1/* 2 * Copyright © 2014 Timothy Arceri 3 * 4 * Permission is hereby granted, free of charge, to any person obtaining a 5 * copy of this software and associated documentation files (the "Software"), 6 * to deal in the Software without restriction, including without limitation 7 * the rights to use, copy, modify, merge, publish, distribute, sublicense, 8 * and/or sell copies of the Software, and to permit persons to whom the 9 * Software is furnished to do so, subject to the following conditions: 10 * 11 * The above copyright notice and this permission notice (including the next 12 * paragraph) shall be included in all copies or substantial portions of the 13 * Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING 20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS 21 * IN THE SOFTWARE. 22 * 23 * Author: 24 * Timothy Arceri <t_arceri@yahoo.com.au> 25 * 26 */ 27 28#include "main/sse_minmax.h" 29#include "util/macros.h" 30#include <smmintrin.h> 31#include <stdint.h> 32 33void 34_mesa_uint_array_min_max(const unsigned *ui_indices, unsigned *min_index, 35 unsigned *max_index, const unsigned count) 36{ 37 unsigned max_ui = 0; 38 unsigned min_ui = ~0U; 39 unsigned i = 0; 40 unsigned aligned_count = count; 41 42 /* handle the first few values without SSE until the pointer is aligned */ 43 while (((uintptr_t)ui_indices & 15) && aligned_count) { 44 if (*ui_indices > max_ui) 45 max_ui = *ui_indices; 46 if (*ui_indices < min_ui) 47 min_ui = *ui_indices; 48 49 aligned_count--; 50 ui_indices++; 51 } 52 53 /* TODO: The actual threshold for SSE begin useful may be higher than 8. 54 * Some careful microbenchmarks and measurement are required to 55 * find the actual tipping point. 56 */ 57 if (aligned_count >= 8) { 58 alignas(16) unsigned max_arr[4]; 59 alignas(16) unsigned min_arr[4]; 60 unsigned vec_count; 61 __m128i max_ui4 = _mm_setzero_si128(); 62 __m128i min_ui4 = _mm_set1_epi32(~0U); 63 __m128i ui_indices4; 64 __m128i *ui_indices_ptr; 65 66 vec_count = aligned_count & ~0x3; 67 ui_indices_ptr = (__m128i *)ui_indices; 68 for (i = 0; i < vec_count / 4; i++) { 69 ui_indices4 = _mm_load_si128(&ui_indices_ptr[i]); 70 max_ui4 = _mm_max_epu32(ui_indices4, max_ui4); 71 min_ui4 = _mm_min_epu32(ui_indices4, min_ui4); 72 } 73 74 _mm_store_si128((__m128i *)max_arr, max_ui4); 75 _mm_store_si128((__m128i *)min_arr, min_ui4); 76 77 for (i = 0; i < 4; i++) { 78 if (max_arr[i] > max_ui) 79 max_ui = max_arr[i]; 80 if (min_arr[i] < min_ui) 81 min_ui = min_arr[i]; 82 } 83 i = vec_count; 84 } 85 86 for (; i < aligned_count; i++) { 87 if (ui_indices[i] > max_ui) 88 max_ui = ui_indices[i]; 89 if (ui_indices[i] < min_ui) 90 min_ui = ui_indices[i]; 91 } 92 93 *min_index = min_ui; 94 *max_index = max_ui; 95} 96