Imported Upstream version 1.4
[deb_x265.git] / source / common / primitives.h
1 /*****************************************************************************
2 * Copyright (C) 2013 x265 project
3 *
4 * Authors: Steve Borho <steve@borho.org>
5 * Mandar Gurav <mandar@multicorewareinc.com>
6 * Deepthi Devaki Akkoorath <deepthidevaki@multicorewareinc.com>
7 * Mahesh Pittala <mahesh@multicorewareinc.com>
8 * Rajesh Paulraj <rajesh@multicorewareinc.com>
9 * Praveen Kumar Tiwari <praveen@multicorewareinc.com>
10 * Min Chen <chenm003@163.com>
11 *
12 * This program is free software; you can redistribute it and/or modify
13 * it under the terms of the GNU General Public License as published by
14 * the Free Software Foundation; either version 2 of the License, or
15 * (at your option) any later version.
16 *
17 * This program is distributed in the hope that it will be useful,
18 * but WITHOUT ANY WARRANTY; without even the implied warranty of
19 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
20 * GNU General Public License for more details.
21 *
22 * You should have received a copy of the GNU General Public License
23 * along with this program; if not, write to the Free Software
24 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02111, USA.
25 *
26 * This program is also available under a commercial proprietary license.
27 * For more information, contact us at license @ x265.com.
28 *****************************************************************************/
29
30 #ifndef X265_PRIMITIVES_H
31 #define X265_PRIMITIVES_H
32
33 #include "common.h"
34 #include "cpu.h"
35
36 namespace x265 {
37 // x265 private namespace
38
39 enum LumaPartitions
40 {
41 // Square
42 LUMA_4x4, LUMA_8x8, LUMA_16x16, LUMA_32x32, LUMA_64x64,
43 // Rectangular
44 LUMA_8x4, LUMA_4x8,
45 LUMA_16x8, LUMA_8x16,
46 LUMA_32x16, LUMA_16x32,
47 LUMA_64x32, LUMA_32x64,
48 // Asymmetrical (0.75, 0.25)
49 LUMA_16x12, LUMA_12x16, LUMA_16x4, LUMA_4x16,
50 LUMA_32x24, LUMA_24x32, LUMA_32x8, LUMA_8x32,
51 LUMA_64x48, LUMA_48x64, LUMA_64x16, LUMA_16x64,
52 NUM_LUMA_PARTITIONS
53 };
54
55 // 4:2:0 chroma partition sizes. These enums are just a convenience for indexing into the
56 // chroma primitive arrays when instantiating templates. The function tables should always
57 // be indexed by the luma partition enum
58 enum Chroma420Partitions
59 {
60 CHROMA_2x2, CHROMA_4x4, CHROMA_8x8, CHROMA_16x16, CHROMA_32x32,
61 CHROMA_4x2, CHROMA_2x4,
62 CHROMA_8x4, CHROMA_4x8,
63 CHROMA_16x8, CHROMA_8x16,
64 CHROMA_32x16, CHROMA_16x32,
65 CHROMA_8x6, CHROMA_6x8, CHROMA_8x2, CHROMA_2x8,
66 CHROMA_16x12, CHROMA_12x16, CHROMA_16x4, CHROMA_4x16,
67 CHROMA_32x24, CHROMA_24x32, CHROMA_32x8, CHROMA_8x32,
68 NUM_CHROMA_PARTITIONS
69 };
70
71 enum Chroma422Partitions
72 {
73 CHROMA422_2x4, CHROMA422_4x8, CHROMA422_8x16, CHROMA422_16x32, CHROMA422_32x64,
74 CHROMA422_4x4, CHROMA422_2x8,
75 CHROMA422_8x8, CHROMA422_4x16,
76 CHROMA422_16x16, CHROMA422_8x32,
77 CHROMA422_32x32, CHROMA422_16x64,
78 CHROMA422_8x12, CHROMA422_6x16, CHROMA422_8x4, CHROMA422_2x16,
79 CHROMA422_16x24, CHROMA422_12x32, CHROMA422_16x8, CHROMA422_4x32,
80 CHROMA422_32x48, CHROMA422_24x64, CHROMA422_32x16, CHROMA422_8x64,
81 NUM_CHROMA_PARTITIONS422
82 };
83
84 enum SquareBlocks // Routines can be indexed using log2n(width)-2
85 {
86 BLOCK_4x4,
87 BLOCK_8x8,
88 BLOCK_16x16,
89 BLOCK_32x32,
90 BLOCK_64x64,
91 NUM_SQUARE_BLOCKS
92 };
93
94 enum { NUM_TR_SIZE = 4 };
95
96 // NOTE: Not all DCT functions support dest stride
97 enum Dcts
98 {
99 DST_4x4,
100 DCT_4x4,
101 DCT_8x8,
102 DCT_16x16,
103 DCT_32x32,
104 NUM_DCTS
105 };
106
107 enum IDcts
108 {
109 IDST_4x4,
110 IDCT_4x4,
111 IDCT_8x8,
112 IDCT_16x16,
113 IDCT_32x32,
114 NUM_IDCTS
115 };
116
117 // Returns a LumaPartitions enum for the given size, always expected to return a valid enum
118 inline int partitionFromSizes(int width, int height)
119 {
120 X265_CHECK(((width | height) & ~(4 | 8 | 16 | 32 | 64)) == 0, "Invalid block width/height\n");
121 extern const uint8_t lumaPartitionMapTable[];
122 int w = (width >> 2) - 1;
123 int h = (height >> 2) - 1;
124 int part = (int)lumaPartitionMapTable[(w << 4) + h];
125 X265_CHECK(part != 255, "Invalid block width %d height %d\n", width, height);
126 return part;
127 }
128
129 inline int partitionFromLog2Size(int log2Size)
130 {
131 X265_CHECK(2 <= log2Size && log2Size <= 6, "Invalid block size\n");
132 return log2Size - 2;
133 }
134
135 typedef int (*pixelcmp_t)(pixel *fenc, intptr_t fencstride, pixel *fref, intptr_t frefstride); // fenc is aligned
136 typedef int (*pixelcmp_ss_t)(int16_t *fenc, intptr_t fencstride, int16_t *fref, intptr_t frefstride);
137 typedef int (*pixelcmp_sp_t)(int16_t *fenc, intptr_t fencstride, pixel *fref, intptr_t frefstride);
138 typedef int (*pixel_ssd_s_t)(int16_t *fenc, intptr_t fencstride);
139 typedef void (*pixelcmp_x4_t)(pixel *fenc, pixel *fref0, pixel *fref1, pixel *fref2, pixel *fref3, intptr_t frefstride, int32_t *res);
140 typedef void (*pixelcmp_x3_t)(pixel *fenc, pixel *fref0, pixel *fref1, pixel *fref2, intptr_t frefstride, int32_t *res);
141 typedef void (*blockcpy_sp_t)(int bx, int by, int16_t *dst, intptr_t dstride, pixel *src, intptr_t sstride); // dst is aligned
142 typedef void (*blockcpy_sc_t)(int bx, int by, int16_t *dst, intptr_t dstride, uint8_t *src, intptr_t sstride); // dst is aligned
143 typedef void (*pixelsub_ps_t)(int bx, int by, int16_t *dst, intptr_t dstride, pixel *src0, pixel *src1, intptr_t sstride0, intptr_t sstride1);
144 typedef void (*pixelavg_pp_t)(pixel *dst, intptr_t dstride, pixel *src0, intptr_t sstride0, pixel *src1, intptr_t sstride1, int weight);
145 typedef void (*blockfill_s_t)(int16_t *dst, intptr_t dstride, int16_t val);
146
147 typedef void (*intra_pred_t)(pixel* dst, intptr_t dstStride, pixel *refLeft, pixel *refAbove, int dirMode, int bFilter);
148 typedef void (*intra_allangs_t)(pixel *dst, pixel *above0, pixel *left0, pixel *above1, pixel *left1, int bLuma);
149
150 typedef void (*cvt16to32_shl_t)(int32_t *dst, int16_t *src, intptr_t, int, int);
151 typedef void (*cvt16to32_shr_t)(int32_t *dst, int16_t *src, intptr_t, int, int);
152 typedef void (*cvt32to16_shr_t)(int16_t *dst, int32_t *src, intptr_t, int, int);
153 typedef void (*cvt32to16_shl_t)(int16_t *dst, int32_t *src, intptr_t, int);
154 typedef uint32_t (*copy_cnt_t)(int16_t* coeff, int16_t* residual, intptr_t stride);
155 typedef void (*copy_shr_t)(int16_t *dst, int16_t *src, intptr_t stride, int shift, int size);
156 typedef void (*copy_shl_t)(int16_t *dst, int16_t *src, intptr_t stride, int shift);
157
158 typedef void (*dct_t)(int16_t *src, int32_t *dst, intptr_t stride);
159 typedef void (*idct_t)(int32_t *src, int16_t *dst, intptr_t stride);
160 typedef void (*denoiseDct_t)(int32_t* dctCoef, uint32_t* resSum, uint16_t* offset, int numCoeff);
161
162 typedef void (*calcresidual_t)(pixel *fenc, pixel *pred, int16_t *residual, intptr_t stride);
163 typedef void (*calcrecon_t)(pixel* pred, int16_t* residual, int16_t* reconqt, pixel *reconipred, int stride, int strideqt, int strideipred);
164 typedef void (*transpose_t)(pixel* dst, pixel* src, intptr_t stride);
165 typedef uint32_t (*quant_t)(int32_t *coef, int32_t *quantCoeff, int32_t *deltaU, int16_t *qCoef, int qBits, int add, int numCoeff);
166 typedef uint32_t (*nquant_t)(int32_t *coef, int32_t *quantCoeff, int16_t *qCoef, int qBits, int add, int numCoeff);
167 typedef void (*dequant_scaling_t)(const int16_t* src, const int32_t *dequantCoef, int32_t* dst, int num, int mcqp_miper, int shift);
168 typedef void (*dequant_normal_t)(const int16_t* quantCoef, int32_t* coef, int num, int scale, int shift);
169 typedef int (*count_nonzero_t)(const int16_t *quantCoeff, int numCoeff);
170
171 typedef void (*weightp_pp_t)(pixel *src, pixel *dst, intptr_t stride, int width, int height, int w0, int round, int shift, int offset);
172 typedef void (*weightp_sp_t)(int16_t *src, pixel *dst, intptr_t srcStride, intptr_t dstStride, int width, int height, int w0, int round, int shift, int offset);
173 typedef void (*scale_t)(pixel *dst, pixel *src, intptr_t stride);
174 typedef void (*downscale_t)(pixel *src0, pixel *dstf, pixel *dsth, pixel *dstv, pixel *dstc,
175 intptr_t src_stride, intptr_t dst_stride, int width, int height);
176 typedef void (*extendCURowBorder_t)(pixel* txt, intptr_t stride, int width, int height, int marginX);
177 typedef void (*ssim_4x4x2_core_t)(const pixel *pix1, intptr_t stride1, const pixel *pix2, intptr_t stride2, int sums[2][4]);
178 typedef float (*ssim_end4_t)(int sum0[5][4], int sum1[5][4], int width);
179 typedef uint64_t (*var_t)(pixel *pix, intptr_t stride);
180 typedef void (*plane_copy_deinterleave_t)(pixel *dstu, intptr_t dstuStride, pixel *dstv, intptr_t dstvStride, pixel *src, intptr_t srcStride, int w, int h);
181
182 typedef void (*filter_pp_t) (pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
183 typedef void (*filter_hps_t) (pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx, int isRowExt);
184 typedef void (*filter_ps_t) (pixel *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx);
185 typedef void (*filter_sp_t) (int16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int coeffIdx);
186 typedef void (*filter_ss_t) (int16_t *src, intptr_t srcStride, int16_t *dst, intptr_t dstStride, int coeffIdx);
187 typedef void (*filter_hv_pp_t) (pixel *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int idxX, int idxY);
188 typedef void (*filter_p2s_t)(pixel *src, intptr_t srcStride, int16_t *dst, int width, int height);
189
190 typedef void (*copy_pp_t)(pixel *dst, intptr_t dstride, pixel *src, intptr_t sstride); // dst is aligned
191 typedef void (*copy_sp_t)(pixel *dst, intptr_t dstStride, int16_t *src, intptr_t srcStride);
192 typedef void (*copy_ps_t)(int16_t *dst, intptr_t dstStride, pixel *src, intptr_t srcStride);
193 typedef void (*copy_ss_t)(int16_t *dst, intptr_t dstStride, int16_t *src, intptr_t srcStride);
194
195 typedef void (*pixel_sub_ps_t)(int16_t *dst, intptr_t dstride, pixel *src0, pixel *src1, intptr_t sstride0, intptr_t sstride1);
196 typedef void (*pixel_add_ps_t)(pixel *a, intptr_t dstride, pixel *b0, int16_t *b1, intptr_t sstride0, intptr_t sstride1);
197 typedef void (*addAvg_t)(int16_t* src0, int16_t* src1, pixel* dst, intptr_t src0Stride, intptr_t src1Stride, intptr_t dstStride);
198
199 typedef void (*saoCuOrgE0_t)(pixel * rec, int8_t * offsetEo, int width, int8_t signLeft);
200 typedef void (*planecopy_cp_t) (uint8_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int width, int height, int shift);
201 typedef void (*planecopy_sp_t) (uint16_t *src, intptr_t srcStride, pixel *dst, intptr_t dstStride, int width, int height, int shift, uint16_t mask);
202
203 typedef void (*cutree_propagate_cost) (int *dst, uint16_t *propagateIn, int32_t *intraCosts, uint16_t *interCosts, int32_t *invQscales, double *fpsFactor, int len);
204
205 /* Define a structure containing function pointers to optimized encoder
206 * primitives. Each pointer can reference either an assembly routine,
207 * a vectorized primitive, or a C function. */
208 struct EncoderPrimitives
209 {
210 pixelcmp_t sad[NUM_LUMA_PARTITIONS]; // Sum of Differences for each size
211 pixelcmp_x3_t sad_x3[NUM_LUMA_PARTITIONS]; // Sum of Differences 3x for each size
212 pixelcmp_x4_t sad_x4[NUM_LUMA_PARTITIONS]; // Sum of Differences 4x for each size
213 pixelcmp_t sse_pp[NUM_LUMA_PARTITIONS]; // Sum of Square Error (pixel, pixel) fenc alignment not assumed
214 pixelcmp_ss_t sse_ss[NUM_LUMA_PARTITIONS]; // Sum of Square Error (short, short) fenc alignment not assumed
215 pixelcmp_sp_t sse_sp[NUM_LUMA_PARTITIONS]; // Sum of Square Error (short, pixel) fenc alignment not assumed
216 pixel_ssd_s_t ssd_s[NUM_SQUARE_BLOCKS - 1]; // Sum of Square Error (short) fenc alignment not assumed
217 pixelcmp_t satd[NUM_LUMA_PARTITIONS]; // Sum of Transformed differences (HADAMARD)
218 pixelcmp_t sa8d_inter[NUM_LUMA_PARTITIONS]; // sa8d primitives for motion search partitions
219 pixelcmp_t sa8d[NUM_SQUARE_BLOCKS]; // sa8d primitives for square intra blocks
220 pixelcmp_t psy_cost_pp[NUM_SQUARE_BLOCKS]; // difference in AC energy between two blocks
221 pixelcmp_ss_t psy_cost_ss[NUM_SQUARE_BLOCKS];
222
223 blockfill_s_t blockfill_s[NUM_SQUARE_BLOCKS]; // block fill with value
224 cvt16to32_shl_t cvt16to32_shl;
225 cvt16to32_shr_t cvt16to32_shr[NUM_SQUARE_BLOCKS - 1];
226 cvt32to16_shr_t cvt32to16_shr;
227 cvt32to16_shl_t cvt32to16_shl[NUM_SQUARE_BLOCKS - 1];
228 copy_cnt_t copy_cnt[NUM_SQUARE_BLOCKS - 1];
229 copy_shr_t copy_shr;
230 copy_shl_t copy_shl[NUM_SQUARE_BLOCKS - 1];
231
232 copy_pp_t luma_copy_pp[NUM_LUMA_PARTITIONS];
233 copy_sp_t luma_copy_sp[NUM_LUMA_PARTITIONS];
234 copy_ps_t luma_copy_ps[NUM_LUMA_PARTITIONS];
235 copy_ss_t luma_copy_ss[NUM_LUMA_PARTITIONS];
236 pixel_sub_ps_t luma_sub_ps[NUM_SQUARE_BLOCKS];
237 pixel_add_ps_t luma_add_ps[NUM_SQUARE_BLOCKS];
238 copy_pp_t square_copy_pp[NUM_SQUARE_BLOCKS];
239 copy_sp_t square_copy_sp[NUM_SQUARE_BLOCKS];
240 copy_ps_t square_copy_ps[NUM_SQUARE_BLOCKS];
241 copy_ss_t square_copy_ss[NUM_SQUARE_BLOCKS];
242
243 filter_pp_t luma_hpp[NUM_LUMA_PARTITIONS];
244 filter_hps_t luma_hps[NUM_LUMA_PARTITIONS];
245 filter_pp_t luma_vpp[NUM_LUMA_PARTITIONS];
246 filter_ps_t luma_vps[NUM_LUMA_PARTITIONS];
247 filter_sp_t luma_vsp[NUM_LUMA_PARTITIONS];
248 filter_ss_t luma_vss[NUM_LUMA_PARTITIONS];
249 filter_hv_pp_t luma_hvpp[NUM_LUMA_PARTITIONS];
250 filter_p2s_t luma_p2s;
251 filter_p2s_t chroma_p2s[X265_CSP_COUNT];
252
253 weightp_sp_t weight_sp;
254 weightp_pp_t weight_pp;
255 pixelavg_pp_t pixelavg_pp[NUM_LUMA_PARTITIONS];
256 addAvg_t luma_addAvg[NUM_LUMA_PARTITIONS];
257
258 intra_pred_t intra_pred[NUM_INTRA_MODE][NUM_TR_SIZE];
259 intra_allangs_t intra_pred_allangs[NUM_TR_SIZE];
260 scale_t scale1D_128to64;
261 scale_t scale2D_64to32;
262
263 dct_t dct[NUM_DCTS];
264 idct_t idct[NUM_IDCTS];
265 quant_t quant;
266 nquant_t nquant;
267 dequant_scaling_t dequant_scaling;
268 dequant_normal_t dequant_normal;
269 count_nonzero_t count_nonzero;
270 denoiseDct_t denoiseDct;
271
272 calcresidual_t calcresidual[NUM_SQUARE_BLOCKS];
273 transpose_t transpose[NUM_SQUARE_BLOCKS];
274
275 var_t var[NUM_SQUARE_BLOCKS];
276 ssim_4x4x2_core_t ssim_4x4x2_core;
277 ssim_end4_t ssim_end_4;
278
279 downscale_t frame_init_lowres_core;
280 plane_copy_deinterleave_t plane_copy_deinterleave_c;
281 extendCURowBorder_t extendRowBorder;
282 // sao primitives
283 saoCuOrgE0_t saoCuOrgE0;
284 planecopy_cp_t planecopy_cp;
285 planecopy_sp_t planecopy_sp;
286
287 cutree_propagate_cost propagateCost;
288
289 struct
290 {
291 filter_pp_t filter_vpp[NUM_LUMA_PARTITIONS];
292 filter_ps_t filter_vps[NUM_LUMA_PARTITIONS];
293 filter_sp_t filter_vsp[NUM_LUMA_PARTITIONS];
294 filter_ss_t filter_vss[NUM_LUMA_PARTITIONS];
295 filter_pp_t filter_hpp[NUM_LUMA_PARTITIONS];
296 filter_hps_t filter_hps[NUM_LUMA_PARTITIONS];
297 addAvg_t addAvg[NUM_LUMA_PARTITIONS];
298 copy_pp_t copy_pp[NUM_LUMA_PARTITIONS];
299 copy_sp_t copy_sp[NUM_LUMA_PARTITIONS];
300 copy_ps_t copy_ps[NUM_LUMA_PARTITIONS];
301 copy_ss_t copy_ss[NUM_LUMA_PARTITIONS];
302 pixel_sub_ps_t sub_ps[NUM_SQUARE_BLOCKS];
303 pixel_add_ps_t add_ps[NUM_SQUARE_BLOCKS];
304 } chroma[4]; // X265_CSP_COUNT - do not want to include x265.h here
305 };
306
307 void extendPicBorder(pixel* recon, intptr_t stride, int width, int height, int marginX, int marginY);
308
309 /* This copy of the table is what gets used by the encoder.
310 * It must be initialized before the encoder begins. */
311 extern EncoderPrimitives primitives;
312
313 void Setup_C_Primitives(EncoderPrimitives &p);
314 void Setup_Instrinsic_Primitives(EncoderPrimitives &p, int cpuMask);
315 void Setup_Assembly_Primitives(EncoderPrimitives &p, int cpuMask);
316 void Setup_Alias_Primitives(EncoderPrimitives &p);
317 }
318
319 #endif // ifndef X265_PRIMITIVES_H