r4): Make n an index_type. (arandom_r8): Likewise. (arandom_r10): Likewise. (arandom_r16): Likewise. (scramble_seed): Use size_t for array index in loop. * m4/cshift1.m4: Make i an index_type. * m4/eoshift1.m4: Likewise. * m4/eoshift3.m4: Likewise. * m4/in_pack.m4: Make n an index_type. * m4/in_unpack.m4: Likewise. * m4/reshape.m4: Make n and dim index_type's. * m4/shape.m4: Make n an index_type. * m4/spread.m4: Likewise, use index_type argument rather than copying to int. * runtime/bounds.c (bounds_ifunction_return): Make n an index_type. * runtime/in_pack_generic.c (internal_pack): Likewise. * runtime/in_unpack_generic.c (internal_unpack): Make n and size index_type's. From-SVN: r257234
266 lines
6.7 KiB
C
266 lines
6.7 KiB
C
/* Special implementation of the SPREAD intrinsic
|
|
Copyright (C) 2008-2018 Free Software Foundation, Inc.
|
|
Contributed by Thomas Koenig <tkoenig@gcc.gnu.org>, based on
|
|
spread_generic.c written by Paul Brook <paul@nowt.org>
|
|
|
|
This file is part of the GNU Fortran 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 3 of the License, or (at your option) any later version.
|
|
|
|
Ligbfortran 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.
|
|
|
|
Under Section 7 of GPL version 3, you are granted additional
|
|
permissions described in the GCC Runtime Library Exception, version
|
|
3.1, as published by the Free Software Foundation.
|
|
|
|
You should have received a copy of the GNU General Public License and
|
|
a copy of the GCC Runtime Library Exception along with this program;
|
|
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
|
<http://www.gnu.org/licenses/>. */
|
|
|
|
#include "libgfortran.h"
|
|
#include <string.h>
|
|
|
|
|
|
#if defined (HAVE_GFC_INTEGER_4)
|
|
|
|
void
|
|
spread_i4 (gfc_array_i4 *ret, const gfc_array_i4 *source,
|
|
const index_type along, const index_type pncopies)
|
|
{
|
|
/* r.* indicates the return array. */
|
|
index_type rstride[GFC_MAX_DIMENSIONS];
|
|
index_type rstride0;
|
|
index_type rdelta = 0;
|
|
index_type rrank;
|
|
index_type rs;
|
|
GFC_INTEGER_4 *rptr;
|
|
GFC_INTEGER_4 * restrict dest;
|
|
/* s.* indicates the source array. */
|
|
index_type sstride[GFC_MAX_DIMENSIONS];
|
|
index_type sstride0;
|
|
index_type srank;
|
|
const GFC_INTEGER_4 *sptr;
|
|
|
|
index_type count[GFC_MAX_DIMENSIONS];
|
|
index_type extent[GFC_MAX_DIMENSIONS];
|
|
index_type n;
|
|
index_type dim;
|
|
index_type ncopies;
|
|
|
|
srank = GFC_DESCRIPTOR_RANK(source);
|
|
|
|
rrank = srank + 1;
|
|
if (rrank > GFC_MAX_DIMENSIONS)
|
|
runtime_error ("return rank too large in spread()");
|
|
|
|
if (along > rrank)
|
|
runtime_error ("dim outside of rank in spread()");
|
|
|
|
ncopies = pncopies;
|
|
|
|
if (ret->base_addr == NULL)
|
|
{
|
|
|
|
size_t ub, stride;
|
|
|
|
/* The front end has signalled that we need to populate the
|
|
return array descriptor. */
|
|
GFC_DTYPE_COPY_SETRANK(ret,source,rrank);
|
|
dim = 0;
|
|
rs = 1;
|
|
for (n = 0; n < rrank; n++)
|
|
{
|
|
stride = rs;
|
|
if (n == along - 1)
|
|
{
|
|
ub = ncopies - 1;
|
|
rdelta = rs;
|
|
rs *= ncopies;
|
|
}
|
|
else
|
|
{
|
|
count[dim] = 0;
|
|
extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
rstride[dim] = rs;
|
|
|
|
ub = extent[dim] - 1;
|
|
rs *= extent[dim];
|
|
dim++;
|
|
}
|
|
GFC_DIMENSION_SET(ret->dim[n], 0, ub, stride);
|
|
}
|
|
ret->offset = 0;
|
|
|
|
/* xmallocarray allocates a single byte for zero size. */
|
|
ret->base_addr = xmallocarray (rs, sizeof(GFC_INTEGER_4));
|
|
if (rs <= 0)
|
|
return;
|
|
}
|
|
else
|
|
{
|
|
int zero_sized;
|
|
|
|
zero_sized = 0;
|
|
|
|
dim = 0;
|
|
if (GFC_DESCRIPTOR_RANK(ret) != rrank)
|
|
runtime_error ("rank mismatch in spread()");
|
|
|
|
if (unlikely (compile_options.bounds_check))
|
|
{
|
|
for (n = 0; n < rrank; n++)
|
|
{
|
|
index_type ret_extent;
|
|
|
|
ret_extent = GFC_DESCRIPTOR_EXTENT(ret,n);
|
|
if (n == along - 1)
|
|
{
|
|
rdelta = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
|
|
if (ret_extent != ncopies)
|
|
runtime_error("Incorrect extent in return value of SPREAD"
|
|
" intrinsic in dimension %ld: is %ld,"
|
|
" should be %ld", (long int) n+1,
|
|
(long int) ret_extent, (long int) ncopies);
|
|
}
|
|
else
|
|
{
|
|
count[dim] = 0;
|
|
extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
if (ret_extent != extent[dim])
|
|
runtime_error("Incorrect extent in return value of SPREAD"
|
|
" intrinsic in dimension %ld: is %ld,"
|
|
" should be %ld", (long int) n+1,
|
|
(long int) ret_extent,
|
|
(long int) extent[dim]);
|
|
|
|
if (extent[dim] <= 0)
|
|
zero_sized = 1;
|
|
sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
rstride[dim] = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
dim++;
|
|
}
|
|
}
|
|
}
|
|
else
|
|
{
|
|
for (n = 0; n < rrank; n++)
|
|
{
|
|
if (n == along - 1)
|
|
{
|
|
rdelta = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
}
|
|
else
|
|
{
|
|
count[dim] = 0;
|
|
extent[dim] = GFC_DESCRIPTOR_EXTENT(source,dim);
|
|
if (extent[dim] <= 0)
|
|
zero_sized = 1;
|
|
sstride[dim] = GFC_DESCRIPTOR_STRIDE(source,dim);
|
|
rstride[dim] = GFC_DESCRIPTOR_STRIDE(ret,n);
|
|
dim++;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (zero_sized)
|
|
return;
|
|
|
|
if (sstride[0] == 0)
|
|
sstride[0] = 1;
|
|
}
|
|
sstride0 = sstride[0];
|
|
rstride0 = rstride[0];
|
|
rptr = ret->base_addr;
|
|
sptr = source->base_addr;
|
|
|
|
while (sptr)
|
|
{
|
|
/* Spread this element. */
|
|
dest = rptr;
|
|
for (n = 0; n < ncopies; n++)
|
|
{
|
|
*dest = *sptr;
|
|
dest += rdelta;
|
|
}
|
|
/* Advance to the next element. */
|
|
sptr += sstride0;
|
|
rptr += rstride0;
|
|
count[0]++;
|
|
n = 0;
|
|
while (count[n] == extent[n])
|
|
{
|
|
/* When we get to the end of a dimension, reset it and increment
|
|
the next dimension. */
|
|
count[n] = 0;
|
|
/* We could precalculate these products, but this is a less
|
|
frequently used path so probably not worth it. */
|
|
sptr -= sstride[n] * extent[n];
|
|
rptr -= rstride[n] * extent[n];
|
|
n++;
|
|
if (n >= srank)
|
|
{
|
|
/* Break out of the loop. */
|
|
sptr = NULL;
|
|
break;
|
|
}
|
|
else
|
|
{
|
|
count[n]++;
|
|
sptr += sstride[n];
|
|
rptr += rstride[n];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* This version of spread_internal treats the special case of a scalar
|
|
source. This is much simpler than the more general case above. */
|
|
|
|
void
|
|
spread_scalar_i4 (gfc_array_i4 *ret, const GFC_INTEGER_4 *source,
|
|
const index_type along, const index_type ncopies)
|
|
{
|
|
GFC_INTEGER_4 * restrict dest;
|
|
index_type stride;
|
|
|
|
if (GFC_DESCRIPTOR_RANK (ret) != 1)
|
|
runtime_error ("incorrect destination rank in spread()");
|
|
|
|
if (along > 1)
|
|
runtime_error ("dim outside of rank in spread()");
|
|
|
|
if (ret->base_addr == NULL)
|
|
{
|
|
ret->base_addr = xmallocarray (ncopies, sizeof (GFC_INTEGER_4));
|
|
ret->offset = 0;
|
|
GFC_DIMENSION_SET(ret->dim[0], 0, ncopies - 1, 1);
|
|
}
|
|
else
|
|
{
|
|
if (ncopies - 1 > (GFC_DESCRIPTOR_EXTENT(ret,0) - 1)
|
|
/ GFC_DESCRIPTOR_STRIDE(ret,0))
|
|
runtime_error ("dim too large in spread()");
|
|
}
|
|
|
|
dest = ret->base_addr;
|
|
stride = GFC_DESCRIPTOR_STRIDE(ret,0);
|
|
|
|
for (index_type n = 0; n < ncopies; n++)
|
|
{
|
|
*dest = *source;
|
|
dest += stride;
|
|
}
|
|
}
|
|
|
|
#endif
|
|
|