595287d3b81663d2e9e3596c9589b291a22d7462
[mesa.git] / src / gallium / auxiliary / util / u_blit.c
1 /**************************************************************************
2 *
3 * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
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 /**
29 * @file
30 * Copy/blit pixel rect between surfaces
31 *
32 * @author Brian Paul
33 */
34
35
36 #include "pipe/p_context.h"
37 #include "util/u_debug.h"
38 #include "pipe/p_defines.h"
39 #include "util/u_inlines.h"
40 #include "pipe/p_shader_tokens.h"
41 #include "pipe/p_state.h"
42
43 #include "util/u_blit.h"
44 #include "util/u_draw_quad.h"
45 #include "util/u_format.h"
46 #include "util/u_math.h"
47 #include "util/u_memory.h"
48 #include "util/u_sampler.h"
49 #include "util/u_texture.h"
50 #include "util/u_simple_shaders.h"
51
52 #include "cso_cache/cso_context.h"
53
54
55 struct blit_state
56 {
57 struct pipe_context *pipe;
58 struct cso_context *cso;
59
60 struct pipe_blend_state blend_write_color;
61 struct pipe_depth_stencil_alpha_state dsa_keep_depthstencil;
62 struct pipe_rasterizer_state rasterizer;
63 struct pipe_sampler_state sampler;
64 struct pipe_viewport_state viewport;
65 struct pipe_vertex_element velem[2];
66
67 void *vs;
68 void *fs[PIPE_MAX_TEXTURE_TYPES][TGSI_WRITEMASK_XYZW + 1];
69
70 struct pipe_resource *vbuf; /**< quad vertices */
71 unsigned vbuf_slot;
72
73 float vertices[4][2][4]; /**< vertex/texcoords for quad */
74 };
75
76
77 /**
78 * Create state object for blit.
79 * Intended to be created once and re-used for many blit() calls.
80 */
81 struct blit_state *
82 util_create_blit(struct pipe_context *pipe, struct cso_context *cso)
83 {
84 struct blit_state *ctx;
85 uint i;
86
87 ctx = CALLOC_STRUCT(blit_state);
88 if (!ctx)
89 return NULL;
90
91 ctx->pipe = pipe;
92 ctx->cso = cso;
93
94 /* disabled blending/masking */
95 ctx->blend_write_color.rt[0].colormask = PIPE_MASK_RGBA;
96
97 /* rasterizer */
98 ctx->rasterizer.cull_face = PIPE_FACE_NONE;
99 ctx->rasterizer.half_pixel_center = 1;
100 ctx->rasterizer.bottom_edge_rule = 1;
101 ctx->rasterizer.depth_clip = 1;
102
103 /* samplers */
104 ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
105 ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
106 ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
107 ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
108 ctx->sampler.min_img_filter = 0; /* set later */
109 ctx->sampler.mag_img_filter = 0; /* set later */
110
111 /* vertex elements state */
112 for (i = 0; i < 2; i++) {
113 ctx->velem[i].src_offset = i * 4 * sizeof(float);
114 ctx->velem[i].instance_divisor = 0;
115 ctx->velem[i].vertex_buffer_index = cso_get_aux_vertex_buffer_slot(cso);
116 ctx->velem[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
117 }
118
119 ctx->vbuf = NULL;
120
121 /* init vertex data that doesn't change */
122 for (i = 0; i < 4; i++) {
123 ctx->vertices[i][0][3] = 1.0f; /* w */
124 ctx->vertices[i][1][3] = 1.0f; /* q */
125 }
126
127 return ctx;
128 }
129
130
131 /**
132 * Destroy a blit context
133 */
134 void
135 util_destroy_blit(struct blit_state *ctx)
136 {
137 struct pipe_context *pipe = ctx->pipe;
138 unsigned i, j;
139
140 if (ctx->vs)
141 pipe->delete_vs_state(pipe, ctx->vs);
142
143 for (i = 0; i < Elements(ctx->fs); i++) {
144 for (j = 0; j < Elements(ctx->fs[i]); j++) {
145 if (ctx->fs[i][j])
146 pipe->delete_fs_state(pipe, ctx->fs[i][j]);
147 }
148 }
149
150 pipe_resource_reference(&ctx->vbuf, NULL);
151
152 FREE(ctx);
153 }
154
155
156 /**
157 * Helper function to set the fragment shaders.
158 */
159 static INLINE void
160 set_fragment_shader(struct blit_state *ctx, uint writemask,
161 enum pipe_texture_target pipe_tex)
162 {
163 if (!ctx->fs[pipe_tex][writemask]) {
164 unsigned tgsi_tex = util_pipe_tex_to_tgsi_tex(pipe_tex, 0);
165
166 ctx->fs[pipe_tex][writemask] =
167 util_make_fragment_tex_shader_writemask(ctx->pipe, tgsi_tex,
168 TGSI_INTERPOLATE_LINEAR,
169 writemask);
170 }
171
172 cso_set_fragment_shader_handle(ctx->cso, ctx->fs[pipe_tex][writemask]);
173 }
174
175
176 /**
177 * Helper function to set the vertex shader.
178 */
179 static INLINE void
180 set_vertex_shader(struct blit_state *ctx)
181 {
182 /* vertex shader - still required to provide the linkage between
183 * fragment shader input semantics and vertex_element/buffers.
184 */
185 if (!ctx->vs) {
186 const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
187 TGSI_SEMANTIC_GENERIC };
188 const uint semantic_indexes[] = { 0, 0 };
189 ctx->vs = util_make_vertex_passthrough_shader(ctx->pipe, 2,
190 semantic_names,
191 semantic_indexes);
192 }
193
194 cso_set_vertex_shader_handle(ctx->cso, ctx->vs);
195 }
196
197
198 /**
199 * Get offset of next free slot in vertex buffer for quad vertices.
200 */
201 static unsigned
202 get_next_slot( struct blit_state *ctx )
203 {
204 const unsigned max_slots = 4096 / sizeof ctx->vertices;
205
206 if (ctx->vbuf_slot >= max_slots) {
207 pipe_resource_reference(&ctx->vbuf, NULL);
208 ctx->vbuf_slot = 0;
209 }
210
211 if (!ctx->vbuf) {
212 ctx->vbuf = pipe_buffer_create(ctx->pipe->screen,
213 PIPE_BIND_VERTEX_BUFFER,
214 PIPE_USAGE_STREAM,
215 max_slots * sizeof ctx->vertices);
216 }
217
218 return ctx->vbuf_slot++ * sizeof ctx->vertices;
219 }
220
221
222
223
224 /**
225 * Setup vertex data for the textured quad we'll draw.
226 * Note: y=0=top
227 *
228 * FIXME: We should call util_map_texcoords2d_onto_cubemap
229 * for cubemaps.
230 */
231 static unsigned
232 setup_vertex_data_tex(struct blit_state *ctx,
233 unsigned src_target,
234 unsigned src_face,
235 float x0, float y0, float x1, float y1,
236 float s0, float t0, float s1, float t1,
237 float z)
238 {
239 unsigned offset;
240
241 ctx->vertices[0][0][0] = x0;
242 ctx->vertices[0][0][1] = y0;
243 ctx->vertices[0][0][2] = z;
244 ctx->vertices[0][1][0] = s0; /*s*/
245 ctx->vertices[0][1][1] = t0; /*t*/
246 ctx->vertices[0][1][2] = 0; /*r*/
247
248 ctx->vertices[1][0][0] = x1;
249 ctx->vertices[1][0][1] = y0;
250 ctx->vertices[1][0][2] = z;
251 ctx->vertices[1][1][0] = s1; /*s*/
252 ctx->vertices[1][1][1] = t0; /*t*/
253 ctx->vertices[1][1][2] = 0; /*r*/
254
255 ctx->vertices[2][0][0] = x1;
256 ctx->vertices[2][0][1] = y1;
257 ctx->vertices[2][0][2] = z;
258 ctx->vertices[2][1][0] = s1;
259 ctx->vertices[2][1][1] = t1;
260 ctx->vertices[3][1][2] = 0;
261
262 ctx->vertices[3][0][0] = x0;
263 ctx->vertices[3][0][1] = y1;
264 ctx->vertices[3][0][2] = z;
265 ctx->vertices[3][1][0] = s0;
266 ctx->vertices[3][1][1] = t1;
267 ctx->vertices[3][1][2] = 0;
268
269 if (src_target == PIPE_TEXTURE_CUBE ||
270 src_target == PIPE_TEXTURE_CUBE_ARRAY) {
271 /* Map cubemap texture coordinates inplace. */
272 const unsigned stride = sizeof ctx->vertices[0] / sizeof ctx->vertices[0][0][0];
273 util_map_texcoords2d_onto_cubemap(src_face,
274 &ctx->vertices[0][1][0], stride,
275 &ctx->vertices[0][1][0], stride,
276 TRUE);
277 }
278
279 offset = get_next_slot( ctx );
280
281 if (ctx->vbuf) {
282 pipe_buffer_write_nooverlap(ctx->pipe, ctx->vbuf,
283 offset, sizeof(ctx->vertices), ctx->vertices);
284 }
285
286 return offset;
287 }
288
289
290 /**
291 * \return TRUE if two regions overlap, FALSE otherwise
292 */
293 static boolean
294 regions_overlap(int srcX0, int srcY0,
295 int srcX1, int srcY1,
296 int dstX0, int dstY0,
297 int dstX1, int dstY1)
298 {
299 if (MAX2(srcX0, srcX1) < MIN2(dstX0, dstX1))
300 return FALSE; /* src completely left of dst */
301
302 if (MAX2(dstX0, dstX1) < MIN2(srcX0, srcX1))
303 return FALSE; /* dst completely left of src */
304
305 if (MAX2(srcY0, srcY1) < MIN2(dstY0, dstY1))
306 return FALSE; /* src completely above dst */
307
308 if (MAX2(dstY0, dstY1) < MIN2(srcY0, srcY1))
309 return FALSE; /* dst completely above src */
310
311 return TRUE; /* some overlap */
312 }
313
314
315 /**
316 * Can we blit from src format to dest format with a simple copy?
317 */
318 static boolean
319 formats_compatible(enum pipe_format src_format,
320 enum pipe_format dst_format)
321 {
322 if (src_format == dst_format) {
323 return TRUE;
324 }
325 else {
326 const struct util_format_description *src_desc =
327 util_format_description(src_format);
328 const struct util_format_description *dst_desc =
329 util_format_description(dst_format);
330 return util_is_format_compatible(src_desc, dst_desc);
331 }
332 }
333
334
335 /**
336 * Copy pixel block from src surface to dst surface.
337 * Overlapping regions are acceptable.
338 * Flipping and stretching are supported.
339 * \param filter one of PIPE_TEX_MIPFILTER_NEAREST/LINEAR
340 * \param writemask controls which channels in the dest surface are sourced
341 * from the src surface. Disabled channels are sourced
342 * from (0,0,0,1).
343 */
344 void
345 util_blit_pixels(struct blit_state *ctx,
346 struct pipe_resource *src_tex,
347 unsigned src_level,
348 int srcX0, int srcY0,
349 int srcX1, int srcY1,
350 int srcZ0,
351 struct pipe_surface *dst,
352 int dstX0, int dstY0,
353 int dstX1, int dstY1,
354 float z, uint filter,
355 uint writemask, uint zs_writemask)
356 {
357 struct pipe_context *pipe = ctx->pipe;
358 enum pipe_format src_format, dst_format;
359 const int srcW = abs(srcX1 - srcX0);
360 const int srcH = abs(srcY1 - srcY0);
361 boolean overlap;
362 boolean is_stencil, is_depth, blit_depth, blit_stencil;
363 const struct util_format_description *src_desc =
364 util_format_description(src_tex->format);
365 struct pipe_blit_info info;
366
367 assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
368 filter == PIPE_TEX_MIPFILTER_LINEAR);
369
370 assert(src_level <= src_tex->last_level);
371
372 /* do the regions overlap? */
373 overlap = src_tex == dst->texture &&
374 dst->u.tex.level == src_level &&
375 dst->u.tex.first_layer == srcZ0 &&
376 regions_overlap(srcX0, srcY0, srcX1, srcY1,
377 dstX0, dstY0, dstX1, dstY1);
378
379 src_format = util_format_linear(src_tex->format);
380 dst_format = util_format_linear(dst->texture->format);
381
382 /* See whether we will blit depth or stencil. */
383 is_depth = util_format_has_depth(src_desc);
384 is_stencil = util_format_has_stencil(src_desc);
385
386 blit_depth = is_depth && (zs_writemask & BLIT_WRITEMASK_Z);
387 blit_stencil = is_stencil && (zs_writemask & BLIT_WRITEMASK_STENCIL);
388
389 assert((writemask && !zs_writemask && !is_depth && !is_stencil) ||
390 (!writemask && (blit_depth || blit_stencil)));
391
392 /*
393 * XXX: z parameter is deprecated. dst->u.tex.first_layer
394 * specificies the destination layer.
395 */
396 assert(z == 0.0f);
397
398 /*
399 * Check for simple case: no format conversion, no flipping, no stretching,
400 * no overlapping, same number of samples.
401 * Filter mode should not matter since there's no stretching.
402 */
403 if (formats_compatible(src_format, dst_format) &&
404 src_tex->nr_samples == dst->texture->nr_samples &&
405 is_stencil == blit_stencil &&
406 is_depth == blit_depth &&
407 srcX0 < srcX1 &&
408 dstX0 < dstX1 &&
409 srcY0 < srcY1 &&
410 dstY0 < dstY1 &&
411 (dstX1 - dstX0) == (srcX1 - srcX0) &&
412 (dstY1 - dstY0) == (srcY1 - srcY0) &&
413 !overlap) {
414 struct pipe_box src_box;
415 src_box.x = srcX0;
416 src_box.y = srcY0;
417 src_box.z = srcZ0;
418 src_box.width = srcW;
419 src_box.height = srcH;
420 src_box.depth = 1;
421 pipe->resource_copy_region(pipe,
422 dst->texture, dst->u.tex.level,
423 dstX0, dstY0, dst->u.tex.first_layer,/* dest */
424 src_tex, src_level,
425 &src_box);
426 return;
427 }
428
429 memset(&info, 0, sizeof info);
430 info.dst.resource = dst->texture;
431 info.dst.level = dst->u.tex.level;
432 info.dst.box.x = dstX0;
433 info.dst.box.y = dstY0;
434 info.dst.box.z = dst->u.tex.first_layer;
435 info.dst.box.width = dstX1 - dstX0;
436 info.dst.box.height = dstY1 - dstY0;
437 assert(info.dst.box.width >= 0);
438 assert(info.dst.box.height >= 0);
439 info.dst.box.depth = 1;
440 info.dst.format = dst->texture->format;
441 info.src.resource = src_tex;
442 info.src.level = src_level;
443 info.src.box.x = srcX0;
444 info.src.box.y = srcY0;
445 info.src.box.z = srcZ0;
446 info.src.box.width = srcX1 - srcX0;
447 info.src.box.height = srcY1 - srcY0;
448 info.src.box.depth = 1;
449 info.src.format = src_tex->format;
450 info.mask = writemask | (zs_writemask << 4);
451 info.filter = filter;
452 info.scissor_enable = 0;
453
454 pipe->blit(pipe, &info);
455 }
456
457
458 /**
459 * Copy pixel block from src sampler view to dst surface.
460 *
461 * The sampler view's first_level field indicates the source
462 * mipmap level to use.
463 *
464 * The sampler view's first_layer indicate the layer to use, but for
465 * cube maps it must point to the first face. Face is passed in src_face.
466 *
467 * The main advantage over util_blit_pixels is that it allows to specify swizzles in
468 * pipe_sampler_view::swizzle_?.
469 *
470 * But there is no control over blitting Z and/or stencil.
471 */
472 void
473 util_blit_pixels_tex(struct blit_state *ctx,
474 struct pipe_sampler_view *src_sampler_view,
475 int srcX0, int srcY0,
476 int srcX1, int srcY1,
477 unsigned src_face,
478 struct pipe_surface *dst,
479 int dstX0, int dstY0,
480 int dstX1, int dstY1,
481 float z, uint filter)
482 {
483 boolean normalized = src_sampler_view->texture->target != PIPE_TEXTURE_RECT;
484 struct pipe_framebuffer_state fb;
485 float s0, t0, s1, t1;
486 unsigned offset;
487 struct pipe_resource *tex = src_sampler_view->texture;
488
489 assert(filter == PIPE_TEX_MIPFILTER_NEAREST ||
490 filter == PIPE_TEX_MIPFILTER_LINEAR);
491
492 assert(tex);
493 assert(tex->width0 != 0);
494 assert(tex->height0 != 0);
495
496 s0 = (float) srcX0;
497 s1 = (float) srcX1;
498 t0 = (float) srcY0;
499 t1 = (float) srcY1;
500
501 if(normalized)
502 {
503 /* normalize according to the mipmap level's size */
504 int level = src_sampler_view->u.tex.first_level;
505 float w = (float) u_minify(tex->width0, level);
506 float h = (float) u_minify(tex->height0, level);
507 s0 /= w;
508 s1 /= w;
509 t0 /= h;
510 t1 /= h;
511 }
512
513 assert(ctx->pipe->screen->is_format_supported(ctx->pipe->screen, dst->format,
514 PIPE_TEXTURE_2D,
515 dst->texture->nr_samples,
516 PIPE_BIND_RENDER_TARGET));
517
518 /* save state (restored below) */
519 cso_save_blend(ctx->cso);
520 cso_save_depth_stencil_alpha(ctx->cso);
521 cso_save_rasterizer(ctx->cso);
522 cso_save_sample_mask(ctx->cso);
523 cso_save_samplers(ctx->cso, PIPE_SHADER_FRAGMENT);
524 cso_save_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT);
525 cso_save_stream_outputs(ctx->cso);
526 cso_save_viewport(ctx->cso);
527 cso_save_framebuffer(ctx->cso);
528 cso_save_fragment_shader(ctx->cso);
529 cso_save_vertex_shader(ctx->cso);
530 cso_save_geometry_shader(ctx->cso);
531 cso_save_vertex_elements(ctx->cso);
532 cso_save_aux_vertex_buffer_slot(ctx->cso);
533
534 /* set misc state we care about */
535 cso_set_blend(ctx->cso, &ctx->blend_write_color);
536 cso_set_depth_stencil_alpha(ctx->cso, &ctx->dsa_keep_depthstencil);
537 cso_set_sample_mask(ctx->cso, ~0);
538 cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
539 cso_set_vertex_elements(ctx->cso, 2, ctx->velem);
540 cso_set_stream_outputs(ctx->cso, 0, NULL, 0);
541
542 /* sampler */
543 ctx->sampler.normalized_coords = normalized;
544 ctx->sampler.min_img_filter = filter;
545 ctx->sampler.mag_img_filter = filter;
546 cso_single_sampler(ctx->cso, PIPE_SHADER_FRAGMENT, 0, &ctx->sampler);
547 cso_single_sampler_done(ctx->cso, PIPE_SHADER_FRAGMENT);
548
549 /* viewport */
550 ctx->viewport.scale[0] = 0.5f * dst->width;
551 ctx->viewport.scale[1] = 0.5f * dst->height;
552 ctx->viewport.scale[2] = 0.5f;
553 ctx->viewport.scale[3] = 1.0f;
554 ctx->viewport.translate[0] = 0.5f * dst->width;
555 ctx->viewport.translate[1] = 0.5f * dst->height;
556 ctx->viewport.translate[2] = 0.5f;
557 ctx->viewport.translate[3] = 0.0f;
558 cso_set_viewport(ctx->cso, &ctx->viewport);
559
560 /* texture */
561 cso_set_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT, 1, &src_sampler_view);
562
563 /* shaders */
564 set_fragment_shader(ctx, TGSI_WRITEMASK_XYZW,
565 src_sampler_view->texture->target);
566 set_vertex_shader(ctx);
567 cso_set_geometry_shader_handle(ctx->cso, NULL);
568
569 /* drawing dest */
570 memset(&fb, 0, sizeof(fb));
571 fb.width = dst->width;
572 fb.height = dst->height;
573 fb.nr_cbufs = 1;
574 fb.cbufs[0] = dst;
575 cso_set_framebuffer(ctx->cso, &fb);
576
577 /* draw quad */
578 offset = setup_vertex_data_tex(ctx,
579 src_sampler_view->texture->target,
580 src_face,
581 (float) dstX0 / dst->width * 2.0f - 1.0f,
582 (float) dstY0 / dst->height * 2.0f - 1.0f,
583 (float) dstX1 / dst->width * 2.0f - 1.0f,
584 (float) dstY1 / dst->height * 2.0f - 1.0f,
585 s0, t0, s1, t1,
586 z);
587
588 util_draw_vertex_buffer(ctx->pipe, ctx->cso, ctx->vbuf,
589 cso_get_aux_vertex_buffer_slot(ctx->cso),
590 offset,
591 PIPE_PRIM_TRIANGLE_FAN,
592 4, /* verts */
593 2); /* attribs/vert */
594
595 /* restore state we changed */
596 cso_restore_blend(ctx->cso);
597 cso_restore_depth_stencil_alpha(ctx->cso);
598 cso_restore_rasterizer(ctx->cso);
599 cso_restore_sample_mask(ctx->cso);
600 cso_restore_samplers(ctx->cso, PIPE_SHADER_FRAGMENT);
601 cso_restore_sampler_views(ctx->cso, PIPE_SHADER_FRAGMENT);
602 cso_restore_viewport(ctx->cso);
603 cso_restore_framebuffer(ctx->cso);
604 cso_restore_fragment_shader(ctx->cso);
605 cso_restore_vertex_shader(ctx->cso);
606 cso_restore_geometry_shader(ctx->cso);
607 cso_restore_vertex_elements(ctx->cso);
608 cso_restore_aux_vertex_buffer_slot(ctx->cso);
609 cso_restore_stream_outputs(ctx->cso);
610 }