f19cfc38d7b7f011cbe852120df9739dad0c01b4
[mesa.git] / src / gallium / auxiliary / vl / vl_idct.c
1 /**************************************************************************
2 *
3 * Copyright 2010 Christian König
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28 #include "vl_idct.h"
29 #include "vl_vertex_buffers.h"
30 #include "util/u_draw.h"
31 #include <assert.h>
32 #include <pipe/p_context.h>
33 #include <pipe/p_screen.h>
34 #include <util/u_inlines.h>
35 #include <util/u_sampler.h>
36 #include <util/u_format.h>
37 #include <tgsi/tgsi_ureg.h>
38 #include "vl_types.h"
39
40 #define BLOCK_WIDTH 8
41 #define BLOCK_HEIGHT 8
42
43 #define SCALE_FACTOR_16_TO_9 (32768.0f / 256.0f)
44
45 #define STAGE1_SCALE 4.0f
46 #define STAGE2_SCALE (SCALE_FACTOR_16_TO_9 / STAGE1_SCALE)
47
48 #define NR_RENDER_TARGETS 1
49
50 enum VS_INPUT
51 {
52 VS_I_RECT,
53 VS_I_VPOS,
54
55 NUM_VS_INPUTS
56 };
57
58 enum VS_OUTPUT
59 {
60 VS_O_VPOS,
61 VS_O_BLOCK,
62 VS_O_TEX,
63 VS_O_START
64 };
65
66 static const float const_matrix[8][8] = {
67 { 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.3535530f, 0.353553f, 0.3535530f },
68 { 0.4903930f, 0.4157350f, 0.2777850f, 0.0975451f, -0.0975452f, -0.2777850f, -0.415735f, -0.4903930f },
69 { 0.4619400f, 0.1913420f, -0.1913420f, -0.4619400f, -0.4619400f, -0.1913420f, 0.191342f, 0.4619400f },
70 { 0.4157350f, -0.0975452f, -0.4903930f, -0.2777850f, 0.2777850f, 0.4903930f, 0.097545f, -0.4157350f },
71 { 0.3535530f, -0.3535530f, -0.3535530f, 0.3535540f, 0.3535530f, -0.3535540f, -0.353553f, 0.3535530f },
72 { 0.2777850f, -0.4903930f, 0.0975452f, 0.4157350f, -0.4157350f, -0.0975451f, 0.490393f, -0.2777850f },
73 { 0.1913420f, -0.4619400f, 0.4619400f, -0.1913420f, -0.1913410f, 0.4619400f, -0.461940f, 0.1913420f },
74 { 0.0975451f, -0.2777850f, 0.4157350f, -0.4903930f, 0.4903930f, -0.4157350f, 0.277786f, -0.0975458f }
75 };
76
77 static void *
78 create_vert_shader(struct vl_idct *idct)
79 {
80 struct ureg_program *shader;
81 struct ureg_src scale;
82 struct ureg_src vrect, vpos;
83 struct ureg_dst t_vpos;
84 struct ureg_dst o_vpos, o_block, o_tex, o_start;
85
86 shader = ureg_create(TGSI_PROCESSOR_VERTEX);
87 if (!shader)
88 return NULL;
89
90 t_vpos = ureg_DECL_temporary(shader);
91
92 vrect = ureg_DECL_vs_input(shader, VS_I_RECT);
93 vpos = ureg_DECL_vs_input(shader, VS_I_VPOS);
94
95 o_vpos = ureg_DECL_output(shader, TGSI_SEMANTIC_POSITION, VS_O_VPOS);
96 o_block = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK);
97 o_tex = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX);
98 o_start = ureg_DECL_output(shader, TGSI_SEMANTIC_GENERIC, VS_O_START);
99
100 /*
101 * scale = (BLOCK_WIDTH, BLOCK_HEIGHT) / (dst.width, dst.height)
102 *
103 * t_vpos = vpos + vrect
104 * o_vpos.xy = t_vpos * scale
105 * o_vpos.zw = vpos
106 *
107 * o_block = vrect
108 * o_tex = t_pos
109 * o_start = vpos * scale
110 *
111 */
112 scale = ureg_imm2f(shader,
113 (float)BLOCK_WIDTH / idct->buffer_width,
114 (float)BLOCK_HEIGHT / idct->buffer_height);
115
116 ureg_ADD(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), vpos, vrect);
117 ureg_MUL(shader, ureg_writemask(t_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos), scale);
118 ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
119 ureg_MOV(shader, ureg_writemask(o_vpos, TGSI_WRITEMASK_ZW), vpos);
120
121 ureg_MOV(shader, ureg_writemask(o_block, TGSI_WRITEMASK_XY), vrect);
122 ureg_MOV(shader, ureg_writemask(o_tex, TGSI_WRITEMASK_XY), ureg_src(t_vpos));
123 ureg_MUL(shader, ureg_writemask(o_start, TGSI_WRITEMASK_XY), vpos, scale);
124
125 ureg_release_temporary(shader, t_vpos);
126
127 ureg_END(shader);
128
129 return ureg_create_shader_and_destroy(shader, idct->pipe);
130 }
131
132 static void
133 fetch_four(struct ureg_program *shader, struct ureg_dst m[2],
134 struct ureg_src tc, struct ureg_src sampler,
135 struct ureg_src start, struct ureg_src block,
136 bool right_side, bool transposed, float size)
137 {
138 struct ureg_dst t_tc;
139 unsigned wm_start = (right_side == transposed) ? TGSI_WRITEMASK_X : TGSI_WRITEMASK_Y;
140 unsigned wm_tc = (right_side == transposed) ? TGSI_WRITEMASK_Y : TGSI_WRITEMASK_X;
141
142 t_tc = ureg_DECL_temporary(shader);
143 m[0] = ureg_DECL_temporary(shader);
144 m[1] = ureg_DECL_temporary(shader);
145
146 /*
147 * t_tc.x = right_side ? start.x : tc.x
148 * t_tc.y = right_side ? tc.y : start.y
149 * m[0..1] = tex(t_tc++, sampler)
150 */
151 if(!right_side) {
152 ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_X));
153 ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_Y));
154 } else {
155 ureg_MOV(shader, ureg_writemask(t_tc, wm_start), ureg_scalar(start, TGSI_SWIZZLE_Y));
156 ureg_MOV(shader, ureg_writemask(t_tc, wm_tc), ureg_scalar(tc, TGSI_SWIZZLE_X));
157 }
158
159 #if NR_RENDER_TARGETS == 8
160 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_scalar(block, TGSI_SWIZZLE_X));
161 #else
162 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Z), ureg_imm1f(shader, 0.0f));
163 #endif
164
165 ureg_TEX(shader, m[0], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler);
166 ureg_ADD(shader, ureg_writemask(t_tc, wm_start), ureg_src(t_tc), ureg_imm1f(shader, 1.0f / size));
167 ureg_TEX(shader, m[1], TGSI_TEXTURE_3D, ureg_src(t_tc), sampler);
168
169 ureg_release_temporary(shader, t_tc);
170 }
171
172 static void
173 matrix_mul(struct ureg_program *shader, struct ureg_dst dst, struct ureg_dst l[2], struct ureg_dst r[2])
174 {
175 struct ureg_dst tmp[2];
176 unsigned i;
177
178 for(i = 0; i < 2; ++i) {
179 tmp[i] = ureg_DECL_temporary(shader);
180 }
181
182 /*
183 * tmp[0..1] = dot4(m[0][0..1], m[1][0..1])
184 * dst = tmp[0] + tmp[1]
185 */
186 ureg_DP4(shader, ureg_writemask(tmp[0], TGSI_WRITEMASK_X), ureg_src(l[0]), ureg_src(r[0]));
187 ureg_DP4(shader, ureg_writemask(tmp[1], TGSI_WRITEMASK_X), ureg_src(l[1]), ureg_src(r[1]));
188 ureg_ADD(shader, dst,
189 ureg_scalar(ureg_src(tmp[0]), TGSI_SWIZZLE_X),
190 ureg_scalar(ureg_src(tmp[1]), TGSI_SWIZZLE_X));
191
192 for(i = 0; i < 2; ++i) {
193 ureg_release_temporary(shader, tmp[i]);
194 }
195 }
196
197 static void *
198 create_transpose_frag_shader(struct vl_idct *idct)
199 {
200 struct pipe_resource *transpose = idct->textures.individual.transpose;
201 struct pipe_resource *intermediate = idct->textures.individual.intermediate;
202
203 struct ureg_program *shader;
204
205 struct ureg_src block, tex, sampler[2];
206 struct ureg_src start[2];
207
208 struct ureg_dst l[2], r[2];
209 struct ureg_dst tmp, fragment;
210
211 shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
212 if (!shader)
213 return NULL;
214
215 block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
216 tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_CONSTANT);
217
218 sampler[0] = ureg_DECL_sampler(shader, 0);
219 sampler[1] = ureg_DECL_sampler(shader, 1);
220
221 start[0] = ureg_imm1f(shader, 0.0f);
222 start[1] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
223
224 fetch_four(shader, l, block, sampler[0], start[0], block, false, false, transpose->width0);
225 fetch_four(shader, r, tex, sampler[1], start[1], block, true, false, intermediate->height0);
226
227 fragment = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, 0);
228
229 tmp = ureg_DECL_temporary(shader);
230 matrix_mul(shader, ureg_writemask(tmp, TGSI_WRITEMASK_X), l, r);
231 ureg_MUL(shader, fragment, ureg_src(tmp), ureg_imm1f(shader, STAGE2_SCALE));
232
233 ureg_release_temporary(shader, tmp);
234 ureg_release_temporary(shader, l[0]);
235 ureg_release_temporary(shader, l[1]);
236 ureg_release_temporary(shader, r[0]);
237 ureg_release_temporary(shader, r[1]);
238
239 ureg_END(shader);
240
241 return ureg_create_shader_and_destroy(shader, idct->pipe);
242 }
243
244 static void *
245 create_matrix_frag_shader(struct vl_idct *idct)
246 {
247 struct pipe_resource *matrix = idct->textures.individual.matrix;
248 struct pipe_resource *source = idct->textures.individual.source;
249
250 struct ureg_program *shader;
251
252 struct ureg_src tex, block, sampler[2];
253 struct ureg_src start[2];
254
255 struct ureg_dst l[4][2], r[2];
256 struct ureg_dst t_tc, tmp, fragment[NR_RENDER_TARGETS];
257
258 unsigned i, j;
259
260 shader = ureg_create(TGSI_PROCESSOR_FRAGMENT);
261 if (!shader)
262 return NULL;
263
264 t_tc = ureg_DECL_temporary(shader);
265 tmp = ureg_DECL_temporary(shader);
266
267 tex = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_TEX, TGSI_INTERPOLATE_LINEAR);
268 block = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_BLOCK, TGSI_INTERPOLATE_LINEAR);
269
270 sampler[0] = ureg_DECL_sampler(shader, 1);
271 sampler[1] = ureg_DECL_sampler(shader, 0);
272
273 start[0] = ureg_DECL_fs_input(shader, TGSI_SEMANTIC_GENERIC, VS_O_START, TGSI_INTERPOLATE_CONSTANT);
274 start[1] = ureg_imm1f(shader, 0.0f);
275
276 for (i = 0; i < NR_RENDER_TARGETS; ++i)
277 fragment[i] = ureg_DECL_output(shader, TGSI_SEMANTIC_COLOR, i);
278
279 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y), tex);
280 for (i = 0; i < 4; ++i) {
281 fetch_four(shader, l[i], ureg_src(t_tc), sampler[0], start[0], block, false, false, source->width0);
282 ureg_MUL(shader, l[i][0], ureg_src(l[i][0]), ureg_imm1f(shader, STAGE1_SCALE));
283 ureg_MUL(shader, l[i][1], ureg_src(l[i][1]), ureg_imm1f(shader, STAGE1_SCALE));
284 if(i != 3)
285 ureg_ADD(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_Y),
286 ureg_src(t_tc), ureg_imm1f(shader, 1.0f / source->height0));
287 }
288
289 for (i = 0; i < NR_RENDER_TARGETS; ++i) {
290
291 #if NR_RENDER_TARGETS == 8
292 ureg_MOV(shader, ureg_writemask(t_tc, TGSI_WRITEMASK_X), ureg_imm1f(shader, 1.0f / BLOCK_WIDTH * i));
293 fetch_four(shader, r, ureg_src(t_tc), sampler[1], start[1], block, true, true, matrix->width0);
294 #elif NR_RENDER_TARGETS == 1
295 fetch_four(shader, r, block, sampler[1], start[1], block, true, true, matrix->width0);
296 #else
297 #error invalid number of render targets
298 #endif
299
300 for (j = 0; j < 4; ++j) {
301 matrix_mul(shader, ureg_writemask(fragment[i], TGSI_WRITEMASK_X << j), l[j], r);
302 }
303 ureg_release_temporary(shader, r[0]);
304 ureg_release_temporary(shader, r[1]);
305 }
306
307 ureg_release_temporary(shader, t_tc);
308 ureg_release_temporary(shader, tmp);
309
310 for (i = 0; i < 4; ++i) {
311 ureg_release_temporary(shader, l[i][0]);
312 ureg_release_temporary(shader, l[i][1]);
313 }
314
315 ureg_END(shader);
316
317 return ureg_create_shader_and_destroy(shader, idct->pipe);
318 }
319
320 static bool
321 init_shaders(struct vl_idct *idct)
322 {
323 idct->vs = create_vert_shader(idct);
324 idct->matrix_fs = create_matrix_frag_shader(idct);
325 idct->transpose_fs = create_transpose_frag_shader(idct);
326
327 return
328 idct->vs != NULL &&
329 idct->transpose_fs != NULL &&
330 idct->matrix_fs != NULL;
331 }
332
333 static void
334 cleanup_shaders(struct vl_idct *idct)
335 {
336 idct->pipe->delete_vs_state(idct->pipe, idct->vs);
337 idct->pipe->delete_fs_state(idct->pipe, idct->matrix_fs);
338 idct->pipe->delete_fs_state(idct->pipe, idct->transpose_fs);
339 }
340
341 static bool
342 init_buffers(struct vl_idct *idct)
343 {
344 struct pipe_resource template;
345 struct pipe_sampler_view sampler_view;
346 struct pipe_vertex_element vertex_elems[2];
347 unsigned i;
348
349 memset(&template, 0, sizeof(struct pipe_resource));
350 template.last_level = 0;
351 template.depth0 = 1;
352 template.bind = PIPE_BIND_SAMPLER_VIEW;
353 template.flags = 0;
354
355 template.target = PIPE_TEXTURE_2D;
356 template.format = PIPE_FORMAT_R16G16B16A16_SNORM;
357 template.width0 = idct->destination->width0 / 4;
358 template.height0 = idct->destination->height0;
359 template.depth0 = 1;
360 template.usage = PIPE_USAGE_STREAM;
361 idct->textures.individual.source = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
362
363 template.target = PIPE_TEXTURE_3D;
364 template.format = PIPE_FORMAT_R16G16B16A16_SNORM;
365 template.width0 = idct->destination->width0 / NR_RENDER_TARGETS;
366 template.height0 = idct->destination->height0 / 4;
367 template.depth0 = NR_RENDER_TARGETS;
368 template.usage = PIPE_USAGE_STATIC;
369 idct->textures.individual.intermediate = idct->pipe->screen->resource_create(idct->pipe->screen, &template);
370
371 for (i = 0; i < 4; ++i) {
372 if(idct->textures.all[i] == NULL)
373 return false; /* a texture failed to allocate */
374
375 u_sampler_view_default_template(&sampler_view, idct->textures.all[i], idct->textures.all[i]->format);
376 idct->sampler_views.all[i] = idct->pipe->create_sampler_view(idct->pipe, idct->textures.all[i], &sampler_view);
377 }
378
379 idct->vertex_bufs.individual.quad = vl_vb_upload_quads(idct->pipe, idct->max_blocks, &vertex_elems[VS_I_RECT]);
380
381 if(idct->vertex_bufs.individual.quad.buffer == NULL)
382 return false;
383
384 /* Pos element */
385 vertex_elems[VS_I_VPOS].src_format = PIPE_FORMAT_R32G32_FLOAT;
386
387 idct->vertex_bufs.individual.pos = vl_vb_create_buffer(idct->pipe, idct->max_blocks, &vertex_elems[VS_I_VPOS], 1, 1);
388
389 if(idct->vertex_bufs.individual.pos.buffer == NULL)
390 return false;
391
392 idct->vertex_elems_state = idct->pipe->create_vertex_elements_state(idct->pipe, 2, vertex_elems);
393
394 return true;
395 }
396
397 static void
398 cleanup_buffers(struct vl_idct *idct)
399 {
400 unsigned i;
401
402 assert(idct);
403
404 for (i = 0; i < 4; ++i) {
405 pipe_sampler_view_reference(&idct->sampler_views.all[i], NULL);
406 pipe_resource_reference(&idct->textures.all[i], NULL);
407 }
408
409 idct->pipe->delete_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
410 pipe_resource_reference(&idct->vertex_bufs.individual.quad.buffer, NULL);
411 pipe_resource_reference(&idct->vertex_bufs.individual.pos.buffer, NULL);
412 }
413
414 static void
415 init_state(struct vl_idct *idct)
416 {
417 struct pipe_sampler_state sampler;
418 struct pipe_rasterizer_state rs_state;
419 unsigned i;
420
421 idct->viewport[0].scale[0] = idct->textures.individual.intermediate->width0;
422 idct->viewport[0].scale[1] = idct->textures.individual.intermediate->height0;
423
424 idct->viewport[1].scale[0] = idct->destination->width0;
425 idct->viewport[1].scale[1] = idct->destination->height0;
426
427 idct->fb_state[0].width = idct->textures.individual.intermediate->width0;
428 idct->fb_state[0].height = idct->textures.individual.intermediate->height0;
429
430 idct->fb_state[0].nr_cbufs = NR_RENDER_TARGETS;
431 for(i = 0; i < NR_RENDER_TARGETS; ++i) {
432 idct->fb_state[0].cbufs[i] = idct->pipe->screen->get_tex_surface(
433 idct->pipe->screen, idct->textures.individual.intermediate, 0, 0, i,
434 PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
435 }
436
437 idct->fb_state[1].width = idct->destination->width0;
438 idct->fb_state[1].height = idct->destination->height0;
439
440 idct->fb_state[1].nr_cbufs = 1;
441 idct->fb_state[1].cbufs[0] = idct->pipe->screen->get_tex_surface(
442 idct->pipe->screen, idct->destination, 0, 0, 0,
443 PIPE_BIND_SAMPLER_VIEW | PIPE_BIND_RENDER_TARGET);
444
445 for(i = 0; i < 2; ++i) {
446 idct->viewport[i].scale[2] = 1;
447 idct->viewport[i].scale[3] = 1;
448 idct->viewport[i].translate[0] = 0;
449 idct->viewport[i].translate[1] = 0;
450 idct->viewport[i].translate[2] = 0;
451 idct->viewport[i].translate[3] = 0;
452
453 idct->fb_state[i].zsbuf = NULL;
454 }
455
456 for (i = 0; i < 4; ++i) {
457 memset(&sampler, 0, sizeof(sampler));
458 sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
459 sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
460 sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
461 sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
462 sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
463 sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
464 sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
465 sampler.compare_func = PIPE_FUNC_ALWAYS;
466 sampler.normalized_coords = 1;
467 /*sampler.shadow_ambient = ; */
468 /*sampler.lod_bias = ; */
469 sampler.min_lod = 0;
470 /*sampler.max_lod = ; */
471 /*sampler.border_color[0] = ; */
472 /*sampler.max_anisotropy = ; */
473 idct->samplers.all[i] = idct->pipe->create_sampler_state(idct->pipe, &sampler);
474 }
475
476 memset(&rs_state, 0, sizeof(rs_state));
477 /*rs_state.sprite_coord_enable */
478 rs_state.sprite_coord_mode = PIPE_SPRITE_COORD_UPPER_LEFT;
479 rs_state.point_quad_rasterization = true;
480 rs_state.point_size = BLOCK_WIDTH;
481 rs_state.gl_rasterization_rules = false;
482 idct->rs_state = idct->pipe->create_rasterizer_state(idct->pipe, &rs_state);
483 }
484
485 static void
486 cleanup_state(struct vl_idct *idct)
487 {
488 unsigned i;
489
490 for(i = 0; i < NR_RENDER_TARGETS; ++i) {
491 idct->pipe->screen->tex_surface_destroy(idct->fb_state[0].cbufs[i]);
492 }
493
494 idct->pipe->screen->tex_surface_destroy(idct->fb_state[1].cbufs[0]);
495
496 for (i = 0; i < 4; ++i)
497 idct->pipe->delete_sampler_state(idct->pipe, idct->samplers.all[i]);
498
499 idct->pipe->delete_rasterizer_state(idct->pipe, idct->rs_state);
500 }
501
502 struct pipe_resource *
503 vl_idct_upload_matrix(struct pipe_context *pipe)
504 {
505 struct pipe_resource template, *matrix;
506 struct pipe_transfer *buf_transfer;
507 unsigned i, j, pitch;
508 float *f;
509
510 struct pipe_box rect =
511 {
512 0, 0, 0,
513 BLOCK_WIDTH,
514 BLOCK_HEIGHT,
515 1
516 };
517
518 memset(&template, 0, sizeof(struct pipe_resource));
519 template.target = PIPE_TEXTURE_2D;
520 template.format = PIPE_FORMAT_R32G32B32A32_FLOAT;
521 template.last_level = 0;
522 template.width0 = 2;
523 template.height0 = 8;
524 template.depth0 = 1;
525 template.usage = PIPE_USAGE_IMMUTABLE;
526 template.bind = PIPE_BIND_SAMPLER_VIEW;
527 template.flags = 0;
528
529 matrix = pipe->screen->resource_create(pipe->screen, &template);
530
531 /* matrix */
532 buf_transfer = pipe->get_transfer
533 (
534 pipe, matrix,
535 u_subresource(0, 0),
536 PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
537 &rect
538 );
539 pitch = buf_transfer->stride / sizeof(float);
540
541 f = pipe->transfer_map(pipe, buf_transfer);
542 for(i = 0; i < BLOCK_HEIGHT; ++i)
543 for(j = 0; j < BLOCK_WIDTH; ++j)
544 f[i * pitch + j] = const_matrix[j][i]; // transpose
545
546 pipe->transfer_unmap(pipe, buf_transfer);
547 pipe->transfer_destroy(pipe, buf_transfer);
548
549 return matrix;
550 }
551
552 bool
553 vl_idct_init(struct vl_idct *idct, struct pipe_context *pipe, struct pipe_resource *dst, struct pipe_resource *matrix)
554 {
555 assert(idct && pipe && dst);
556
557 idct->pipe = pipe;
558 idct->buffer_width = dst->width0;
559 idct->buffer_height = dst->height0;
560
561 pipe_resource_reference(&idct->textures.individual.matrix, matrix);
562 pipe_resource_reference(&idct->textures.individual.transpose, matrix);
563 pipe_resource_reference(&idct->destination, dst);
564
565 idct->max_blocks =
566 align(idct->destination->width0, BLOCK_WIDTH) / BLOCK_WIDTH *
567 align(idct->destination->height0, BLOCK_HEIGHT) / BLOCK_HEIGHT *
568 idct->destination->depth0;
569
570 if(!init_buffers(idct))
571 return false;
572
573 if(!init_shaders(idct)) {
574 cleanup_buffers(idct);
575 return false;
576 }
577
578 if(!vl_vb_init(&idct->blocks, idct->max_blocks, 2)) {
579 cleanup_shaders(idct);
580 cleanup_buffers(idct);
581 return false;
582 }
583
584 init_state(idct);
585
586 vl_idct_map_buffers(idct);
587
588 return true;
589 }
590
591 void
592 vl_idct_cleanup(struct vl_idct *idct)
593 {
594 vl_idct_unmap_buffers(idct);
595
596 vl_vb_cleanup(&idct->blocks);
597 cleanup_shaders(idct);
598 cleanup_buffers(idct);
599
600 cleanup_state(idct);
601
602 pipe_resource_reference(&idct->destination, NULL);
603 }
604
605 void
606 vl_idct_map_buffers(struct vl_idct *idct)
607 {
608 assert(idct);
609
610 struct pipe_box rect =
611 {
612 0, 0, 0,
613 idct->textures.individual.source->width0,
614 idct->textures.individual.source->height0,
615 1
616 };
617
618 idct->tex_transfer = idct->pipe->get_transfer
619 (
620 idct->pipe, idct->textures.individual.source,
621 u_subresource(0, 0),
622 PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
623 &rect
624 );
625
626 idct->texels = idct->pipe->transfer_map(idct->pipe, idct->tex_transfer);
627 }
628
629 void
630 vl_idct_add_block(struct vl_idct *idct, unsigned x, unsigned y, short *block)
631 {
632 struct vertex2f v;
633 unsigned tex_pitch;
634 short *texels;
635
636 unsigned i;
637
638 assert(idct);
639
640 tex_pitch = idct->tex_transfer->stride / sizeof(short);
641 texels = idct->texels + y * tex_pitch * BLOCK_HEIGHT + x * BLOCK_WIDTH;
642
643 for (i = 0; i < BLOCK_HEIGHT; ++i)
644 memcpy(texels + i * tex_pitch, block + i * BLOCK_WIDTH, BLOCK_WIDTH * sizeof(short));
645
646 v.x = x;
647 v.y = y;
648 vl_vb_add_block(&idct->blocks, (float*)&v);
649 }
650
651 void
652 vl_idct_unmap_buffers(struct vl_idct *idct)
653 {
654 assert(idct);
655
656 idct->pipe->transfer_unmap(idct->pipe, idct->tex_transfer);
657 idct->pipe->transfer_destroy(idct->pipe, idct->tex_transfer);
658 }
659
660 void
661 vl_idct_flush(struct vl_idct *idct)
662 {
663 struct pipe_transfer *vec_transfer;
664 void *vectors;
665 unsigned num_verts;
666
667 assert(idct);
668
669 vectors = pipe_buffer_map
670 (
671 idct->pipe,
672 idct->vertex_bufs.individual.pos.buffer,
673 PIPE_TRANSFER_WRITE | PIPE_TRANSFER_DISCARD,
674 &vec_transfer
675 );
676
677 num_verts = vl_vb_upload(&idct->blocks, vectors);
678
679 pipe_buffer_unmap(idct->pipe, idct->vertex_bufs.individual.pos.buffer, vec_transfer);
680
681 if(num_verts > 0) {
682
683 idct->pipe->bind_rasterizer_state(idct->pipe, idct->rs_state);
684 idct->pipe->set_vertex_buffers(idct->pipe, 2, idct->vertex_bufs.all);
685 idct->pipe->bind_vertex_elements_state(idct->pipe, idct->vertex_elems_state);
686 idct->pipe->bind_vs_state(idct->pipe, idct->vs);
687
688 /* first stage */
689 idct->pipe->set_framebuffer_state(idct->pipe, &idct->fb_state[0]);
690 idct->pipe->set_viewport_state(idct->pipe, &idct->viewport[0]);
691 idct->pipe->set_fragment_sampler_views(idct->pipe, 2, idct->sampler_views.stage[0]);
692 idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[0]);
693 idct->pipe->bind_fs_state(idct->pipe, idct->matrix_fs);
694 util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts);
695
696 /* second stage */
697 idct->pipe->set_framebuffer_state(idct->pipe, &idct->fb_state[1]);
698 idct->pipe->set_viewport_state(idct->pipe, &idct->viewport[1]);
699 idct->pipe->set_fragment_sampler_views(idct->pipe, 2, idct->sampler_views.stage[1]);
700 idct->pipe->bind_fragment_sampler_states(idct->pipe, 2, idct->samplers.stage[1]);
701 idct->pipe->bind_fs_state(idct->pipe, idct->transpose_fs);
702 util_draw_arrays(idct->pipe, PIPE_PRIM_QUADS, 0, num_verts);
703 }
704 }