| 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 | |
| 21 | //----------------------------------------------------------------------------- |
| 22 | |
| 23 | #if !WEBP_NEON_OMIT_C_CODE |
| 24 | static uint16_t clip(int v, int max) { |
| 25 | return (v < 0) ? 0 : (v > max) ? max : (uint16_t)v; |
| 26 | } |
| 27 | |
| 28 | static uint64_t SharpYuvUpdateY_C(const uint16_t* ref, const uint16_t* src, |
| 29 | uint16_t* dst, int len, int bit_depth) { |
| 30 | uint64_t diff = 0; |
| 31 | int i; |
| 32 | const int max_y = (1 << bit_depth) - 1; |
| 33 | for (i = 0; i < len; ++i) { |
| 34 | const int diff_y = ref[i] - src[i]; |
| 35 | const int new_y = (int)dst[i] + diff_y; |
| 36 | dst[i] = clip(new_y, max_y); |
| 37 | diff += (uint64_t)abs(diff_y); |
| 38 | } |
| 39 | return diff; |
| 40 | } |
| 41 | |
| 42 | static void SharpYuvUpdateRGB_C(const int16_t* ref, const int16_t* src, |
| 43 | int16_t* dst, int len) { |
| 44 | int i; |
| 45 | for (i = 0; i < len; ++i) { |
| 46 | const int diff_uv = ref[i] - src[i]; |
| 47 | dst[i] += diff_uv; |
| 48 | } |
| 49 | } |
| 50 | |
| 51 | static void SharpYuvFilterRow_C(const int16_t* A, const int16_t* B, int len, |
| 52 | const uint16_t* best_y, uint16_t* out, |
| 53 | int bit_depth) { |
| 54 | int i; |
| 55 | const int max_y = (1 << bit_depth) - 1; |
| 56 | for (i = 0; i < len; ++i, ++A, ++B) { |
| 57 | const int v0 = (A[0] * 9 + A[1] * 3 + B[0] * 3 + B[1] + 8) >> 4; |
| 58 | const int v1 = (A[1] * 9 + A[0] * 3 + B[1] * 3 + B[0] + 8) >> 4; |
| 59 | out[2 * i + 0] = clip(best_y[2 * i + 0] + v0, max_y); |
| 60 | out[2 * i + 1] = clip(best_y[2 * i + 1] + v1, max_y); |
| 61 | } |
| 62 | } |
| 63 | #endif // !WEBP_NEON_OMIT_C_CODE |
| 64 | |
| 65 | //----------------------------------------------------------------------------- |
| 66 | |
| 67 | uint64_t (*SharpYuvUpdateY)(const uint16_t* src, const uint16_t* ref, |
| 68 | uint16_t* dst, int len, int bit_depth); |
| 69 | void (*SharpYuvUpdateRGB)(const int16_t* src, const int16_t* ref, int16_t* dst, |
| 70 | int len); |
| 71 | void (*SharpYuvFilterRow)(const int16_t* A, const int16_t* B, int len, |
| 72 | const uint16_t* best_y, uint16_t* out, |
| 73 | int bit_depth); |
| 74 | |
| 75 | extern VP8CPUInfo SharpYuvGetCPUInfo; |
| 76 | extern void InitSharpYuvSSE2(void); |
| 77 | extern void InitSharpYuvNEON(void); |
| 78 | |
| 79 | void SharpYuvInitDsp(void) { |
| 80 | #if !WEBP_NEON_OMIT_C_CODE |
| 81 | SharpYuvUpdateY = SharpYuvUpdateY_C; |
| 82 | SharpYuvUpdateRGB = SharpYuvUpdateRGB_C; |
| 83 | SharpYuvFilterRow = SharpYuvFilterRow_C; |
| 84 | #endif |
| 85 | |
| 86 | if (SharpYuvGetCPUInfo != NULL) { |
| 87 | #if defined(WEBP_HAVE_SSE2) |
| 88 | if (SharpYuvGetCPUInfo(kSSE2)) { |
| 89 | InitSharpYuvSSE2(); |
| 90 | } |
| 91 | #endif // WEBP_HAVE_SSE2 |
| 92 | } |
| 93 | |
| 94 | #if defined(WEBP_HAVE_NEON) |
| 95 | if (WEBP_NEON_OMIT_C_CODE || |
| 96 | (SharpYuvGetCPUInfo != NULL && SharpYuvGetCPUInfo(kNEON))) { |
| 97 | InitSharpYuvNEON(); |
| 98 | } |
| 99 | #endif // WEBP_HAVE_NEON |
| 100 | |
| 101 | assert(SharpYuvUpdateY != NULL); |
| 102 | assert(SharpYuvUpdateRGB != NULL); |
| 103 | assert(SharpYuvFilterRow != NULL); |
| 104 | } |
| 105 | |