1// Copyright 2010 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// YUV->RGB conversion functions
11//
12// Author: Skal (pascal.massimino@gmail.com)
13
14#include "src/dsp/yuv.h"
15
16#include <assert.h>
17#include <stdlib.h>
18
19//-----------------------------------------------------------------------------
20// Plain-C version
21
22#define ROW_FUNC(FUNC_NAME, FUNC, XSTEP) \
23static void FUNC_NAME(const uint8_t* y, \
24 const uint8_t* u, const uint8_t* v, \
25 uint8_t* dst, int len) { \
26 const uint8_t* const end = dst + (len & ~1) * (XSTEP); \
27 while (dst != end) { \
28 FUNC(y[0], u[0], v[0], dst); \
29 FUNC(y[1], u[0], v[0], dst + (XSTEP)); \
30 y += 2; \
31 ++u; \
32 ++v; \
33 dst += 2 * (XSTEP); \
34 } \
35 if (len & 1) { \
36 FUNC(y[0], u[0], v[0], dst); \
37 } \
38} \
39
40// All variants implemented.
41ROW_FUNC(YuvToRgbRow, VP8YuvToRgb, 3)
42ROW_FUNC(YuvToBgrRow, VP8YuvToBgr, 3)
43ROW_FUNC(YuvToRgbaRow, VP8YuvToRgba, 4)
44ROW_FUNC(YuvToBgraRow, VP8YuvToBgra, 4)
45ROW_FUNC(YuvToArgbRow, VP8YuvToArgb, 4)
46ROW_FUNC(YuvToRgba4444Row, VP8YuvToRgba4444, 2)
47ROW_FUNC(YuvToRgb565Row, VP8YuvToRgb565, 2)
48
49#undef ROW_FUNC
50
51// Main call for processing a plane with a WebPSamplerRowFunc function:
52void WebPSamplerProcessPlane(const uint8_t* y, int y_stride,
53 const uint8_t* u, const uint8_t* v, int uv_stride,
54 uint8_t* dst, int dst_stride,
55 int width, int height, WebPSamplerRowFunc func) {
56 int j;
57 for (j = 0; j < height; ++j) {
58 func(y, u, v, dst, width);
59 y += y_stride;
60 if (j & 1) {
61 u += uv_stride;
62 v += uv_stride;
63 }
64 dst += dst_stride;
65 }
66}
67
68//-----------------------------------------------------------------------------
69// Main call
70
71WebPSamplerRowFunc WebPSamplers[MODE_LAST];
72
73extern VP8CPUInfo VP8GetCPUInfo;
74extern void WebPInitSamplersSSE2(void);
75extern void WebPInitSamplersSSE41(void);
76extern void WebPInitSamplersMIPS32(void);
77extern void WebPInitSamplersMIPSdspR2(void);
78
79WEBP_DSP_INIT_FUNC(WebPInitSamplers) {
80 WebPSamplers[MODE_RGB] = YuvToRgbRow;
81 WebPSamplers[MODE_RGBA] = YuvToRgbaRow;
82 WebPSamplers[MODE_BGR] = YuvToBgrRow;
83 WebPSamplers[MODE_BGRA] = YuvToBgraRow;
84 WebPSamplers[MODE_ARGB] = YuvToArgbRow;
85 WebPSamplers[MODE_RGBA_4444] = YuvToRgba4444Row;
86 WebPSamplers[MODE_RGB_565] = YuvToRgb565Row;
87 WebPSamplers[MODE_rgbA] = YuvToRgbaRow;
88 WebPSamplers[MODE_bgrA] = YuvToBgraRow;
89 WebPSamplers[MODE_Argb] = YuvToArgbRow;
90 WebPSamplers[MODE_rgbA_4444] = YuvToRgba4444Row;
91
92 // If defined, use CPUInfo() to overwrite some pointers with faster versions.
93 if (VP8GetCPUInfo != NULL) {
94#if defined(WEBP_HAVE_SSE2)
95 if (VP8GetCPUInfo(kSSE2)) {
96 WebPInitSamplersSSE2();
97 }
98#endif // WEBP_HAVE_SSE2
99#if defined(WEBP_HAVE_SSE41)
100 if (VP8GetCPUInfo(kSSE4_1)) {
101 WebPInitSamplersSSE41();
102 }
103#endif // WEBP_HAVE_SSE41
104#if defined(WEBP_USE_MIPS32)
105 if (VP8GetCPUInfo(kMIPS32)) {
106 WebPInitSamplersMIPS32();
107 }
108#endif // WEBP_USE_MIPS32
109#if defined(WEBP_USE_MIPS_DSP_R2)
110 if (VP8GetCPUInfo(kMIPSdspR2)) {
111 WebPInitSamplersMIPSdspR2();
112 }
113#endif // WEBP_USE_MIPS_DSP_R2
114 }
115}
116
117//-----------------------------------------------------------------------------
118// ARGB -> YUV converters
119
120static void ConvertARGBToY_C(const uint32_t* argb, uint8_t* y, int width) {
121 int i;
122 for (i = 0; i < width; ++i) {
123 const uint32_t p = argb[i];
124 y[i] = VP8RGBToY((p >> 16) & 0xff, (p >> 8) & 0xff, (p >> 0) & 0xff,
125 YUV_HALF);
126 }
127}
128
129void WebPConvertARGBToUV_C(const uint32_t* argb, uint8_t* u, uint8_t* v,
130 int src_width, int do_store) {
131 // No rounding. Last pixel is dealt with separately.
132 const int uv_width = src_width >> 1;
133 int i;
134 for (i = 0; i < uv_width; ++i) {
135 const uint32_t v0 = argb[2 * i + 0];
136 const uint32_t v1 = argb[2 * i + 1];
137 // VP8RGBToU/V expects four accumulated pixels. Hence we need to
138 // scale r/g/b value by a factor 2. We just shift v0/v1 one bit less.
139 const int r = ((v0 >> 15) & 0x1fe) + ((v1 >> 15) & 0x1fe);
140 const int g = ((v0 >> 7) & 0x1fe) + ((v1 >> 7) & 0x1fe);
141 const int b = ((v0 << 1) & 0x1fe) + ((v1 << 1) & 0x1fe);
142 const int tmp_u = VP8RGBToU(r, g, b, YUV_HALF << 2);
143 const int tmp_v = VP8RGBToV(r, g, b, YUV_HALF << 2);
144 if (do_store) {
145 u[i] = tmp_u;
146 v[i] = tmp_v;
147 } else {
148 // Approximated average-of-four. But it's an acceptable diff.
149 u[i] = (u[i] + tmp_u + 1) >> 1;
150 v[i] = (v[i] + tmp_v + 1) >> 1;
151 }
152 }
153 if (src_width & 1) { // last pixel
154 const uint32_t v0 = argb[2 * i + 0];
155 const int r = (v0 >> 14) & 0x3fc;
156 const int g = (v0 >> 6) & 0x3fc;
157 const int b = (v0 << 2) & 0x3fc;
158 const int tmp_u = VP8RGBToU(r, g, b, YUV_HALF << 2);
159 const int tmp_v = VP8RGBToV(r, g, b, YUV_HALF << 2);
160 if (do_store) {
161 u[i] = tmp_u;
162 v[i] = tmp_v;
163 } else {
164 u[i] = (u[i] + tmp_u + 1) >> 1;
165 v[i] = (v[i] + tmp_v + 1) >> 1;
166 }
167 }
168}
169
170//-----------------------------------------------------------------------------
171
172static void ConvertRGB24ToY_C(const uint8_t* rgb, uint8_t* y, int width) {
173 int i;
174 for (i = 0; i < width; ++i, rgb += 3) {
175 y[i] = VP8RGBToY(rgb[0], rgb[1], rgb[2], YUV_HALF);
176 }
177}
178
179static void ConvertBGR24ToY_C(const uint8_t* bgr, uint8_t* y, int width) {
180 int i;
181 for (i = 0; i < width; ++i, bgr += 3) {
182 y[i] = VP8RGBToY(bgr[2], bgr[1], bgr[0], YUV_HALF);
183 }
184}
185
186void WebPConvertRGBA32ToUV_C(const uint16_t* rgb,
187 uint8_t* u, uint8_t* v, int width) {
188 int i;
189 for (i = 0; i < width; i += 1, rgb += 4) {
190 const int r = rgb[0], g = rgb[1], b = rgb[2];
191 u[i] = VP8RGBToU(r, g, b, YUV_HALF << 2);
192 v[i] = VP8RGBToV(r, g, b, YUV_HALF << 2);
193 }
194}
195
196//-----------------------------------------------------------------------------
197
198void (*WebPConvertRGB24ToY)(const uint8_t* rgb, uint8_t* y, int width);
199void (*WebPConvertBGR24ToY)(const uint8_t* bgr, uint8_t* y, int width);
200void (*WebPConvertRGBA32ToUV)(const uint16_t* rgb,
201 uint8_t* u, uint8_t* v, int width);
202
203void (*WebPConvertARGBToY)(const uint32_t* argb, uint8_t* y, int width);
204void (*WebPConvertARGBToUV)(const uint32_t* argb, uint8_t* u, uint8_t* v,
205 int src_width, int do_store);
206
207extern void WebPInitConvertARGBToYUVSSE2(void);
208extern void WebPInitConvertARGBToYUVSSE41(void);
209extern void WebPInitConvertARGBToYUVNEON(void);
210
211WEBP_DSP_INIT_FUNC(WebPInitConvertARGBToYUV) {
212 WebPConvertARGBToY = ConvertARGBToY_C;
213 WebPConvertARGBToUV = WebPConvertARGBToUV_C;
214
215 WebPConvertRGB24ToY = ConvertRGB24ToY_C;
216 WebPConvertBGR24ToY = ConvertBGR24ToY_C;
217
218 WebPConvertRGBA32ToUV = WebPConvertRGBA32ToUV_C;
219
220 if (VP8GetCPUInfo != NULL) {
221#if defined(WEBP_HAVE_SSE2)
222 if (VP8GetCPUInfo(kSSE2)) {
223 WebPInitConvertARGBToYUVSSE2();
224 }
225#endif // WEBP_HAVE_SSE2
226#if defined(WEBP_HAVE_SSE41)
227 if (VP8GetCPUInfo(kSSE4_1)) {
228 WebPInitConvertARGBToYUVSSE41();
229 }
230#endif // WEBP_HAVE_SSE41
231 }
232
233#if defined(WEBP_HAVE_NEON)
234 if (WEBP_NEON_OMIT_C_CODE ||
235 (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
236 WebPInitConvertARGBToYUVNEON();
237 }
238#endif // WEBP_HAVE_NEON
239
240 assert(WebPConvertARGBToY != NULL);
241 assert(WebPConvertARGBToUV != NULL);
242 assert(WebPConvertRGB24ToY != NULL);
243 assert(WebPConvertBGR24ToY != NULL);
244 assert(WebPConvertRGBA32ToUV != NULL);
245}
246