1/* { dg-do run } */
2/* { dg-require-effective-target fma4 } */
3/* { dg-options "-O0 -mfma4" } */
4
5#include "fma4-check.h"
6
7#include <x86intrin.h>
8#include <string.h>
9
10#define NUM 20
11
12union
13{
14  __m128 x[NUM];
15  float f[NUM * 4];
16  __m128d y[NUM];
17  double d[NUM * 2];
18} dst, res, src1, src2, src3;
19
20
21/* Note that in macc*,msub*,mnmacc* and mnsub* instructions, the intermdediate
22   product is not rounded, only the addition is rounded. */
23
24static void
25init_maccps ()
26{
27  int i;
28  for (i = 0; i < NUM * 4; i++)
29    {
30      src1.f[i] = i;
31      src2.f[i] = i + 10;
32      src3.f[i] = i + 20;
33    }
34}
35
36static void
37init_maccpd ()
38{
39  int i;
40  for (i = 0; i < NUM * 4; i++)
41    {
42      src1.d[i] = i;
43      src2.d[i] = i + 10;
44      src3.d[i] = i + 20;
45    }
46}
47
48static int
49check_maccps ()
50{
51  int i, j, check_fails = 0;
52  for (i = 0; i < NUM * 4; i = i + 4)
53    for (j = 0; j < 4; j++)
54      {
55	res.f[i + j] = (src1.f[i + j] * src2.f[i + j]) + src3.f[i + j];
56	if (dst.f[i + j] != res.f[i + j])
57	  check_fails++;
58      }
59  return check_fails++;
60}
61
62static int
63check_maccpd ()
64{
65  int i, j, check_fails = 0;
66  for (i = 0; i < NUM * 2; i = i + 2)
67    for (j = 0; j < 2; j++)
68      {
69	res.d[i + j] = (src1.d[i + j] * src2.d[i + j]) + src3.d[i + j];
70	if (dst.d[i + j] != res.d[i + j])
71	  check_fails++;
72      }
73  return check_fails++;
74}
75
76
77static int
78check_maccss ()
79{
80  int i, j, check_fails = 0;
81  for (i = 0; i < NUM * 4; i= i + 4)
82    {
83      res.f[i] = (src1.f[i] * src2.f[i]) + src3.f[i];
84      if (dst.f[i] != res.f[i])
85	check_fails++;
86    }
87  return check_fails++;
88}
89
90static int
91check_maccsd ()
92{
93  int i, j, check_fails = 0;
94  for (i = 0; i < NUM * 2; i = i + 2)
95    {
96      res.d[i] = (src1.d[i] * src2.d[i]) + src3.d[i];
97      if (dst.d[i] != res.d[i])
98	check_fails++;
99    }
100  return check_fails++;
101}
102
103static void
104fma4_test (void)
105{
106  int i;
107
108  init_maccps ();
109
110  for (i = 0; i < NUM; i++)
111    dst.x[i] = _mm_macc_ps (src1.x[i], src2.x[i], src3.x[i]);
112
113  if (check_maccps ())
114    abort ();
115
116  for (i = 0; i < NUM; i++)
117    dst.x[i] = _mm_macc_ss (src1.x[i], src2.x[i], src3.x[i]);
118
119  if (check_maccss ())
120    abort ();
121
122  init_maccpd ();
123
124  for (i = 0; i < NUM; i++)
125    dst.y[i] = _mm_macc_pd (src1.y[i], src2.y[i], src3.y[i]);
126
127  if (check_maccpd ())
128    abort ();
129
130  for (i = 0; i < NUM; i++)
131    dst.y[i] = _mm_macc_sd (src1.y[i], src2.y[i], src3.y[i]);
132
133  if (check_maccsd ())
134    abort ();
135
136}
137