1 /* { dg-do compile { target { powerpc*-*-* } } } */
2 /* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */
3 /* { dg-require-effective-target powerpc_vsx_ok } */
4 /* { dg-options "-O3 -ftree-vectorize -mcpu=power7 -ffast-math -ffp-contract=off" } */
5 /* { dg-final { scan-assembler-times "xvmadd" 2 } } */
6 /* { dg-final { scan-assembler-times "xsmadd" 1 } } */
7 /* { dg-final { scan-assembler-times "fmadds" 1 } } */
8 /* { dg-final { scan-assembler-times "xvmsub" 2 } } */
9 /* { dg-final { scan-assembler-times "xsmsub" 1 } } */
10 /* { dg-final { scan-assembler-times "fmsubs" 1 } } */
11 /* { dg-final { scan-assembler-times "xvnmadd" 2 } } */
12 /* { dg-final { scan-assembler-times "xsnmadd" 1 } } */
13 /* { dg-final { scan-assembler-times "fnmadds" 1 } } */
14 /* { dg-final { scan-assembler-times "xvnmsub" 2 } } */
15 /* { dg-final { scan-assembler-times "xsnmsub" 1 } } */
16 /* { dg-final { scan-assembler-times "fnmsubs" 1 } } */
18 /* Only the functions calling the bulitin should generate an appropriate (a *
19 b) + c instruction. */
22 builtin_fma (double b, double c, double d)
24 return __builtin_fma (b, c, d); /* xsmadd{a,m}dp */
28 builtin_fms (double b, double c, double d)
30 return __builtin_fma (b, c, -d); /* xsmsub{a,b}dp */
34 builtin_fnma (double b, double c, double d)
36 return - __builtin_fma (b, c, d); /* xsnmadd{a,b}dp */
40 builtin_fnms (double b, double c, double d)
42 return - __builtin_fma (b, c, -d); /* xsnmsub{a,b}dp */
46 builtin_fmaf (float b, float c, float d)
48 return __builtin_fmaf (b, c, d); /* fmadds */
52 builtin_fmsf (float b, float c, float d)
54 return __builtin_fmaf (b, c, -d); /* fmsubs */
58 builtin_fnmaf (float b, float c, float d)
60 return - __builtin_fmaf (b, c, d); /* fnmadds */
64 builtin_fnmsf (float b, float c, float d)
66 return - __builtin_fmaf (b, c, -d); /* fnmsubs */
70 normal_fma (double b, double c, double d)
72 return (b * c) + d; /* fmul/fadd */
76 normal_fmaf (float b, float c, float d)
78 return (b * c) + d; /* fmuls/fadds */
85 double vda[SIZE] __attribute__((__aligned__(32)));
86 double vdb[SIZE] __attribute__((__aligned__(32)));
87 double vdc[SIZE] __attribute__((__aligned__(32)));
88 double vdd[SIZE] __attribute__((__aligned__(32)));
90 float vfa[SIZE] __attribute__((__aligned__(32)));
91 float vfb[SIZE] __attribute__((__aligned__(32)));
92 float vfc[SIZE] __attribute__((__aligned__(32)));
93 float vfd[SIZE] __attribute__((__aligned__(32)));
100 for (i = 0; i < SIZE; i++)
101 vda[i] = __builtin_fma (vdb[i], vdc[i], vdd[i]); /* xvmadd{a,m}dp */
109 for (i = 0; i < SIZE; i++)
110 vda[i] = __builtin_fma (vdb[i], vdc[i], -vdd[i]); /* xvmsub{a,m}dp */
118 for (i = 0; i < SIZE; i++)
119 vda[i] = - __builtin_fma (vdb[i], vdc[i], vdd[i]); /* xvnmadd{a,m}dp */
127 for (i = 0; i < SIZE; i++)
128 vda[i] = - __builtin_fma (vdb[i], vdc[i], -vdd[i]); /* xvnmsub{a,m}dp */
136 for (i = 0; i < SIZE; i++)
137 vfa[i] = __builtin_fmaf (vfb[i], vfc[i], vfd[i]); /* xvmadd{a,m}sp */
145 for (i = 0; i < SIZE; i++)
146 vfa[i] = __builtin_fmaf (vfb[i], vfc[i], -vfd[i]); /* xvmsub{a,m}sp */
154 for (i = 0; i < SIZE; i++)
155 vfa[i] = - __builtin_fmaf (vfb[i], vfc[i], vfd[i]); /* xvnmadd{a,m}sp */
163 for (i = 0; i < SIZE; i++)
164 vfa[i] = - __builtin_fmaf (vfb[i], vfc[i], -vfd[i]); /* xvnmsub{a,m}sp */
172 for (i = 0; i < SIZE; i++)
173 vda[i] = (vdb[i] * vdc[i]) + vdd[i]; /* xvmadd{a,m}dp */
181 for (i = 0; i < SIZE; i++)
182 vfa[i] = (vfb[i] * vfc[i]) + vfd[i]; /* xvmadd{a,m}sp */