2 * Copyright 2013 The LibYuv Project Authors. All rights reserved.
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
11 #ifndef INCLUDE_LIBYUV_SCALE_ROW_H_ // NOLINT
12 #define INCLUDE_LIBYUV_SCALE_ROW_H_
14 #include "libyuv/basic_types.h"
21 #if defined(__pnacl__) || defined(__CLR_VER) || defined(COVERAGE_ENABLED) || \
22 defined(TARGET_IPHONE_SIMULATOR)
23 #define LIBYUV_DISABLE_X86
26 // The following are available on all x86 platforms:
27 #if !defined(LIBYUV_DISABLE_X86) && \
28 (defined(_M_IX86) || defined(__x86_64__) || defined(__i386__))
29 #define HAS_SCALEROWDOWN2_SSE2
30 #define HAS_SCALEROWDOWN4_SSE2
31 #define HAS_SCALEROWDOWN34_SSSE3
32 #define HAS_SCALEROWDOWN38_SSSE3
33 #define HAS_SCALEADDROWS_SSE2
34 #define HAS_SCALEFILTERCOLS_SSSE3
35 #define HAS_SCALECOLSUP2_SSE2
36 #define HAS_SCALEARGBROWDOWN2_SSE2
37 #define HAS_SCALEARGBROWDOWNEVEN_SSE2
38 #define HAS_SCALEARGBCOLS_SSE2
39 #define HAS_SCALEARGBFILTERCOLS_SSSE3
40 #define HAS_SCALEARGBCOLSUP2_SSE2
41 #define HAS_FIXEDDIV_X86
42 #define HAS_FIXEDDIV1_X86
45 // The following are available on Neon platforms:
46 #if !defined(LIBYUV_DISABLE_NEON) && !defined(__native_client__) && \
47 (defined(__ARM_NEON__) || defined(LIBYUV_NEON))
48 #define HAS_SCALEROWDOWN2_NEON
49 #define HAS_SCALEROWDOWN4_NEON
50 #define HAS_SCALEROWDOWN34_NEON
51 #define HAS_SCALEROWDOWN38_NEON
52 #define HAS_SCALEARGBROWDOWNEVEN_NEON
53 #define HAS_SCALEARGBROWDOWN2_NEON
54 #elif !defined(LIBYUV_DISABLE_NEON) && !defined(__native_client__) && \
55 (defined(__aarch64__) || defined(LIBYUV_NEON))
56 /* #define HAS_SCALEROWDOWN2_NEON */
57 /* #define HAS_SCALEROWDOWN4_NEON */
58 /* #define HAS_SCALEROWDOWN34_NEON */
59 /* #define HAS_SCALEROWDOWN38_NEON */
60 /* #define HAS_SCALEARGBROWDOWNEVEN_NEON */
61 /* #define HAS_SCALEARGBROWDOWN2_NEON */
64 // The following are available on Mips platforms:
65 #if !defined(LIBYUV_DISABLE_MIPS) && !defined(__native_client__) && \
66 defined(__mips__) && defined(__mips_dsp) && (__mips_dsp_rev >= 2)
67 #define HAS_SCALEROWDOWN2_MIPS_DSPR2
68 #define HAS_SCALEROWDOWN4_MIPS_DSPR2
69 #define HAS_SCALEROWDOWN34_MIPS_DSPR2
70 #define HAS_SCALEROWDOWN38_MIPS_DSPR2
73 // Scale ARGB vertically with bilinear interpolation.
74 void ScalePlaneVertical(int src_height,
75 int dst_width, int dst_height,
76 int src_stride, int dst_stride,
77 const uint8* src_argb, uint8* dst_argb,
79 int bpp, enum FilterMode filtering);
81 void ScalePlaneVertical_16(int src_height,
82 int dst_width, int dst_height,
83 int src_stride, int dst_stride,
84 const uint16* src_argb, uint16* dst_argb,
86 int wpp, enum FilterMode filtering);
88 // Simplify the filtering based on scale factors.
89 enum FilterMode ScaleFilterReduce(int src_width, int src_height,
90 int dst_width, int dst_height,
91 enum FilterMode filtering);
93 // Divide num by div and return as 16.16 fixed point result.
94 int FixedDiv_C(int num, int div);
95 int FixedDiv_X86(int num, int div);
96 // Divide num - 1 by div - 1 and return as 16.16 fixed point result.
97 int FixedDiv1_C(int num, int div);
98 int FixedDiv1_X86(int num, int div);
99 #ifdef HAS_FIXEDDIV_X86
100 #define FixedDiv FixedDiv_X86
101 #define FixedDiv1 FixedDiv1_X86
103 #define FixedDiv FixedDiv_C
104 #define FixedDiv1 FixedDiv1_C
107 // Compute slope values for stepping.
108 void ScaleSlope(int src_width, int src_height,
109 int dst_width, int dst_height,
110 enum FilterMode filtering,
111 int* x, int* y, int* dx, int* dy);
113 void ScaleRowDown2_C(const uint8* src_ptr, ptrdiff_t src_stride,
114 uint8* dst, int dst_width);
115 void ScaleRowDown2_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
116 uint16* dst, int dst_width);
117 void ScaleRowDown2Linear_C(const uint8* src_ptr, ptrdiff_t src_stride,
118 uint8* dst, int dst_width);
119 void ScaleRowDown2Linear_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
120 uint16* dst, int dst_width);
121 void ScaleRowDown2Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
122 uint8* dst, int dst_width);
123 void ScaleRowDown2Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
124 uint16* dst, int dst_width);
125 void ScaleRowDown4_C(const uint8* src_ptr, ptrdiff_t src_stride,
126 uint8* dst, int dst_width);
127 void ScaleRowDown4_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
128 uint16* dst, int dst_width);
129 void ScaleRowDown4Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
130 uint8* dst, int dst_width);
131 void ScaleRowDown4Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
132 uint16* dst, int dst_width);
133 void ScaleRowDown34_C(const uint8* src_ptr, ptrdiff_t src_stride,
134 uint8* dst, int dst_width);
135 void ScaleRowDown34_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
136 uint16* dst, int dst_width);
137 void ScaleRowDown34_0_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
138 uint8* d, int dst_width);
139 void ScaleRowDown34_0_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
140 uint16* d, int dst_width);
141 void ScaleRowDown34_1_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
142 uint8* d, int dst_width);
143 void ScaleRowDown34_1_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
144 uint16* d, int dst_width);
145 void ScaleCols_C(uint8* dst_ptr, const uint8* src_ptr,
146 int dst_width, int x, int dx);
147 void ScaleCols_16_C(uint16* dst_ptr, const uint16* src_ptr,
148 int dst_width, int x, int dx);
149 void ScaleColsUp2_C(uint8* dst_ptr, const uint8* src_ptr,
150 int dst_width, int, int);
151 void ScaleColsUp2_16_C(uint16* dst_ptr, const uint16* src_ptr,
152 int dst_width, int, int);
153 void ScaleFilterCols_C(uint8* dst_ptr, const uint8* src_ptr,
154 int dst_width, int x, int dx);
155 void ScaleFilterCols_16_C(uint16* dst_ptr, const uint16* src_ptr,
156 int dst_width, int x, int dx);
157 void ScaleFilterCols64_C(uint8* dst_ptr, const uint8* src_ptr,
158 int dst_width, int x, int dx);
159 void ScaleFilterCols64_16_C(uint16* dst_ptr, const uint16* src_ptr,
160 int dst_width, int x, int dx);
161 void ScaleRowDown38_C(const uint8* src_ptr, ptrdiff_t src_stride,
162 uint8* dst, int dst_width);
163 void ScaleRowDown38_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
164 uint16* dst, int dst_width);
165 void ScaleRowDown38_3_Box_C(const uint8* src_ptr,
166 ptrdiff_t src_stride,
167 uint8* dst_ptr, int dst_width);
168 void ScaleRowDown38_3_Box_16_C(const uint16* src_ptr,
169 ptrdiff_t src_stride,
170 uint16* dst_ptr, int dst_width);
171 void ScaleRowDown38_2_Box_C(const uint8* src_ptr, ptrdiff_t src_stride,
172 uint8* dst_ptr, int dst_width);
173 void ScaleRowDown38_2_Box_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
174 uint16* dst_ptr, int dst_width);
175 void ScaleAddRows_C(const uint8* src_ptr, ptrdiff_t src_stride,
176 uint16* dst_ptr, int src_width, int src_height);
177 void ScaleAddRows_16_C(const uint16* src_ptr, ptrdiff_t src_stride,
178 uint32* dst_ptr, int src_width, int src_height);
179 void ScaleARGBRowDown2_C(const uint8* src_argb,
180 ptrdiff_t src_stride,
181 uint8* dst_argb, int dst_width);
182 void ScaleARGBRowDown2Linear_C(const uint8* src_argb,
183 ptrdiff_t src_stride,
184 uint8* dst_argb, int dst_width);
185 void ScaleARGBRowDown2Box_C(const uint8* src_argb, ptrdiff_t src_stride,
186 uint8* dst_argb, int dst_width);
187 void ScaleARGBRowDownEven_C(const uint8* src_argb, ptrdiff_t src_stride,
189 uint8* dst_argb, int dst_width);
190 void ScaleARGBRowDownEvenBox_C(const uint8* src_argb,
191 ptrdiff_t src_stride,
193 uint8* dst_argb, int dst_width);
194 void ScaleARGBCols_C(uint8* dst_argb, const uint8* src_argb,
195 int dst_width, int x, int dx);
196 void ScaleARGBCols64_C(uint8* dst_argb, const uint8* src_argb,
197 int dst_width, int x, int dx);
198 void ScaleARGBColsUp2_C(uint8* dst_argb, const uint8* src_argb,
199 int dst_width, int, int);
200 void ScaleARGBFilterCols_C(uint8* dst_argb, const uint8* src_argb,
201 int dst_width, int x, int dx);
202 void ScaleARGBFilterCols64_C(uint8* dst_argb, const uint8* src_argb,
203 int dst_width, int x, int dx);
205 void ScaleRowDown2_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
206 uint8* dst_ptr, int dst_width);
207 void ScaleRowDown2Linear_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
208 uint8* dst_ptr, int dst_width);
209 void ScaleRowDown2Box_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
210 uint8* dst_ptr, int dst_width);
211 void ScaleRowDown2_Unaligned_SSE2(const uint8* src_ptr,
212 ptrdiff_t src_stride,
213 uint8* dst_ptr, int dst_width);
214 void ScaleRowDown2Linear_Unaligned_SSE2(const uint8* src_ptr,
215 ptrdiff_t src_stride,
216 uint8* dst_ptr, int dst_width);
217 void ScaleRowDown2Box_Unaligned_SSE2(const uint8* src_ptr,
218 ptrdiff_t src_stride,
219 uint8* dst_ptr, int dst_width);
220 void ScaleRowDown4_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
221 uint8* dst_ptr, int dst_width);
222 void ScaleRowDown4Box_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
223 uint8* dst_ptr, int dst_width);
224 void ScaleRowDown34_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
225 uint8* dst_ptr, int dst_width);
226 void ScaleRowDown34_1_Box_SSSE3(const uint8* src_ptr,
227 ptrdiff_t src_stride,
228 uint8* dst_ptr, int dst_width);
229 void ScaleRowDown34_0_Box_SSSE3(const uint8* src_ptr,
230 ptrdiff_t src_stride,
231 uint8* dst_ptr, int dst_width);
232 void ScaleRowDown38_SSSE3(const uint8* src_ptr, ptrdiff_t src_stride,
233 uint8* dst_ptr, int dst_width);
234 void ScaleRowDown38_3_Box_SSSE3(const uint8* src_ptr,
235 ptrdiff_t src_stride,
236 uint8* dst_ptr, int dst_width);
237 void ScaleRowDown38_2_Box_SSSE3(const uint8* src_ptr,
238 ptrdiff_t src_stride,
239 uint8* dst_ptr, int dst_width);
240 void ScaleAddRows_SSE2(const uint8* src_ptr, ptrdiff_t src_stride,
241 uint16* dst_ptr, int src_width,
243 void ScaleFilterCols_SSSE3(uint8* dst_ptr, const uint8* src_ptr,
244 int dst_width, int x, int dx);
245 void ScaleColsUp2_SSE2(uint8* dst_ptr, const uint8* src_ptr,
246 int dst_width, int x, int dx);
247 void ScaleARGBRowDown2_SSE2(const uint8* src_argb,
248 ptrdiff_t src_stride,
249 uint8* dst_argb, int dst_width);
250 void ScaleARGBRowDown2Linear_SSE2(const uint8* src_argb,
251 ptrdiff_t src_stride,
252 uint8* dst_argb, int dst_width);
253 void ScaleARGBRowDown2Box_SSE2(const uint8* src_argb,
254 ptrdiff_t src_stride,
255 uint8* dst_argb, int dst_width);
256 void ScaleARGBRowDownEven_SSE2(const uint8* src_argb, ptrdiff_t src_stride,
258 uint8* dst_argb, int dst_width);
259 void ScaleARGBRowDownEvenBox_SSE2(const uint8* src_argb,
260 ptrdiff_t src_stride,
262 uint8* dst_argb, int dst_width);
263 void ScaleARGBCols_SSE2(uint8* dst_argb, const uint8* src_argb,
264 int dst_width, int x, int dx);
265 void ScaleARGBFilterCols_SSSE3(uint8* dst_argb, const uint8* src_argb,
266 int dst_width, int x, int dx);
267 void ScaleARGBColsUp2_SSE2(uint8* dst_argb, const uint8* src_argb,
268 int dst_width, int x, int dx);
270 void ScaleARGBRowDownEven_NEON(const uint8* src_argb, int src_stride,
272 uint8* dst_argb, int dst_width);
273 void ScaleARGBRowDownEvenBox_NEON(const uint8* src_argb, int src_stride,
275 uint8* dst_argb, int dst_width);
276 void ScaleARGBRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
277 uint8* dst, int dst_width);
278 void ScaleARGBRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
279 uint8* dst, int dst_width);
281 // ScaleRowDown2Box also used by planar functions
282 // NEON downscalers with interpolation.
284 // Note - not static due to reuse in convert for 444 to 420.
285 void ScaleRowDown2_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
286 uint8* dst, int dst_width);
288 void ScaleRowDown2Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
289 uint8* dst, int dst_width);
291 void ScaleRowDown4_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
292 uint8* dst_ptr, int dst_width);
293 void ScaleRowDown4Box_NEON(const uint8* src_ptr, ptrdiff_t src_stride,
294 uint8* dst_ptr, int dst_width);
296 // Down scale from 4 to 3 pixels. Use the neon multilane read/write
297 // to load up the every 4th pixel into a 4 different registers.
298 // Point samples 32 pixels to 24 pixels.
299 void ScaleRowDown34_NEON(const uint8* src_ptr,
300 ptrdiff_t src_stride,
301 uint8* dst_ptr, int dst_width);
302 void ScaleRowDown34_0_Box_NEON(const uint8* src_ptr,
303 ptrdiff_t src_stride,
304 uint8* dst_ptr, int dst_width);
305 void ScaleRowDown34_1_Box_NEON(const uint8* src_ptr,
306 ptrdiff_t src_stride,
307 uint8* dst_ptr, int dst_width);
310 void ScaleRowDown38_NEON(const uint8* src_ptr,
311 ptrdiff_t src_stride,
312 uint8* dst_ptr, int dst_width);
314 void ScaleRowDown38_3_Box_NEON(const uint8* src_ptr,
315 ptrdiff_t src_stride,
316 uint8* dst_ptr, int dst_width);
318 void ScaleRowDown38_2_Box_NEON(const uint8* src_ptr,
319 ptrdiff_t src_stride,
320 uint8* dst_ptr, int dst_width);
322 void ScaleRowDown2_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
323 uint8* dst, int dst_width);
324 void ScaleRowDown2Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
325 uint8* dst, int dst_width);
326 void ScaleRowDown4_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
327 uint8* dst, int dst_width);
328 void ScaleRowDown4Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
329 uint8* dst, int dst_width);
330 void ScaleRowDown34_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
331 uint8* dst, int dst_width);
332 void ScaleRowDown34_0_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
333 uint8* d, int dst_width);
334 void ScaleRowDown34_1_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
335 uint8* d, int dst_width);
336 void ScaleRowDown38_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
337 uint8* dst, int dst_width);
338 void ScaleRowDown38_2_Box_MIPS_DSPR2(const uint8* src_ptr, ptrdiff_t src_stride,
339 uint8* dst_ptr, int dst_width);
340 void ScaleRowDown38_3_Box_MIPS_DSPR2(const uint8* src_ptr,
341 ptrdiff_t src_stride,
342 uint8* dst_ptr, int dst_width);
346 } // namespace libyuv
349 #endif // INCLUDE_LIBYUV_SCALE_ROW_H_ NOLINT