Last active
January 5, 2016 17:18
-
-
Save fsaintjacques/a4cf58fb446489da9a85 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| int bitset_container_or(const bitset_container_t *bitset1, | |
| const bitset_container_t *bitset2, | |
| bitset_container_t *bitsetout) { | |
| #ifdef USEAVX | |
| - bitset_container_or_nocard(bitset1, bitset2, bitsetout); | |
| - bitsetout->cardinality = bitset_container_compute_cardinality(bitsetout); | |
| + const uint64_t *a1 = bitset1->array; | |
| + const uint64_t *a2 = bitset2->array; | |
| + uint64_t *ao = bitsetout->array; | |
| + int sum = 0; | |
| + for (int i = 0; i < BITSET_CONTAINER_SIZE_IN_WORDS / (4 * 8); i++) { | |
| + for (int j = 0; j < 8; ++j) { | |
| + __m256i A1 = _mm256_lddqu_si256((__m256i *)a1 + (i * 8) + j); | |
| + __m256i A2 = _mm256_lddqu_si256((__m256i *)a2 + (i * 8) + j); | |
| + __m256i AO = _mm256_or_si256(A1, A2); | |
| + _mm256_storeu_si256((__m256i *)ao + (i * 8) + j, AO); | |
| + } | |
| + for (int j = 0; j < 8; ++j) { | |
| + sum += _mm_popcnt_u64(ao[(i * 8) + j]); | |
| + sum += _mm_popcnt_u64(ao[(i * 8) + j + 1]); | |
| + sum += _mm_popcnt_u64(ao[(i * 8) + j + 2]); | |
| + sum += _mm_popcnt_u64(ao[(i * 8) + j + 3]); | |
| + } | |
| + } | |
| + bitsetout->cardinality = sum; |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
bitset_container_or_nocard(B1, B2, BO): 0.54 cycles per operation
bitset_container_or(B1, B2, BO): 1.00 cycles per operation
compared to the previous
bitset_container_or_nocard(B1, B2, BO): 0.55 cycles per operation
bitset_container_or(B1, B2, BO): 1.56 cycles per operation