Viewing file:  ssse3-palignr.c (6.02 KB) -rw-r--r--
Select action/file-type:
 (+) |  (+) |  (+) | Code (+) | Session (+) |  (+) | SDB (+) |  (+) |  (+) |  (+) |  (+) |  (+) |
/* { dg-do run } */
/* { dg-options "-O3 -mpower8-vector -Wno-psabi" } */
/* { dg-require-effective-target p8vector_hw } */
#ifndef CHECK_H
#define CHECK_H "ssse3-check.h"
#endif
#ifndef TEST
#define TEST ssse3_test
#endif
#include CHECK_H
#include "ssse3-vals.h"
#include <tmmintrin.h>
#include <string.h>
#ifndef __AVX__
/* Test the 64-bit form */
static void
ssse3_test_palignr (__m64 *i1, __m64 *i2, unsigned int imm, __m64 *r)
{
switch (imm)
{
case 0:
*r = _mm_alignr_pi8 (*i1, *i2, 0);
break;
case 1:
*r = _mm_alignr_pi8 (*i1, *i2, 1);
break;
case 2:
*r = _mm_alignr_pi8 (*i1, *i2, 2);
break;
case 3:
*r = _mm_alignr_pi8 (*i1, *i2, 3);
break;
case 4:
*r = _mm_alignr_pi8 (*i1, *i2, 4);
break;
case 5:
*r = _mm_alignr_pi8 (*i1, *i2, 5);
break;
case 6:
*r = _mm_alignr_pi8 (*i1, *i2, 6);
break;
case 7:
*r = _mm_alignr_pi8 (*i1, *i2, 7);
break;
case 8:
*r = _mm_alignr_pi8 (*i1, *i2, 8);
break;
case 9:
*r = _mm_alignr_pi8 (*i1, *i2, 9);
break;
case 10:
*r = _mm_alignr_pi8 (*i1, *i2, 10);
break;
case 11:
*r = _mm_alignr_pi8 (*i1, *i2, 11);
break;
case 12:
*r = _mm_alignr_pi8 (*i1, *i2, 12);
break;
case 13:
*r = _mm_alignr_pi8 (*i1, *i2, 13);
break;
case 14:
*r = _mm_alignr_pi8 (*i1, *i2, 14);
break;
case 15:
*r = _mm_alignr_pi8 (*i1, *i2, 15);
break;
default:
*r = _mm_alignr_pi8 (*i1, *i2, 16);
break;
}
_mm_empty();
}
#endif
/* Test the 128-bit form */
static void
ssse3_test_palignr128 (__m128i *i1, __m128i *i2, unsigned int imm, __m128i *r)
{
/* Assumes incoming pointers are 16-byte aligned */
switch (imm)
{
case 0:
*r = _mm_alignr_epi8 (*i1, *i2, 0);
break;
case 1:
*r = _mm_alignr_epi8 (*i1, *i2, 1);
break;
case 2:
*r = _mm_alignr_epi8 (*i1, *i2, 2);
break;
case 3:
*r = _mm_alignr_epi8 (*i1, *i2, 3);
break;
case 4:
*r = _mm_alignr_epi8 (*i1, *i2, 4);
break;
case 5:
*r = _mm_alignr_epi8 (*i1, *i2, 5);
break;
case 6:
*r = _mm_alignr_epi8 (*i1, *i2, 6);
break;
case 7:
*r = _mm_alignr_epi8 (*i1, *i2, 7);
break;
case 8:
*r = _mm_alignr_epi8 (*i1, *i2, 8);
break;
case 9:
*r = _mm_alignr_epi8 (*i1, *i2, 9);
break;
case 10:
*r = _mm_alignr_epi8 (*i1, *i2, 10);
break;
case 11:
*r = _mm_alignr_epi8 (*i1, *i2, 11);
break;
case 12:
*r = _mm_alignr_epi8 (*i1, *i2, 12);
break;
case 13:
*r = _mm_alignr_epi8 (*i1, *i2, 13);
break;
case 14:
*r = _mm_alignr_epi8 (*i1, *i2, 14);
break;
case 15:
*r = _mm_alignr_epi8 (*i1, *i2, 15);
break;
case 16:
*r = _mm_alignr_epi8 (*i1, *i2, 16);
break;
case 17:
*r = _mm_alignr_epi8 (*i1, *i2, 17);
break;
case 18:
*r = _mm_alignr_epi8 (*i1, *i2, 18);
break;
case 19:
*r = _mm_alignr_epi8 (*i1, *i2, 19);
break;
case 20:
*r = _mm_alignr_epi8 (*i1, *i2, 20);
break;
case 21:
*r = _mm_alignr_epi8 (*i1, *i2, 21);
break;
case 22:
*r = _mm_alignr_epi8 (*i1, *i2, 22);
break;
case 23:
*r = _mm_alignr_epi8 (*i1, *i2, 23);
break;
case 24:
*r = _mm_alignr_epi8 (*i1, *i2, 24);
break;
case 25:
*r = _mm_alignr_epi8 (*i1, *i2, 25);
break;
case 26:
*r = _mm_alignr_epi8 (*i1, *i2, 26);
break;
case 27:
*r = _mm_alignr_epi8 (*i1, *i2, 27);
break;
case 28:
*r = _mm_alignr_epi8 (*i1, *i2, 28);
break;
case 29:
*r = _mm_alignr_epi8 (*i1, *i2, 29);
break;
case 30:
*r = _mm_alignr_epi8 (*i1, *i2, 30);
break;
case 31:
*r = _mm_alignr_epi8 (*i1, *i2, 31);
break;
default:
*r = _mm_alignr_epi8 (*i1, *i2, 32);
break;
}
}
/* Routine to manually compute the results */
static void
compute_correct_result_128 (signed char *i1, signed char *i2, unsigned int imm,
signed char *r)
{
signed char buf [32];
int i;
memcpy (&buf[0], i2, 16);
memcpy (&buf[16], i1, 16);
for (i = 0; i < 16; i++)
if (imm >= 32 || imm + i >= 32)
r[i] = 0;
else
r[i] = buf[imm + i];
}
#ifndef __AVX__
static void
compute_correct_result_64 (signed char *i1, signed char *i2, unsigned int imm,
signed char *r)
{
signed char buf [16];
int i;
/* Handle the first half */
memcpy (&buf[0], &i2[0], 8);
memcpy (&buf[8], &i1[0], 8);
for (i = 0; i < 8; i++)
if (imm >= 16 || imm + i >= 16)
r[i] = 0;
else
r[i] = buf[imm + i];
/* Handle the second half */
memcpy (&buf[0], &i2[8], 8);
memcpy (&buf[8], &i1[8], 8);
for (i = 0; i < 8; i++)
if (imm >= 16 || imm + i >= 16)
r[i + 8] = 0;
else
r[i + 8] = buf[imm + i];
}
#endif
static void
TEST (void)
{
int i;
union data r __attribute__ ((aligned(16)));
union data ck;
unsigned int imm;
int fail = 0;
for (i = 0; i < ARRAY_SIZE (vals) - 1; i++)
for (imm = 0; imm < 100; imm++)
{
#ifndef __AVX__
/* Manually compute the result */
compute_correct_result_64 (&vals[i + 0].b[0],
&vals[i + 1].b[0], imm, &ck.b[0]);
/* Run the 64-bit tests */
ssse3_test_palignr (&vals[i + 0].ll[0],
&vals[i + 1].ll[0], imm, &r.ll[0]);
ssse3_test_palignr (&vals[i + 0].ll[1],
&vals[i + 1].ll[1], imm, &r.ll[1]);
fail += chk_128 (ck.m[0], r.m[0]);
#endif
/* Recompute the results for 128-bits */
compute_correct_result_128 (&vals[i + 0].b[0],
&vals[i + 1].b[0], imm, &ck.b[0]);
/* Run the 128-bit tests */
ssse3_test_palignr128 (&vals[i + 0].m[0],
&vals[i + 1].m[0], imm, &r.m[0]);
fail += chk_128 (ck.m[0], r.m[0]);
}
if (fail != 0)
abort ();
}
|
