-
-
Save VamuveTV/d5d0d3ee9eb8c3db6b82eba0cb7f36fc to your computer and use it in GitHub Desktop.
This file contains hidden or 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
| static inline __m128 vselect(const __m128 a, const __m128 b, const __m128 mask) | |
| { | |
| return _mm_or_ps(_mm_and_ps(mask, a), _mm_andnot_ps(mask, b)); | |
| } | |
| static inline __m128i vselect(const __m128i a, const __m128i b, const __m128i mask) | |
| { | |
| return _mm_or_si128(_mm_and_si128(mask, a), _mm_andnot_si128(mask, b)); | |
| } | |
| /* | |
| Sort 4 floats in SSE vector using sorting network and return indices of moved values | |
| */ | |
| inline __m128i v4_sort(__m128& v) | |
| { | |
| auto i = _mm_set_epi32(3, 2, 1, 0); | |
| // Simple sorting network for n=4 | |
| // First pass | |
| auto temp = _mm_shuffle_ps(v, v, _MM_SHUFFLE(1, 0, 3, 2)); | |
| auto cmp = _mm_cmplt_ps(v, temp); | |
| cmp = _mm_shuffle_ps(cmp, cmp, _MM_SHUFFLE(1, 0, 1, 0)); | |
| auto temp2 = vselect(v, temp, cmp); | |
| // Swap indices | |
| i = vselect(i, _mm_shuffle_epi32(i, _MM_SHUFFLE(1, 0, 3, 2)), _mm_castps_si128(cmp)); | |
| // Second pass | |
| temp = _mm_shuffle_ps(temp2, temp2, _MM_SHUFFLE(2, 3, 0, 1)); | |
| cmp = _mm_cmplt_ps(temp2, temp); | |
| cmp = _mm_shuffle_ps(cmp, cmp, _MM_SHUFFLE(2, 0, 2, 0)); | |
| temp2 = vselect(temp2, temp, cmp); | |
| i = vselect(i, _mm_shuffle_epi32(i, _MM_SHUFFLE(2, 3, 0, 1)), _mm_castps_si128(cmp)); | |
| // Third pass | |
| temp = _mm_shuffle_ps(temp2, temp2, _MM_SHUFFLE(3, 1, 2, 0)); | |
| cmp = _mm_cmplt_ps(temp2, temp); | |
| cmp = _mm_shuffle_ps(cmp, cmp, _MM_SHUFFLE(3, 1, 1, 0)); | |
| v = vselect(temp2, temp, cmp); | |
| return vselect(i, _mm_shuffle_epi32(i, _MM_SHUFFLE(3, 1, 2, 0)), _mm_castps_si128(cmp)); | |
| } |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment