Conditionally blends data elements of source vector depending on bits in a mask. The corresponding Intel® AVX2 instruction is VPBLENDD or VPBLENDW .
extern __m128i _mm_blend_epi32(__m128i s1, __m128i s2, const int mask); |
extern __m256i _mm256_blend_epi16(__m256i s1, __m256i s2, const int mask); |
extern __m256i _mm256_blend_epi32(__m256i s1, __m256i s2, const int mask); |
s1 |
integer source vector used for the operation |
s2 |
integer source vector used for the operation |
mask |
8-bit immediate used for the operation |
Performs a blend operation by conditionally copying 16/32-bit [word/doubleword] elements from the s2 and s1 source vectors, depending on mask bits defined in mask. The mask bits are the least significant 8 bits in mask when the 256-bit intrinsics, _mm256_blend_epi16/_mm256_blend_epi32, are used, and 4 bits when the 128-bit intrinsic, _mm_blend_epi32, is used.
Each word/doubleword element of the destination vector is copied from the corresponding word/doubleword element in s2 if a mask bit is 1, or is copied from the corresponding word/doubleword element in s1 if a mask bit is 0.
Result of the blend operation.
Copyright © 1996-2011, Intel Corporation. All rights reserved.