2 /* { dg-require-effective-target fma4 } */
3 /* { dg-options "-O2 -mfma4" } */
5 #include "fma4-check.h"
18 } dst, res, src1, src2, src3;
20 /* Note that in macc*,msub*,mnmacc* and mnsub* instructions, the intermdediate
21 product is not rounded, only the addition is rounded. */
27 for (i = 0; i < NUM * 8; i++)
39 for (i = 0; i < NUM * 4; i++)
50 int i, j, check_fails = 0;
51 for (i = 0; i < NUM * 8; i = i + 8)
52 for (j = 0; j < 8; j++)
54 res.f[i + j] = - (src1.f[i + j] * src2.f[i + j]) + src3.f[i + j];
55 if (dst.f[i + j] != res.f[i + j])
64 int i, j, check_fails = 0;
65 for (i = 0; i < NUM * 4; i = i + 4)
66 for (j = 0; j < 4; j++)
68 res.d[i + j] = - (src1.d[i + j] * src2.d[i + j]) + src3.d[i + j];
69 if (dst.d[i + j] != res.d[i + j])
82 for (i = 0; i < NUM; i++)
83 dst.x[i] = _mm256_nmacc_ps (src1.x[i], src2.x[i], src3.x[i]);
90 for (i = 0; i < NUM; i++)
91 dst.y[i] = _mm256_nmacc_pd (src1.y[i], src2.y[i], src3.y[i]);