1/* { dg-do compile { target { powerpc*-*-* } } } */
2/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
3/* { dg-require-effective-target powerpc_vsx_ok } */
4/* { dg-skip-if "do not override -mcpu" { powerpc*-*-* } { "-mcpu=*" } { "-mcpu=power7" } } */
5/* { dg-options "-O3 -ftree-vectorize -mcpu=power7 -ffast-math" } */
6/* { dg-final { scan-assembler-times "xvmadd" 4 } } */
7/* { dg-final { scan-assembler-times "xsmadd\|fmadd\ " 2 } } */
8/* { dg-final { scan-assembler-times "fmadds" 2 } } */
9/* { dg-final { scan-assembler-times "xvmsub" 2 } } */
10/* { dg-final { scan-assembler-times "xsmsub\|fmsub\ " 1 } } */
11/* { dg-final { scan-assembler-times "fmsubs" 1 } } */
12/* { dg-final { scan-assembler-times "xvnmadd" 2 } } */
13/* { dg-final { scan-assembler-times "xsnmadd\|fnmadd " 1 } } */
14/* { dg-final { scan-assembler-times "fnmadds" 1 } } */
15/* { dg-final { scan-assembler-times "xvnmsub" 2 } } */
16/* { dg-final { scan-assembler-times "xsnmsub\|fnmsub " 1 } } */
17/* { dg-final { scan-assembler-times "fnmsubs" 1 } } */
18
19/* All functions should generate an appropriate (a * b) + c instruction
20   since -mfused-madd is on by default.  */
21
22double
23builtin_fma (double b, double c, double d)
24{
25  return __builtin_fma (b, c, d);			/* xsmadd{a,m}dp */
26}
27
28double
29builtin_fms (double b, double c, double d)
30{
31  return __builtin_fma (b, c, -d);			/* xsmsub{a,b}dp */
32}
33
34double
35builtin_fnma (double b, double c, double d)
36{
37  return - __builtin_fma (b, c, d);			/* xsnmadd{a,b}dp */
38}
39
40double
41builtin_fnms (double b, double c, double d)
42{
43  return - __builtin_fma (b, c, -d);			/* xsnmsub{a,b}dp */
44}
45
46float
47builtin_fmaf (float b, float c, float d)
48{
49  return __builtin_fmaf (b, c, d);			/* fmadds */
50}
51
52float
53builtin_fmsf (float b, float c, float d)
54{
55  return __builtin_fmaf (b, c, -d);			/* fmsubs */
56}
57
58float
59builtin_fnmaf (float b, float c, float d)
60{
61  return - __builtin_fmaf (b, c, d);			/* fnmadds */
62}
63
64float
65builtin_fnmsf (float b, float c, float d)
66{
67  return - __builtin_fmaf (b, c, -d);			/* fnmsubs */
68}
69
70double
71normal_fma (double b, double c, double d)
72{
73  return (b * c) + d;					/* xsmadd{a,m}dp */
74}
75
76float
77normal_fmaf (float b, float c, float d)
78{
79  return (b * c) + d;					/* fmadds */
80}
81
82#ifndef SIZE
83#define SIZE 1024
84#endif
85
86double vda[SIZE] __attribute__((__aligned__(32)));
87double vdb[SIZE] __attribute__((__aligned__(32)));
88double vdc[SIZE] __attribute__((__aligned__(32)));
89double vdd[SIZE] __attribute__((__aligned__(32)));
90
91float vfa[SIZE] __attribute__((__aligned__(32)));
92float vfb[SIZE] __attribute__((__aligned__(32)));
93float vfc[SIZE] __attribute__((__aligned__(32)));
94float vfd[SIZE] __attribute__((__aligned__(32)));
95
96void
97vector_fma (void)
98{
99  int i;
100
101  for (i = 0; i < SIZE; i++)
102    vda[i] = __builtin_fma (vdb[i], vdc[i], vdd[i]);	/* xvmadd{a,m}dp */
103}
104
105void
106vector_fms (void)
107{
108  int i;
109
110  for (i = 0; i < SIZE; i++)
111    vda[i] = __builtin_fma (vdb[i], vdc[i], -vdd[i]);	/* xvmsub{a,m}dp */
112}
113
114void
115vector_fnma (void)
116{
117  int i;
118
119  for (i = 0; i < SIZE; i++)
120    vda[i] = - __builtin_fma (vdb[i], vdc[i], vdd[i]);	/* xvnmadd{a,m}dp */
121}
122
123void
124vector_fnms (void)
125{
126  int i;
127
128  for (i = 0; i < SIZE; i++)
129    vda[i] = - __builtin_fma (vdb[i], vdc[i], -vdd[i]);	/* xvnmsub{a,m}dp */
130}
131
132void
133vector_fmaf (void)
134{
135  int i;
136
137  for (i = 0; i < SIZE; i++)
138    vfa[i] = __builtin_fmaf (vfb[i], vfc[i], vfd[i]);	/* xvmadd{a,m}sp */
139}
140
141void
142vector_fmsf (void)
143{
144  int i;
145
146  for (i = 0; i < SIZE; i++)
147    vfa[i] = __builtin_fmaf (vfb[i], vfc[i], -vfd[i]);	/* xvmsub{a,m}sp */
148}
149
150void
151vector_fnmaf (void)
152{
153  int i;
154
155  for (i = 0; i < SIZE; i++)
156    vfa[i] = - __builtin_fmaf (vfb[i], vfc[i], vfd[i]);	/* xvnmadd{a,m}sp */
157}
158
159void
160vector_fnmsf (void)
161{
162  int i;
163
164  for (i = 0; i < SIZE; i++)
165    vfa[i] = - __builtin_fmaf (vfb[i], vfc[i], -vfd[i]); /* xvnmsub{a,m}sp */
166}
167
168void
169vnormal_fma (void)
170{
171  int i;
172
173  for (i = 0; i < SIZE; i++)
174    vda[i] = (vdb[i] * vdc[i]) + vdd[i];		/* xvmadd{a,m}dp */
175}
176
177void
178vnormal_fmaf (void)
179{
180  int i;
181
182  for (i = 0; i < SIZE; i++)
183    vfa[i] = (vfb[i] * vfc[i]) + vfd[i];		/* xvmadd{a,m}sp */
184}
185