_mm256_madd_epi16ADD TO COMPARE ADDED TO COMPARE
Intel 64-bit (64 bits)/ AVX2
View official documentation
Location:
Arithmetic
>
Vector Multiply, Pairwise Add & Widen
Purpose:
Multiply packed signed 16-bit integers in a and b, producing intermediate signed 32-bit integers. Horizontally add adjacent pairs of intermediate 32-bit integers, and pack the results in output.
Result:
A 256-bit vector of 8 x 32-bit signed integers that hold the result of the pairwise added products.
Example:
#include <immintrin.h>
#include <stdio.h>
int main() {
__m256i a = _mm256_set_epi16(1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16);
__m256i b = _mm256_set_epi16(16, 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1);
__m256i result = _mm256_madd_epi16(a, b);
int32_t* res = (int32_t*)&result;
for(int i = 0; i < 8; i++) {
printf("%d ", res[i]);
}
return 0;
}Prototypes
Assembly Instruction:
vpmaddwd
Usage:
__m256i output =
_mm256_madd_epi16(
__m256i a, __m256i b
)
DB statistics
| SIMD Engines: | 5 |
| C Intrinsics: | 10702 |
| NEON: | 4232 |
| AVX2: | 462 |
| AVX512: | 4955 |
| SSE4.2: | 652 |
| VSX: | 401 |