#endif
#include "avx10-helper.h"
#include <limits.h>
+#include <string.h>
#define SIZE (AVX512F_LEN / 16)
#include "avx512f-mask-type.h"
UNION_TYPE (AVX512F_LEN, h) s;
UNION_TYPE (AVX512F_LEN, i_w) res1, res2, res3;
MASK_TYPE mask = MASK_VALUE;
- short res_ref[SIZE] = { 0 };
+ short res_ref[SIZE] = { 0 }, res_ref2[SIZE] = { 0 };
int i, sign = 1;
for (i = 0; i < SIZE; i++)
res3.x = INTRINSIC (_maskz_ipcvtts_ph_epu8) (mask, s.x);
CALC (s.a, res_ref);
-
-#if AVX512F_LEN == 512
- res1.x = INTRINSIC (_ipcvtts_roundph_epu8) (s.x, 8);
- res2.x = INTRINSIC (_mask_ipcvtts_roundph_epu8) (res2.x, mask, s.x, 8);
- res3.x = INTRINSIC (_maskz_ipcvtts_roundph_epu8) (mask, s.x, 8);
+ memcpy(res_ref2, res_ref, sizeof(res_ref));
if (UNION_CHECK (AVX512F_LEN, i_w) (res1, res_ref))
abort ();
MASK_ZERO (i_w) (res_ref, mask, SIZE);
if (UNION_CHECK (AVX512F_LEN, i_w) (res3, res_ref))
abort ();
+
+#if AVX512F_LEN == 512
+ for (i = 0; i < SIZE; i++)
+ res2.a[i] = DEFAULT_VALUE;
+
+ res1.x = INTRINSIC (_ipcvtts_roundph_epu8) (s.x, 8);
+ res2.x = INTRINSIC (_mask_ipcvtts_roundph_epu8) (res2.x, mask, s.x, 8);
+ res3.x = INTRINSIC (_maskz_ipcvtts_roundph_epu8) (mask, s.x, 8);
+
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res1, res_ref2))
+ abort ();
+
+ MASK_MERGE (i_w) (res_ref2, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res2, res_ref2))
+ abort ();
+
+ MASK_ZERO (i_w) (res_ref2, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, i_w) (res3, res_ref2))
+ abort ();
#endif
}