1 /* { dg-require-effective-target fma4 } */
2 /* { dg-options "-O2 -mfma4" } */
4 #include "fma4-check.h"
17 } dst, res, src1, src2, src3;
19 /* Note that in macc*,msub*,mnmacc* and mnsub* instructions, the intermdediate
20 product is not rounded, only the addition is rounded. */
26 for (i = 0; i < NUM * 8; i++)
38 for (i = 0; i < NUM * 4; i++)
49 int i, j, check_fails = 0;
50 for (i = 0; i < NUM * 8; i = i + 8)
51 for (j = 0; j < 8; j++)
53 res.f[i + j] = - (src1.f[i + j] * src2.f[i + j]) - src3.f[i + j];
54 if (dst.f[i + j] != res.f[i + j])
63 int i, j, check_fails = 0;
64 for (i = 0; i < NUM * 4; i = i + 4)
65 for (j = 0; j < 4; j++)
67 res.d[i + j] = - (src1.d[i + j] * src2.d[i + j]) - src3.d[i + j];
68 if (dst.d[i + j] != res.d[i + j])
81 for (i = 0; i < NUM; i++)
82 dst.x[i] = _mm256_nmsub_ps (src1.x[i], src2.x[i], src3.x[i]);
84 if (check_nmsubps (&dst.x[i], &src1.f[i * 4], &src2.f[i * 4], &src3.f[i * 4]))
89 for (i = 0; i < NUM; i++)
90 dst.y[i] = _mm256_nmsub_pd (src1.y[i], src2.y[i], src3.y[i]);
92 if (check_nmsubpd (&dst.y[i], &src1.d[i * 2], &src2.d[i * 2], &src3.d[i * 2]))