gcc/libgfortran/m4/eoshift1.m4
Janne Blomqvist 7a15726687 Use pointer sized array indices.
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
2018-01-31 16:16:22 +02:00

325 lines
8.5 KiB
Plaintext

`/* Implementation of the EOSHIFT intrinsic
Copyright (C) 2002-2018 Free Software Foundation, Inc.
Contributed 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.
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.
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>'
include(iparm.m4)dnl
`#if defined (HAVE_'atype_name`)
static void
eoshift1 (gfc_array_char * const restrict ret,
const gfc_array_char * const restrict array,
const 'atype` * const restrict h,
const char * const restrict pbound,
const 'atype_name` * const restrict pwhich,
const char * filler, index_type filler_len)
{
/* r.* indicates the return array. */
index_type rstride[GFC_MAX_DIMENSIONS];
index_type rstride0;
index_type roffset;
char *rptr;
char * restrict dest;
/* s.* indicates the source array. */
index_type sstride[GFC_MAX_DIMENSIONS];
index_type sstride0;
index_type soffset;
const char *sptr;
const char *src;
/* h.* indicates the shift array. */
index_type hstride[GFC_MAX_DIMENSIONS];
index_type hstride0;
const 'atype_name` *hptr;
index_type count[GFC_MAX_DIMENSIONS];
index_type extent[GFC_MAX_DIMENSIONS];
index_type dim;
index_type len;
index_type n;
index_type size;
index_type arraysize;
int which;
'atype_name` sh;
'atype_name` delta;
/* The compiler cannot figure out that these are set, initialize
them to avoid warnings. */
len = 0;
soffset = 0;
roffset = 0;
size = GFC_DESCRIPTOR_SIZE(array);
if (pwhich)
which = *pwhich - 1;
else
which = 0;
extent[0] = 1;
count[0] = 0;
arraysize = size0 ((array_t *) array);
if (ret->base_addr == NULL)
{
ret->offset = 0;
GFC_DTYPE_COPY(ret,array);
for (index_type i = 0; i < GFC_DESCRIPTOR_RANK (array); i++)
{
index_type ub, str;
ub = GFC_DESCRIPTOR_EXTENT(array,i) - 1;
if (i == 0)
str = 1;
else
str = GFC_DESCRIPTOR_EXTENT(ret,i-1)
* GFC_DESCRIPTOR_STRIDE(ret,i-1);
GFC_DIMENSION_SET(ret->dim[i], 0, ub, str);
}
/* xmallocarray allocates a single byte for zero size. */
ret->base_addr = xmallocarray (arraysize, size);
}
else if (unlikely (compile_options.bounds_check))
{
bounds_equal_extents ((array_t *) ret, (array_t *) array,
"return value", "EOSHIFT");
}
if (unlikely (compile_options.bounds_check))
{
bounds_reduced_extents ((array_t *) h, (array_t *) array, which,
"SHIFT argument", "EOSHIFT");
}
if (arraysize == 0)
return;
n = 0;
for (dim = 0; dim < GFC_DESCRIPTOR_RANK (array); dim++)
{
if (dim == which)
{
roffset = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
if (roffset == 0)
roffset = size;
soffset = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
if (soffset == 0)
soffset = size;
len = GFC_DESCRIPTOR_EXTENT(array,dim);
}
else
{
count[n] = 0;
extent[n] = GFC_DESCRIPTOR_EXTENT(array,dim);
rstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(ret,dim);
sstride[n] = GFC_DESCRIPTOR_STRIDE_BYTES(array,dim);
hstride[n] = GFC_DESCRIPTOR_STRIDE(h,n);
n++;
}
}
if (sstride[0] == 0)
sstride[0] = size;
if (rstride[0] == 0)
rstride[0] = size;
if (hstride[0] == 0)
hstride[0] = 1;
dim = GFC_DESCRIPTOR_RANK (array);
rstride0 = rstride[0];
sstride0 = sstride[0];
hstride0 = hstride[0];
rptr = ret->base_addr;
sptr = array->base_addr;
hptr = h->base_addr;
while (rptr)
{
/* Do the shift for this dimension. */
sh = *hptr;
if (( sh >= 0 ? sh : -sh ) > len)
{
delta = len;
sh = len;
}
else
delta = (sh >= 0) ? sh: -sh;
if (sh > 0)
{
src = &sptr[delta * soffset];
dest = rptr;
}
else
{
src = sptr;
dest = &rptr[delta * roffset];
}
/* If the elements are contiguous, perform a single block move. */
if (soffset == size && roffset == size)
{
size_t chunk = size * (len - delta);
memcpy (dest, src, chunk);
dest += chunk;
}
else
{
for (n = 0; n < len - delta; n++)
{
memcpy (dest, src, size);
dest += roffset;
src += soffset;
}
}
if (sh < 0)
dest = rptr;
n = delta;
if (pbound)
while (n--)
{
memcpy (dest, pbound, size);
dest += roffset;
}
else
while (n--)
{
index_type i;
if (filler_len == 1)
memset (dest, filler[0], size);
else
for (i = 0; i < size; i += filler_len)
memcpy (&dest[i], filler, filler_len);
dest += roffset;
}
/* Advance to the next section. */
rptr += rstride0;
sptr += sstride0;
hptr += hstride0;
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. */
rptr -= rstride[n] * extent[n];
sptr -= sstride[n] * extent[n];
hptr -= hstride[n] * extent[n];
n++;
if (n >= dim - 1)
{
/* Break out of the loop. */
rptr = NULL;
break;
}
else
{
count[n]++;
rptr += rstride[n];
sptr += sstride[n];
hptr += hstride[n];
}
}
}
}
void eoshift1_'atype_kind` (gfc_array_char * const restrict,
const gfc_array_char * const restrict,
const 'atype` * const restrict, const char * const restrict,
const 'atype_name` * const restrict);
export_proto(eoshift1_'atype_kind`);
void
eoshift1_'atype_kind` (gfc_array_char * const restrict ret,
const gfc_array_char * const restrict array,
const 'atype` * const restrict h,
const char * const restrict pbound,
const 'atype_name` * const restrict pwhich)
{
eoshift1 (ret, array, h, pbound, pwhich, "\0", 1);
}
void eoshift1_'atype_kind`_char (gfc_array_char * const restrict,
GFC_INTEGER_4,
const gfc_array_char * const restrict,
const 'atype` * const restrict,
const char * const restrict,
const 'atype_name` * const restrict,
GFC_INTEGER_4, GFC_INTEGER_4);
export_proto(eoshift1_'atype_kind`_char);
void
eoshift1_'atype_kind`_char (gfc_array_char * const restrict ret,
GFC_INTEGER_4 ret_length __attribute__((unused)),
const gfc_array_char * const restrict array,
const 'atype` * const restrict h,
const char * const restrict pbound,
const 'atype_name` * const restrict pwhich,
GFC_INTEGER_4 array_length __attribute__((unused)),
GFC_INTEGER_4 bound_length __attribute__((unused)))
{
eoshift1 (ret, array, h, pbound, pwhich, " ", 1);
}
void eoshift1_'atype_kind`_char4 (gfc_array_char * const restrict,
GFC_INTEGER_4,
const gfc_array_char * const restrict,
const 'atype` * const restrict,
const char * const restrict,
const 'atype_name` * const restrict,
GFC_INTEGER_4, GFC_INTEGER_4);
export_proto(eoshift1_'atype_kind`_char4);
void
eoshift1_'atype_kind`_char4 (gfc_array_char * const restrict ret,
GFC_INTEGER_4 ret_length __attribute__((unused)),
const gfc_array_char * const restrict array,
const 'atype` * const restrict h,
const char * const restrict pbound,
const 'atype_name` * const restrict pwhich,
GFC_INTEGER_4 array_length __attribute__((unused)),
GFC_INTEGER_4 bound_length __attribute__((unused)))
{
static const gfc_char4_t space = (unsigned char) ''` ''`;
eoshift1 (ret, array, h, pbound, pwhich,
(const char *) &space, sizeof (gfc_char4_t));
}
#endif'