sharpyuv_dsp.c (3418B)
1 // Copyright 2022 Google Inc. All Rights Reserved. 2 // 3 // Use of this source code is governed by a BSD-style license 4 // that can be found in the COPYING file in the root of the source 5 // tree. An additional intellectual property rights grant can be found 6 // in the file PATENTS. All contributing project authors may 7 // be found in the AUTHORS file in the root of the source tree. 8 // ----------------------------------------------------------------------------- 9 // 10 // Speed-critical functions for Sharp YUV. 11 // 12 // Author: Skal (pascal.massimino@gmail.com) 13 14 #include "sharpyuv/sharpyuv_dsp.h" 15 16 #include <assert.h> 17 #include <stdlib.h> 18 19 #include "sharpyuv/sharpyuv_cpu.h" 20 #include "src/dsp/cpu.h" 21 #include "src/webp/types.h" 22 23 //----------------------------------------------------------------------------- 24 25 #if !WEBP_NEON_OMIT_C_CODE 26 static uint16_t clip(int v, int max) { 27 return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v; 28 } 29 30 static uint64_t SharpYuvUpdateY_C(const uint16_t* ref, const uint16_t* src, 31 uint16_t* dst, int len, int bit_depth) { 32 uint64_t diff = 0; 33 int i; 34 const int max_y = (1 << bit_depth) - 1; 35 for (i = 0; i < len; ++i) { 36 const int diff_y = ref[i] - src[i]; 37 const int new_y = (int)dst[i] + diff_y; 38 dst[i] = clip(new_y, max_y); 39 diff += (uint64_t)abs(diff_y); 40 } 41 return diff; 42 } 43 44 static void SharpYuvUpdateRGB_C(const int16_t* ref, const int16_t* src, 45 int16_t* dst, int len) { 46 int i; 47 for (i = 0; i < len; ++i) { 48 const int diff_uv = ref[i] - src[i]; 49 dst[i] += diff_uv; 50 } 51 } 52 53 static void SharpYuvFilterRow_C(const int16_t* A, const int16_t* B, int len, 54 const uint16_t* best_y, uint16_t* out, 55 int bit_depth) { 56 int i; 57 const int max_y = (1 << bit_depth) - 1; 58 for (i = 0; i < len; ++i, ++A, ++B) { 59 const int v0 = (A[0] * 9 + A[1] * 3 + B[0] * 3 + B[1] + 8) >> 4; 60 const int v1 = (A[1] * 9 + A[0] * 3 + B[1] * 3 + B[0] + 8) >> 4; 61 out[2 * i + 0] = clip(best_y[2 * i + 0] + v0, max_y); 62 out[2 * i + 1] = clip(best_y[2 * i + 1] + v1, max_y); 63 } 64 } 65 #endif // !WEBP_NEON_OMIT_C_CODE 66 67 //----------------------------------------------------------------------------- 68 69 uint64_t (*SharpYuvUpdateY)(const uint16_t* src, const uint16_t* ref, 70 uint16_t* dst, int len, int bit_depth); 71 void (*SharpYuvUpdateRGB)(const int16_t* src, const int16_t* ref, int16_t* dst, 72 int len); 73 void (*SharpYuvFilterRow)(const int16_t* A, const int16_t* B, int len, 74 const uint16_t* best_y, uint16_t* out, int bit_depth); 75 76 extern VP8CPUInfo SharpYuvGetCPUInfo; 77 extern void InitSharpYuvSSE2(void); 78 extern void InitSharpYuvNEON(void); 79 80 void SharpYuvInitDsp(void) { 81 #if !WEBP_NEON_OMIT_C_CODE 82 SharpYuvUpdateY = SharpYuvUpdateY_C; 83 SharpYuvUpdateRGB = SharpYuvUpdateRGB_C; 84 SharpYuvFilterRow = SharpYuvFilterRow_C; 85 #endif 86 87 if (SharpYuvGetCPUInfo != NULL) { 88 #if defined(WEBP_HAVE_SSE2) 89 if (SharpYuvGetCPUInfo(kSSE2)) { 90 InitSharpYuvSSE2(); 91 } 92 #endif // WEBP_HAVE_SSE2 93 } 94 95 #if defined(WEBP_HAVE_NEON) 96 if (WEBP_NEON_OMIT_C_CODE || 97 (SharpYuvGetCPUInfo != NULL && SharpYuvGetCPUInfo(kNEON))) { 98 InitSharpYuvNEON(); 99 } 100 #endif // WEBP_HAVE_NEON 101 102 assert(SharpYuvUpdateY != NULL); 103 assert(SharpYuvUpdateRGB != NULL); 104 assert(SharpYuvFilterRow != NULL); 105 }