1 // Copyright 2011 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 to RGB upsampling functions.
11 //
12 // Author: somnath@google.com (Somnath Banerjee)
13
14 #include "src/dsp/dsp.h"
15 #include "src/dsp/yuv.h"
16
17 #include <assert.h>
18
19 //------------------------------------------------------------------------------
20 // Fancy upsampler
21
22 #ifdef FANCY_UPSAMPLING
23
24 // Fancy upsampling functions to convert YUV to RGB
25 WebPUpsampleLinePairFunc WebPUpsamplers[MODE_LAST];
26
27 // Given samples laid out in a square as:
28 // [a b]
29 // [c d]
30 // we interpolate u/v as:
31 // ([9*a + 3*b + 3*c + d 3*a + 9*b + 3*c + d] + [8 8]) / 16
32 // ([3*a + b + 9*c + 3*d a + 3*b + 3*c + 9*d] [8 8]) / 16
33
34 // We process u and v together stashed into 32bit (16bit each).
35 #define LOAD_UV(u, v) ((u) | ((v) << 16))
36
37 #define UPSAMPLE_FUNC(FUNC_NAME, FUNC, XSTEP) \
38 static void FUNC_NAME(const uint8_t* WEBP_RESTRICT top_y, \
39 const uint8_t* WEBP_RESTRICT bottom_y, \
40 const uint8_t* WEBP_RESTRICT top_u, \
41 const uint8_t* WEBP_RESTRICT top_v, \
42 const uint8_t* WEBP_RESTRICT cur_u, \
43 const uint8_t* WEBP_RESTRICT cur_v, \
44 uint8_t* WEBP_RESTRICT top_dst, \
45 uint8_t* WEBP_RESTRICT bottom_dst, int len) { \
46 int x; \
47 const int last_pixel_pair = (len - 1) >> 1; \
48 uint32_t tl_uv = LOAD_UV(top_u[0], top_v[0]); /* top-left sample */ \
49 uint32_t l_uv = LOAD_UV(cur_u[0], cur_v[0]); /* left-sample */ \
50 assert(top_y != NULL); \
51 { \
52 const uint32_t uv0 = (3 * tl_uv + l_uv + 0x00020002u) >> 2; \
53 FUNC(top_y[0], uv0 & 0xff, (uv0 >> 16), top_dst); \
54 } \
55 if (bottom_y != NULL) { \
56 const uint32_t uv0 = (3 * l_uv + tl_uv + 0x00020002u) >> 2; \
57 FUNC(bottom_y[0], uv0 & 0xff, (uv0 >> 16), bottom_dst); \
58 } \
59 for (x = 1; x <= last_pixel_pair; ++x) { \
60 const uint32_t t_uv = LOAD_UV(top_u[x], top_v[x]); /* top sample */ \
61 const uint32_t uv = LOAD_UV(cur_u[x], cur_v[x]); /* sample */ \
62 /* precompute invariant values associated with first and second diagonals*/\
63 const uint32_t avg = tl_uv + t_uv + l_uv + uv + 0x00080008u; \
64 const uint32_t diag_12 = (avg + 2 * (t_uv + l_uv)) >> 3; \
65 const uint32_t diag_03 = (avg + 2 * (tl_uv + uv)) >> 3; \
66 { \
67 const uint32_t uv0 = (diag_12 + tl_uv) >> 1; \
68 const uint32_t uv1 = (diag_03 + t_uv) >> 1; \
69 FUNC(top_y[2 * x - 1], uv0 & 0xff, (uv0 >> 16), \
70 top_dst + (2 * x - 1) * (XSTEP)); \
71 FUNC(top_y[2 * x - 0], uv1 & 0xff, (uv1 >> 16), \
72 top_dst + (2 * x - 0) * (XSTEP)); \
73 } \
74 if (bottom_y != NULL) { \
75 const uint32_t uv0 = (diag_03 + l_uv) >> 1; \
76 const uint32_t uv1 = (diag_12 + uv) >> 1; \
77 FUNC(bottom_y[2 * x - 1], uv0 & 0xff, (uv0 >> 16), \
78 bottom_dst + (2 * x - 1) * (XSTEP)); \
79 FUNC(bottom_y[2 * x + 0], uv1 & 0xff, (uv1 >> 16), \
80 bottom_dst + (2 * x + 0) * (XSTEP)); \
81 } \
82 tl_uv = t_uv; \
83 l_uv = uv; \
84 } \
85 if (!(len & 1)) { \
86 { \
87 const uint32_t uv0 = (3 * tl_uv + l_uv + 0x00020002u) >> 2; \
88 FUNC(top_y[len - 1], uv0 & 0xff, (uv0 >> 16), \
89 top_dst + (len - 1) * (XSTEP)); \
90 } \
91 if (bottom_y != NULL) { \
92 const uint32_t uv0 = (3 * l_uv + tl_uv + 0x00020002u) >> 2; \
93 FUNC(bottom_y[len - 1], uv0 & 0xff, (uv0 >> 16), \
94 bottom_dst + (len - 1) * (XSTEP)); \
95 } \
96 } \
97 }
98
99 // All variants implemented.
100 #if !WEBP_NEON_OMIT_C_CODE
101 UPSAMPLE_FUNC(UpsampleRgbaLinePair_C, VP8YuvToRgba, 4)
102 UPSAMPLE_FUNC(UpsampleBgraLinePair_C, VP8YuvToBgra, 4)
103 #if !defined(WEBP_REDUCE_CSP)
104 UPSAMPLE_FUNC(UpsampleArgbLinePair_C, VP8YuvToArgb, 4)
105 UPSAMPLE_FUNC(UpsampleRgbLinePair_C, VP8YuvToRgb, 3)
106 UPSAMPLE_FUNC(UpsampleBgrLinePair_C, VP8YuvToBgr, 3)
107 UPSAMPLE_FUNC(UpsampleRgba4444LinePair_C, VP8YuvToRgba4444, 2)
108 UPSAMPLE_FUNC(UpsampleRgb565LinePair_C, VP8YuvToRgb565, 2)
109 #else
110 static void EmptyUpsampleFunc(const uint8_t* top_y, const uint8_t* bottom_y,
111 const uint8_t* top_u, const uint8_t* top_v,
112 const uint8_t* cur_u, const uint8_t* cur_v,
113 uint8_t* top_dst, uint8_t* bottom_dst, int len) {
114 (void)top_y;
115 (void)bottom_y;
116 (void)top_u;
117 (void)top_v;
118 (void)cur_u;
119 (void)cur_v;
120 (void)top_dst;
121 (void)bottom_dst;
122 (void)len;
123 assert(0); // COLORSPACE SUPPORT NOT COMPILED
124 }
125 #define UpsampleArgbLinePair_C EmptyUpsampleFunc
126 #define UpsampleRgbLinePair_C EmptyUpsampleFunc
127 #define UpsampleBgrLinePair_C EmptyUpsampleFunc
128 #define UpsampleRgba4444LinePair_C EmptyUpsampleFunc
129 #define UpsampleRgb565LinePair_C EmptyUpsampleFunc
130 #endif // WEBP_REDUCE_CSP
131
132 #endif
133
134 #undef LOAD_UV
135 #undef UPSAMPLE_FUNC
136
137 #endif // FANCY_UPSAMPLING
138
139 //------------------------------------------------------------------------------
140
141 #if !defined(FANCY_UPSAMPLING)
142 #define DUAL_SAMPLE_FUNC(FUNC_NAME, FUNC) \
143 static void FUNC_NAME(const uint8_t* WEBP_RESTRICT top_y, \
144 const uint8_t* WEBP_RESTRICT bot_y, \
145 const uint8_t* WEBP_RESTRICT top_u, \
146 const uint8_t* WEBP_RESTRICT top_v, \
147 const uint8_t* WEBP_RESTRICT bot_u, \
148 const uint8_t* WEBP_RESTRICT bot_v, \
149 uint8_t* WEBP_RESTRICT top_dst, \
150 uint8_t* WEBP_RESTRICT bot_dst, int len) { \
151 const int half_len = len >> 1; \
152 int x; \
153 assert(top_dst != NULL); \
154 { \
155 for (x = 0; x < half_len; ++x) { \
156 FUNC(top_y[2 * x + 0], top_u[x], top_v[x], top_dst + 8 * x + 0); \
157 FUNC(top_y[2 * x + 1], top_u[x], top_v[x], top_dst + 8 * x + 4); \
158 } \
159 if (len & 1) FUNC(top_y[2 * x + 0], top_u[x], top_v[x], top_dst + 8 * x); \
160 } \
161 if (bot_dst != NULL) { \
162 for (x = 0; x < half_len; ++x) { \
163 FUNC(bot_y[2 * x + 0], bot_u[x], bot_v[x], bot_dst + 8 * x + 0); \
164 FUNC(bot_y[2 * x + 1], bot_u[x], bot_v[x], bot_dst + 8 * x + 4); \
165 } \
166 if (len & 1) FUNC(bot_y[2 * x + 0], bot_u[x], bot_v[x], bot_dst + 8 * x); \
167 } \
168 }
169
DUAL_SAMPLE_FUNC(DualLineSamplerBGRA,VP8YuvToBgra)170 DUAL_SAMPLE_FUNC(DualLineSamplerBGRA, VP8YuvToBgra)
171 DUAL_SAMPLE_FUNC(DualLineSamplerARGB, VP8YuvToArgb)
172 #undef DUAL_SAMPLE_FUNC
173
174 #endif // !FANCY_UPSAMPLING
175
176 WebPUpsampleLinePairFunc WebPGetLinePairConverter(int alpha_is_last) {
177 WebPInitUpsamplers();
178 #ifdef FANCY_UPSAMPLING
179 return WebPUpsamplers[alpha_is_last ? MODE_BGRA : MODE_ARGB];
180 #else
181 return (alpha_is_last ? DualLineSamplerBGRA : DualLineSamplerARGB);
182 #endif
183 }
184
185 //------------------------------------------------------------------------------
186 // YUV444 converter
187
188 #define YUV444_FUNC(FUNC_NAME, FUNC, XSTEP) \
189 extern void FUNC_NAME(const uint8_t* WEBP_RESTRICT y, \
190 const uint8_t* WEBP_RESTRICT u, \
191 const uint8_t* WEBP_RESTRICT v, \
192 uint8_t* WEBP_RESTRICT dst, int len); \
193 void FUNC_NAME(const uint8_t* WEBP_RESTRICT y, \
194 const uint8_t* WEBP_RESTRICT u, \
195 const uint8_t* WEBP_RESTRICT v, \
196 uint8_t* WEBP_RESTRICT dst, int len) { \
197 int i; \
198 for (i = 0; i < len; ++i) FUNC(y[i], u[i], v[i], &dst[i * (XSTEP)]); \
199 }
200
201 YUV444_FUNC(WebPYuv444ToRgba_C, VP8YuvToRgba, 4)
202 YUV444_FUNC(WebPYuv444ToBgra_C, VP8YuvToBgra, 4)
203 #if !defined(WEBP_REDUCE_CSP)
204 YUV444_FUNC(WebPYuv444ToRgb_C, VP8YuvToRgb, 3)
205 YUV444_FUNC(WebPYuv444ToBgr_C, VP8YuvToBgr, 3)
206 YUV444_FUNC(WebPYuv444ToArgb_C, VP8YuvToArgb, 4)
207 YUV444_FUNC(WebPYuv444ToRgba4444_C, VP8YuvToRgba4444, 2)
208 YUV444_FUNC(WebPYuv444ToRgb565_C, VP8YuvToRgb565, 2)
209 #else
210 static void EmptyYuv444Func(const uint8_t* y,
211 const uint8_t* u, const uint8_t* v,
212 uint8_t* dst, int len) {
213 (void)y;
214 (void)u;
215 (void)v;
216 (void)dst;
217 (void)len;
218 }
219 #define WebPYuv444ToRgb_C EmptyYuv444Func
220 #define WebPYuv444ToBgr_C EmptyYuv444Func
221 #define WebPYuv444ToArgb_C EmptyYuv444Func
222 #define WebPYuv444ToRgba4444_C EmptyYuv444Func
223 #define WebPYuv444ToRgb565_C EmptyYuv444Func
224 #endif // WEBP_REDUCE_CSP
225
226 #undef YUV444_FUNC
227
228 WebPYUV444Converter WebPYUV444Converters[MODE_LAST];
229
230 extern VP8CPUInfo VP8GetCPUInfo;
231 extern void WebPInitYUV444ConvertersMIPSdspR2(void);
232 extern void WebPInitYUV444ConvertersSSE2(void);
233 extern void WebPInitYUV444ConvertersSSE41(void);
234
WEBP_DSP_INIT_FUNC(WebPInitYUV444Converters)235 WEBP_DSP_INIT_FUNC(WebPInitYUV444Converters) {
236 WebPYUV444Converters[MODE_RGBA] = WebPYuv444ToRgba_C;
237 WebPYUV444Converters[MODE_BGRA] = WebPYuv444ToBgra_C;
238 WebPYUV444Converters[MODE_RGB] = WebPYuv444ToRgb_C;
239 WebPYUV444Converters[MODE_BGR] = WebPYuv444ToBgr_C;
240 WebPYUV444Converters[MODE_ARGB] = WebPYuv444ToArgb_C;
241 WebPYUV444Converters[MODE_RGBA_4444] = WebPYuv444ToRgba4444_C;
242 WebPYUV444Converters[MODE_RGB_565] = WebPYuv444ToRgb565_C;
243 WebPYUV444Converters[MODE_rgbA] = WebPYuv444ToRgba_C;
244 WebPYUV444Converters[MODE_bgrA] = WebPYuv444ToBgra_C;
245 WebPYUV444Converters[MODE_Argb] = WebPYuv444ToArgb_C;
246 WebPYUV444Converters[MODE_rgbA_4444] = WebPYuv444ToRgba4444_C;
247
248 if (VP8GetCPUInfo != NULL) {
249 #if defined(WEBP_HAVE_SSE2)
250 if (VP8GetCPUInfo(kSSE2)) {
251 WebPInitYUV444ConvertersSSE2();
252 }
253 #endif
254 #if defined(WEBP_HAVE_SSE41)
255 if (VP8GetCPUInfo(kSSE4_1)) {
256 WebPInitYUV444ConvertersSSE41();
257 }
258 #endif
259 #if defined(WEBP_USE_MIPS_DSP_R2)
260 if (VP8GetCPUInfo(kMIPSdspR2)) {
261 WebPInitYUV444ConvertersMIPSdspR2();
262 }
263 #endif
264 }
265 }
266
267 //------------------------------------------------------------------------------
268 // Main calls
269
270 extern void WebPInitUpsamplersSSE2(void);
271 extern void WebPInitUpsamplersSSE41(void);
272 extern void WebPInitUpsamplersNEON(void);
273 extern void WebPInitUpsamplersMIPSdspR2(void);
274 extern void WebPInitUpsamplersMSA(void);
275
WEBP_DSP_INIT_FUNC(WebPInitUpsamplers)276 WEBP_DSP_INIT_FUNC(WebPInitUpsamplers) {
277 #ifdef FANCY_UPSAMPLING
278 #if !WEBP_NEON_OMIT_C_CODE
279 WebPUpsamplers[MODE_RGBA] = UpsampleRgbaLinePair_C;
280 WebPUpsamplers[MODE_BGRA] = UpsampleBgraLinePair_C;
281 WebPUpsamplers[MODE_rgbA] = UpsampleRgbaLinePair_C;
282 WebPUpsamplers[MODE_bgrA] = UpsampleBgraLinePair_C;
283 WebPUpsamplers[MODE_RGB] = UpsampleRgbLinePair_C;
284 WebPUpsamplers[MODE_BGR] = UpsampleBgrLinePair_C;
285 WebPUpsamplers[MODE_ARGB] = UpsampleArgbLinePair_C;
286 WebPUpsamplers[MODE_RGBA_4444] = UpsampleRgba4444LinePair_C;
287 WebPUpsamplers[MODE_RGB_565] = UpsampleRgb565LinePair_C;
288 WebPUpsamplers[MODE_Argb] = UpsampleArgbLinePair_C;
289 WebPUpsamplers[MODE_rgbA_4444] = UpsampleRgba4444LinePair_C;
290 #endif
291
292 // If defined, use CPUInfo() to overwrite some pointers with faster versions.
293 if (VP8GetCPUInfo != NULL) {
294 #if defined(WEBP_HAVE_SSE2)
295 if (VP8GetCPUInfo(kSSE2)) {
296 WebPInitUpsamplersSSE2();
297 }
298 #endif
299 #if defined(WEBP_HAVE_SSE41)
300 if (VP8GetCPUInfo(kSSE4_1)) {
301 WebPInitUpsamplersSSE41();
302 }
303 #endif
304 #if defined(WEBP_USE_MIPS_DSP_R2)
305 if (VP8GetCPUInfo(kMIPSdspR2)) {
306 WebPInitUpsamplersMIPSdspR2();
307 }
308 #endif
309 #if defined(WEBP_USE_MSA)
310 if (VP8GetCPUInfo(kMSA)) {
311 WebPInitUpsamplersMSA();
312 }
313 #endif
314 }
315
316 #if defined(WEBP_HAVE_NEON)
317 if (WEBP_NEON_OMIT_C_CODE ||
318 (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kNEON))) {
319 WebPInitUpsamplersNEON();
320 }
321 #endif
322
323 assert(WebPUpsamplers[MODE_RGBA] != NULL);
324 assert(WebPUpsamplers[MODE_BGRA] != NULL);
325 assert(WebPUpsamplers[MODE_rgbA] != NULL);
326 assert(WebPUpsamplers[MODE_bgrA] != NULL);
327 #if !defined(WEBP_REDUCE_CSP) || !WEBP_NEON_OMIT_C_CODE
328 assert(WebPUpsamplers[MODE_RGB] != NULL);
329 assert(WebPUpsamplers[MODE_BGR] != NULL);
330 assert(WebPUpsamplers[MODE_ARGB] != NULL);
331 assert(WebPUpsamplers[MODE_RGBA_4444] != NULL);
332 assert(WebPUpsamplers[MODE_RGB_565] != NULL);
333 assert(WebPUpsamplers[MODE_Argb] != NULL);
334 assert(WebPUpsamplers[MODE_rgbA_4444] != NULL);
335 #endif
336
337 #endif // FANCY_UPSAMPLING
338 }
339
340 //------------------------------------------------------------------------------
341