1 /**************************************************************************
2 *
3 * Copyright 2007 VMware, Inc.
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 VMWARE 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 * Authors:
30 * Brian Paul
31 */
32
33 #include "main/errors.h"
34
35 #include "main/image.h"
36 #include "main/bufferobj.h"
37 #include "main/blit.h"
38 #include "main/format_pack.h"
39 #include "main/framebuffer.h"
40 #include "main/macros.h"
41 #include "main/mtypes.h"
42 #include "main/pack.h"
43 #include "main/pbo.h"
44 #include "main/readpix.h"
45 #include "main/state.h"
46 #include "main/teximage.h"
47 #include "main/texstore.h"
48 #include "main/glformats.h"
49 #include "program/program.h"
50 #include "program/prog_print.h"
51 #include "program/prog_instruction.h"
52
53 #include "st_atom.h"
54 #include "st_atom_constbuf.h"
55 #include "st_cb_bitmap.h"
56 #include "st_cb_drawpixels.h"
57 #include "st_context.h"
58 #include "st_debug.h"
59 #include "st_draw.h"
60 #include "st_format.h"
61 #include "st_program.h"
62 #include "st_sampler_view.h"
63 #include "st_scissor.h"
64 #include "st_texture.h"
65 #include "st_util.h"
66 #include "st_nir.h"
67
68 #include "pipe/p_context.h"
69 #include "pipe/p_defines.h"
70 #include "util/format/u_format.h"
71 #include "util/u_inlines.h"
72 #include "util/u_math.h"
73 #include "util/u_simple_shaders.h"
74 #include "util/u_tile.h"
75 #include "cso_cache/cso_context.h"
76
77 #include "compiler/nir/nir_builder.h"
78
79 /**
80 * We have a simple glDrawPixels cache to try to optimize the case where the
81 * same image is drawn over and over again. It basically works as follows:
82 *
83 * 1. After we construct a texture map with the image and draw it, we do
84 * not discard the texture. We keep it around, plus we note the
85 * glDrawPixels width, height, format, etc. parameters and keep a copy
86 * of the image in a malloc'd buffer.
87 *
88 * 2. On the next glDrawPixels we check if the parameters match the previous
89 * call. If those match, we check if the image matches the previous image
90 * via a memcmp() call. If everything matches, we re-use the previous
91 * texture, thereby avoiding the cost creating a new texture and copying
92 * the image to it.
93 *
94 * The effectiveness of this cache depends upon:
95 * 1. If the memcmp() finds a difference, it happens relatively quickly.
96 Hopefully, not just the last pixels differ!
97 * 2. If the memcmp() finds no difference, doing that check is faster than
98 * creating and loading a texture.
99 *
100 * Notes:
101 * 1. We don't support any pixel unpacking parameters.
102 * 2. We don't try to cache images in Pixel Buffer Objects.
103 * 3. Instead of saving the whole image, perhaps some sort of reliable
104 * checksum function could be used instead.
105 */
106 #define USE_DRAWPIXELS_CACHE 1
107
108 static nir_ssa_def *
sample_via_nir(nir_builder * b,nir_variable * texcoord,const char * name,int sampler,enum glsl_base_type base_type,nir_alu_type alu_type)109 sample_via_nir(nir_builder *b, nir_variable *texcoord,
110 const char *name, int sampler, enum glsl_base_type base_type,
111 nir_alu_type alu_type)
112 {
113 const struct glsl_type *sampler2D =
114 glsl_sampler_type(GLSL_SAMPLER_DIM_2D, false, false, base_type);
115
116 nir_variable *var =
117 nir_variable_create(b->shader, nir_var_uniform, sampler2D, name);
118 var->data.binding = sampler;
119 var->data.explicit_binding = true;
120
121 nir_deref_instr *deref = nir_build_deref_var(b, var);
122
123 nir_tex_instr *tex = nir_tex_instr_create(b->shader, 3);
124 tex->op = nir_texop_tex;
125 tex->sampler_dim = GLSL_SAMPLER_DIM_2D;
126 tex->coord_components = 2;
127 tex->dest_type = alu_type;
128 tex->src[0].src_type = nir_tex_src_texture_deref;
129 tex->src[0].src = nir_src_for_ssa(&deref->dest.ssa);
130 tex->src[1].src_type = nir_tex_src_sampler_deref;
131 tex->src[1].src = nir_src_for_ssa(&deref->dest.ssa);
132 tex->src[2].src_type = nir_tex_src_coord;
133 tex->src[2].src =
134 nir_src_for_ssa(nir_channels(b, nir_load_var(b, texcoord),
135 (1 << tex->coord_components) - 1));
136
137 nir_ssa_dest_init(&tex->instr, &tex->dest, 4, 32, NULL);
138 nir_builder_instr_insert(b, &tex->instr);
139 return nir_channel(b, &tex->dest.ssa, 0);
140 }
141
142 static void *
make_drawpix_z_stencil_program_nir(struct st_context * st,bool write_depth,bool write_stencil)143 make_drawpix_z_stencil_program_nir(struct st_context *st,
144 bool write_depth,
145 bool write_stencil)
146 {
147 const nir_shader_compiler_options *options =
148 st_get_nir_compiler_options(st, MESA_SHADER_FRAGMENT);
149
150 nir_builder b = nir_builder_init_simple_shader(MESA_SHADER_FRAGMENT, options,
151 "drawpixels %s%s",
152 write_depth ? "Z" : "",
153 write_stencil ? "S" : "");
154
155 nir_variable *texcoord =
156 nir_variable_create(b.shader, nir_var_shader_in, glsl_vec_type(2),
157 "texcoord");
158 texcoord->data.location = VARYING_SLOT_TEX0;
159
160 if (write_depth) {
161 nir_variable *out =
162 nir_variable_create(b.shader, nir_var_shader_out, glsl_float_type(),
163 "gl_FragDepth");
164 out->data.location = FRAG_RESULT_DEPTH;
165 nir_ssa_def *depth = sample_via_nir(&b, texcoord, "depth", 0,
166 GLSL_TYPE_FLOAT, nir_type_float32);
167 nir_store_var(&b, out, depth, 0x1);
168
169 /* Also copy color */
170 nir_variable *color_in =
171 nir_variable_create(b.shader, nir_var_shader_in, glsl_vec_type(4),
172 "v_color");
173 color_in->data.location = VARYING_SLOT_COL0;
174
175 nir_variable *color_out =
176 nir_variable_create(b.shader, nir_var_shader_out, glsl_vec_type(4),
177 "gl_FragColor");
178 color_out->data.location = FRAG_RESULT_COLOR;
179 nir_copy_var(&b, color_out, color_in);
180 }
181
182 if (write_stencil) {
183 nir_variable *out =
184 nir_variable_create(b.shader, nir_var_shader_out, glsl_uint_type(),
185 "gl_FragStencilRefARB");
186 out->data.location = FRAG_RESULT_STENCIL;
187 nir_ssa_def *stencil = sample_via_nir(&b, texcoord, "stencil", 1,
188 GLSL_TYPE_UINT, nir_type_uint32);
189 nir_store_var(&b, out, stencil, 0x1);
190 }
191
192 return st_nir_finish_builtin_shader(st, b.shader);
193 }
194
195 static void *
make_drawpix_zs_to_color_program_nir(struct st_context * st,bool rgba)196 make_drawpix_zs_to_color_program_nir(struct st_context *st,
197 bool rgba)
198 {
199 const nir_shader_compiler_options *options =
200 st_get_nir_compiler_options(st, MESA_SHADER_FRAGMENT);
201
202 nir_builder b = nir_builder_init_simple_shader(MESA_SHADER_FRAGMENT, options,
203 "copypixels ZStoC");
204
205 nir_variable *texcoord =
206 nir_variable_create(b.shader, nir_var_shader_in, glsl_vec_type(2),
207 "texcoord");
208 texcoord->data.location = VARYING_SLOT_TEX0;
209
210 /* Sample depth and stencil */
211 nir_ssa_def *depth = sample_via_nir(&b, texcoord, "depth", 0,
212 GLSL_TYPE_FLOAT, nir_type_float32);
213 nir_ssa_def *stencil = sample_via_nir(&b, texcoord, "stencil", 1,
214 GLSL_TYPE_UINT, nir_type_uint32);
215
216 /* Create the variable to store the output color */
217 nir_variable *color_out =
218 nir_variable_create(b.shader, nir_var_shader_out, glsl_vec_type(4),
219 "make_drawpix_zs_to_color_program_nirgl_FragColor");
220 color_out->data.location = FRAG_RESULT_COLOR;
221
222 nir_ssa_def *shifted_depth = nir_fmul(&b,nir_f2f64(&b, depth), nir_imm_double(&b,0xffffff));
223 nir_ssa_def *int_depth = nir_f2u32(&b,shifted_depth);
224
225 nir_ssa_def *ds[4];
226 ds[0] = nir_ubitfield_extract(&b, stencil, nir_imm_int(&b, 0), nir_imm_int(&b,8));
227 ds[1] = nir_ubitfield_extract(&b, int_depth, nir_imm_int(&b, 0), nir_imm_int(&b,8));
228 ds[2] = nir_ubitfield_extract(&b, int_depth, nir_imm_int(&b, 8), nir_imm_int(&b,8));
229 ds[3] = nir_ubitfield_extract(&b, int_depth, nir_imm_int(&b, 16), nir_imm_int(&b,8));
230
231 nir_ssa_def *ds_comp[4];
232 ds_comp[0] = nir_fsat(&b, nir_fmul_imm(&b, nir_u2f32(&b, ds[3]), 1.0/255.0));
233 ds_comp[1] = nir_fsat(&b, nir_fmul_imm(&b, nir_u2f32(&b, ds[2]), 1.0/255.0));
234 ds_comp[2] = nir_fsat(&b, nir_fmul_imm(&b, nir_u2f32(&b, ds[1]), 1.0/255.0));
235 ds_comp[3] = nir_fsat(&b, nir_fmul_imm(&b, nir_u2f32(&b, ds[0]), 1.0/255.0));
236
237 nir_ssa_def *unpacked_ds = nir_vec4(&b, ds_comp[0], ds_comp[1], ds_comp[2], ds_comp[3]);
238
239 if (rgba) {
240 nir_store_var(&b, color_out, unpacked_ds, 0xf);
241 }
242 else {
243 unsigned zyxw[4] = { 2, 1, 0, 3 };
244 nir_ssa_def *swizzled_ds= nir_swizzle(&b, unpacked_ds, zyxw, 4);
245 nir_store_var(&b, color_out, swizzled_ds, 0xf);
246 }
247
248 return st_nir_finish_builtin_shader(st, b.shader);
249 }
250
251
252 /**
253 * Create fragment program that does a TEX() instruction to get a Z and/or
254 * stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL.
255 * Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX).
256 * Pass fragment color through as-is.
257 *
258 * \return CSO of the fragment shader.
259 */
260 static void *
get_drawpix_z_stencil_program(struct st_context * st,bool write_depth,bool write_stencil)261 get_drawpix_z_stencil_program(struct st_context *st,
262 bool write_depth,
263 bool write_stencil)
264 {
265 const GLuint shaderIndex = write_depth * 2 + write_stencil;
266 void *cso;
267
268 assert(shaderIndex < ARRAY_SIZE(st->drawpix.zs_shaders));
269
270 if (st->drawpix.zs_shaders[shaderIndex]) {
271 /* already have the proper shader */
272 return st->drawpix.zs_shaders[shaderIndex];
273 }
274
275 cso = make_drawpix_z_stencil_program_nir(st, write_depth, write_stencil);
276
277 /* save the new shader */
278 st->drawpix.zs_shaders[shaderIndex] = cso;
279 return cso;
280 }
281
282 /**
283 * Create fragment program that does a TEX() instruction to get a Z and
284 * stencil value value, then writes to FRAG_RESULT_COLOR.
285 * Used for glCopyPixels(GL_DEPTH_STENCIL_TO_RGBA_NV / GL_DEPTH_STENCIL_TO_BGRA_NV).
286 *
287 * \return CSO of the fragment shader.
288 */
289 static void *
get_drawpix_zs_to_color_program(struct st_context * st,bool rgba)290 get_drawpix_zs_to_color_program(struct st_context *st,
291 bool rgba)
292 {
293 void *cso;
294 GLuint shaderIndex;
295
296 if (rgba)
297 shaderIndex = 4;
298 else
299 shaderIndex = 5;
300
301 assert(shaderIndex < ARRAY_SIZE(st->drawpix.zs_shaders));
302
303 if (st->drawpix.zs_shaders[shaderIndex]) {
304 /* already have the proper shader */
305 return st->drawpix.zs_shaders[shaderIndex];
306 }
307
308 cso = make_drawpix_zs_to_color_program_nir(st, rgba);
309
310 /* save the new shader */
311 st->drawpix.zs_shaders[shaderIndex] = cso;
312 return cso;
313 }
314
315 /**
316 * Create a simple vertex shader that just passes through the
317 * vertex position, texcoord, and color.
318 */
319 void
st_make_passthrough_vertex_shader(struct st_context * st)320 st_make_passthrough_vertex_shader(struct st_context *st)
321 {
322 if (st->passthrough_vs)
323 return;
324
325 unsigned inputs[] =
326 { VERT_ATTRIB_POS, VERT_ATTRIB_COLOR0, VERT_ATTRIB_GENERIC0 };
327 unsigned outputs[] =
328 { VARYING_SLOT_POS, VARYING_SLOT_COL0, VARYING_SLOT_TEX0 };
329
330 st->passthrough_vs =
331 st_nir_make_passthrough_shader(st, "drawpixels VS",
332 MESA_SHADER_VERTEX, 3,
333 inputs, outputs, NULL, 0);
334 }
335
336
337 /**
338 * Return a texture internalFormat for drawing/copying an image
339 * of the given format and type.
340 */
341 static GLenum
internal_format(struct gl_context * ctx,GLenum format,GLenum type)342 internal_format(struct gl_context *ctx, GLenum format, GLenum type)
343 {
344 switch (format) {
345 case GL_DEPTH_COMPONENT:
346 switch (type) {
347 case GL_UNSIGNED_SHORT:
348 return GL_DEPTH_COMPONENT16;
349
350 case GL_UNSIGNED_INT:
351 return GL_DEPTH_COMPONENT32;
352
353 case GL_FLOAT:
354 if (ctx->Extensions.ARB_depth_buffer_float)
355 return GL_DEPTH_COMPONENT32F;
356 else
357 return GL_DEPTH_COMPONENT;
358
359 default:
360 return GL_DEPTH_COMPONENT;
361 }
362
363 case GL_DEPTH_STENCIL:
364 switch (type) {
365 case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
366 return GL_DEPTH32F_STENCIL8;
367
368 case GL_UNSIGNED_INT_24_8:
369 default:
370 return GL_DEPTH24_STENCIL8;
371 }
372
373 case GL_STENCIL_INDEX:
374 return GL_STENCIL_INDEX;
375
376 default:
377 if (_mesa_is_enum_format_integer(format)) {
378 switch (type) {
379 case GL_BYTE:
380 return GL_RGBA8I;
381 case GL_UNSIGNED_BYTE:
382 return GL_RGBA8UI;
383 case GL_SHORT:
384 return GL_RGBA16I;
385 case GL_UNSIGNED_SHORT:
386 return GL_RGBA16UI;
387 case GL_INT:
388 return GL_RGBA32I;
389 case GL_UNSIGNED_INT:
390 return GL_RGBA32UI;
391 default:
392 assert(0 && "Unexpected type in internal_format()");
393 return GL_RGBA_INTEGER;
394 }
395 }
396 else {
397 switch (type) {
398 case GL_UNSIGNED_BYTE:
399 case GL_UNSIGNED_INT_8_8_8_8:
400 case GL_UNSIGNED_INT_8_8_8_8_REV:
401 default:
402 return GL_RGBA8;
403
404 case GL_UNSIGNED_BYTE_3_3_2:
405 case GL_UNSIGNED_BYTE_2_3_3_REV:
406 return GL_R3_G3_B2;
407
408 case GL_UNSIGNED_SHORT_4_4_4_4:
409 case GL_UNSIGNED_SHORT_4_4_4_4_REV:
410 return GL_RGBA4;
411
412 case GL_UNSIGNED_SHORT_5_6_5:
413 case GL_UNSIGNED_SHORT_5_6_5_REV:
414 return GL_RGB565;
415
416 case GL_UNSIGNED_SHORT_5_5_5_1:
417 case GL_UNSIGNED_SHORT_1_5_5_5_REV:
418 return GL_RGB5_A1;
419
420 case GL_UNSIGNED_INT_10_10_10_2:
421 case GL_UNSIGNED_INT_2_10_10_10_REV:
422 return GL_RGB10_A2;
423
424 case GL_UNSIGNED_SHORT:
425 case GL_UNSIGNED_INT:
426 return GL_RGBA16;
427
428 case GL_BYTE:
429 return
430 ctx->Extensions.EXT_texture_snorm ? GL_RGBA8_SNORM : GL_RGBA8;
431
432 case GL_SHORT:
433 case GL_INT:
434 return
435 ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
436
437 case GL_HALF_FLOAT_ARB:
438 return
439 ctx->Extensions.ARB_texture_float ? GL_RGBA16F :
440 ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
441
442 case GL_FLOAT:
443 case GL_DOUBLE:
444 return
445 ctx->Extensions.ARB_texture_float ? GL_RGBA32F :
446 ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
447
448 case GL_UNSIGNED_INT_5_9_9_9_REV:
449 assert(ctx->Extensions.EXT_texture_shared_exponent);
450 return GL_RGB9_E5;
451
452 case GL_UNSIGNED_INT_10F_11F_11F_REV:
453 assert(ctx->Extensions.EXT_packed_float);
454 return GL_R11F_G11F_B10F;
455 }
456 }
457 }
458 }
459
460
461 /**
462 * Create a temporary texture to hold an image of the given size.
463 * If width, height are not POT and the driver only handles POT textures,
464 * allocate the next larger size of texture that is POT.
465 */
466 static struct pipe_resource *
alloc_texture(struct st_context * st,GLsizei width,GLsizei height,enum pipe_format texFormat,unsigned bind)467 alloc_texture(struct st_context *st, GLsizei width, GLsizei height,
468 enum pipe_format texFormat, unsigned bind)
469 {
470 struct pipe_resource *pt;
471
472 pt = st_texture_create(st, st->internal_target, texFormat, 0,
473 width, height, 1, 1, 0, bind, false);
474
475 return pt;
476 }
477
478
479 /**
480 * Search the cache for an image which matches the given parameters.
481 * \return pipe_resource pointer if found, NULL if not found.
482 */
483 static struct pipe_resource *
search_drawpixels_cache(struct st_context * st,GLsizei width,GLsizei height,GLenum format,GLenum type,const struct gl_pixelstore_attrib * unpack,const void * pixels)484 search_drawpixels_cache(struct st_context *st,
485 GLsizei width, GLsizei height,
486 GLenum format, GLenum type,
487 const struct gl_pixelstore_attrib *unpack,
488 const void *pixels)
489 {
490 struct pipe_resource *pt = NULL;
491 const GLint bpp = _mesa_bytes_per_pixel(format, type);
492 unsigned i;
493
494 if ((unpack->RowLength != 0 && unpack->RowLength != width) ||
495 unpack->SkipPixels != 0 ||
496 unpack->SkipRows != 0 ||
497 unpack->SwapBytes ||
498 unpack->BufferObj) {
499 /* we don't allow non-default pixel unpacking values */
500 return NULL;
501 }
502
503 /* Search cache entries for a match */
504 for (i = 0; i < ARRAY_SIZE(st->drawpix_cache.entries); i++) {
505 struct drawpix_cache_entry *entry = &st->drawpix_cache.entries[i];
506
507 if (width == entry->width &&
508 height == entry->height &&
509 format == entry->format &&
510 type == entry->type &&
511 pixels == entry->user_pointer &&
512 entry->image) {
513 assert(entry->texture);
514
515 /* check if the pixel data is the same */
516 if (memcmp(pixels, entry->image, width * height * bpp) == 0) {
517 /* Success - found a cache match */
518 pipe_resource_reference(&pt, entry->texture);
519 /* refcount of returned texture should be at least two here. One
520 * reference for the cache to hold on to, one for the caller (which
521 * it will release), and possibly more held by the driver.
522 */
523 assert(pt->reference.count >= 2);
524
525 /* update the age of this entry */
526 entry->age = ++st->drawpix_cache.age;
527
528 return pt;
529 }
530 }
531 }
532
533 /* no cache match found */
534 return NULL;
535 }
536
537
538 /**
539 * Find the oldest entry in the glDrawPixels cache. We'll replace this
540 * one when we need to store a new image.
541 */
542 static struct drawpix_cache_entry *
find_oldest_drawpixels_cache_entry(struct st_context * st)543 find_oldest_drawpixels_cache_entry(struct st_context *st)
544 {
545 unsigned oldest_age = ~0u, oldest_index = ~0u;
546 unsigned i;
547
548 /* Find entry with oldest (lowest) age */
549 for (i = 0; i < ARRAY_SIZE(st->drawpix_cache.entries); i++) {
550 const struct drawpix_cache_entry *entry = &st->drawpix_cache.entries[i];
551 if (entry->age < oldest_age) {
552 oldest_age = entry->age;
553 oldest_index = i;
554 }
555 }
556
557 assert(oldest_index != ~0u);
558
559 return &st->drawpix_cache.entries[oldest_index];
560 }
561
562
563 /**
564 * Try to save the given glDrawPixels image in the cache.
565 */
566 static void
cache_drawpixels_image(struct st_context * st,GLsizei width,GLsizei height,GLenum format,GLenum type,const struct gl_pixelstore_attrib * unpack,const void * pixels,struct pipe_resource * pt)567 cache_drawpixels_image(struct st_context *st,
568 GLsizei width, GLsizei height,
569 GLenum format, GLenum type,
570 const struct gl_pixelstore_attrib *unpack,
571 const void *pixels,
572 struct pipe_resource *pt)
573 {
574 if ((unpack->RowLength == 0 || unpack->RowLength == width) &&
575 unpack->SkipPixels == 0 &&
576 unpack->SkipRows == 0) {
577 const GLint bpp = _mesa_bytes_per_pixel(format, type);
578 struct drawpix_cache_entry *entry =
579 find_oldest_drawpixels_cache_entry(st);
580 assert(entry);
581 entry->width = width;
582 entry->height = height;
583 entry->format = format;
584 entry->type = type;
585 entry->user_pointer = pixels;
586 free(entry->image);
587 entry->image = malloc(width * height * bpp);
588 if (entry->image) {
589 memcpy(entry->image, pixels, width * height * bpp);
590 pipe_resource_reference(&entry->texture, pt);
591 entry->age = ++st->drawpix_cache.age;
592 }
593 else {
594 /* out of memory, free/disable cached texture */
595 entry->width = 0;
596 entry->height = 0;
597 pipe_resource_reference(&entry->texture, NULL);
598 }
599 }
600 }
601
602
603 /**
604 * Make texture containing an image for glDrawPixels image.
605 * If 'pixels' is NULL, leave the texture image data undefined.
606 */
607 static struct pipe_resource *
make_texture(struct st_context * st,GLsizei width,GLsizei height,GLenum format,GLenum type,const struct gl_pixelstore_attrib * unpack,const void * pixels)608 make_texture(struct st_context *st,
609 GLsizei width, GLsizei height, GLenum format, GLenum type,
610 const struct gl_pixelstore_attrib *unpack,
611 const void *pixels)
612 {
613 struct gl_context *ctx = st->ctx;
614 struct pipe_context *pipe = st->pipe;
615 mesa_format mformat;
616 struct pipe_resource *pt = NULL;
617 enum pipe_format pipeFormat;
618 GLenum baseInternalFormat;
619
620 #if USE_DRAWPIXELS_CACHE
621 pt = search_drawpixels_cache(st, width, height, format, type,
622 unpack, pixels);
623 if (pt) {
624 return pt;
625 }
626 #endif
627
628 /* Choose a pixel format for the temp texture which will hold the
629 * image to draw.
630 */
631 pipeFormat = st_choose_matching_format(st, PIPE_BIND_SAMPLER_VIEW,
632 format, type, unpack->SwapBytes);
633
634 if (pipeFormat == PIPE_FORMAT_NONE) {
635 /* Use the generic approach. */
636 GLenum intFormat = internal_format(ctx, format, type);
637
638 pipeFormat = st_choose_format(st, intFormat, format, type,
639 st->internal_target, 0, 0,
640 PIPE_BIND_SAMPLER_VIEW,
641 false, false);
642 assert(pipeFormat != PIPE_FORMAT_NONE);
643 }
644
645 mformat = st_pipe_format_to_mesa_format(pipeFormat);
646 baseInternalFormat = _mesa_get_format_base_format(mformat);
647
648 pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
649 if (!pixels)
650 return NULL;
651
652 /* alloc temporary texture */
653 pt = alloc_texture(st, width, height, pipeFormat, PIPE_BIND_SAMPLER_VIEW);
654 if (!pt) {
655 _mesa_unmap_pbo_source(ctx, unpack);
656 return NULL;
657 }
658
659 {
660 struct pipe_transfer *transfer;
661 GLubyte *dest;
662 const GLbitfield imageTransferStateSave = ctx->_ImageTransferState;
663
664 /* we'll do pixel transfer in a fragment shader */
665 ctx->_ImageTransferState = 0x0;
666
667 /* map texture transfer */
668 dest = pipe_texture_map(pipe, pt, 0, 0,
669 PIPE_MAP_WRITE, 0, 0,
670 width, height, &transfer);
671 if (!dest) {
672 pipe_resource_reference(&pt, NULL);
673 _mesa_unmap_pbo_source(ctx, unpack);
674 return NULL;
675 }
676
677 /* Put image into texture transfer.
678 * Note that the image is actually going to be upside down in
679 * the texture. We deal with that with texcoords.
680 */
681 if ((format == GL_RGBA || format == GL_BGRA)
682 && type == GL_UNSIGNED_BYTE) {
683 /* Use a memcpy-based texstore to avoid software pixel swizzling.
684 * We'll do the necessary swizzling with the pipe_sampler_view to
685 * give much better performance.
686 * XXX in the future, expand this to accomodate more format and
687 * type combinations.
688 */
689 _mesa_memcpy_texture(ctx, 2,
690 mformat, /* mesa_format */
691 transfer->stride, /* dstRowStride, bytes */
692 &dest, /* destSlices */
693 width, height, 1, /* size */
694 format, type, /* src format/type */
695 pixels, /* data source */
696 unpack);
697 }
698 else {
699 ASSERTED bool success;
700 success = _mesa_texstore(ctx, 2, /* dims */
701 baseInternalFormat, /* baseInternalFormat */
702 mformat, /* mesa_format */
703 transfer->stride, /* dstRowStride, bytes */
704 &dest, /* destSlices */
705 width, height, 1, /* size */
706 format, type, /* src format/type */
707 pixels, /* data source */
708 unpack);
709
710 assert(success);
711 }
712
713 /* unmap */
714 pipe_texture_unmap(pipe, transfer);
715
716 /* restore */
717 ctx->_ImageTransferState = imageTransferStateSave;
718 }
719
720 #if USE_DRAWPIXELS_CACHE
721 cache_drawpixels_image(st, width, height, format, type, unpack, pixels, pt);
722 #endif
723
724 _mesa_unmap_pbo_source(ctx, unpack);
725
726 return pt;
727 }
728
729
730 static void
draw_textured_quad(struct gl_context * ctx,GLint x,GLint y,GLfloat z,GLsizei width,GLsizei height,GLfloat zoomX,GLfloat zoomY,struct pipe_sampler_view ** sv,int num_sampler_view,void * driver_vp,void * driver_fp,struct st_fp_variant * fpv,const GLfloat * color,GLboolean invertTex,GLboolean write_depth,GLboolean write_stencil)731 draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
732 GLsizei width, GLsizei height,
733 GLfloat zoomX, GLfloat zoomY,
734 struct pipe_sampler_view **sv,
735 int num_sampler_view,
736 void *driver_vp,
737 void *driver_fp,
738 struct st_fp_variant *fpv,
739 const GLfloat *color,
740 GLboolean invertTex,
741 GLboolean write_depth, GLboolean write_stencil)
742 {
743 struct st_context *st = st_context(ctx);
744 struct pipe_context *pipe = st->pipe;
745 struct cso_context *cso = st->cso_context;
746 const unsigned fb_width = _mesa_geometric_width(ctx->DrawBuffer);
747 const unsigned fb_height = _mesa_geometric_height(ctx->DrawBuffer);
748 GLfloat x0, y0, x1, y1;
749 ASSERTED GLsizei maxSize;
750 boolean normalized = sv[0]->texture->target == PIPE_TEXTURE_2D ||
751 (sv[0]->texture->target == PIPE_TEXTURE_RECT && st->lower_rect_tex);
752 unsigned cso_state_mask;
753
754 assert(sv[0]->texture->target == st->internal_target);
755
756 /* limit checks */
757 /* XXX if DrawPixels image is larger than max texture size, break
758 * it up into chunks.
759 */
760 maxSize = st->screen->get_param(st->screen,
761 PIPE_CAP_MAX_TEXTURE_2D_SIZE);
762 assert(width <= maxSize);
763 assert(height <= maxSize);
764
765 cso_state_mask = (CSO_BIT_RASTERIZER |
766 CSO_BIT_VIEWPORT |
767 CSO_BIT_FRAGMENT_SAMPLERS |
768 CSO_BIT_STREAM_OUTPUTS |
769 CSO_BIT_VERTEX_ELEMENTS |
770 CSO_BITS_ALL_SHADERS);
771 if (write_stencil) {
772 cso_state_mask |= (CSO_BIT_DEPTH_STENCIL_ALPHA |
773 CSO_BIT_BLEND);
774 }
775 cso_save_state(cso, cso_state_mask);
776
777 /* rasterizer state: just scissor */
778 {
779 struct pipe_rasterizer_state rasterizer;
780 memset(&rasterizer, 0, sizeof(rasterizer));
781 rasterizer.clamp_fragment_color = !st->clamp_frag_color_in_shader &&
782 ctx->Color._ClampFragmentColor;
783 rasterizer.half_pixel_center = 1;
784 rasterizer.bottom_edge_rule = 1;
785 rasterizer.depth_clip_near = !ctx->Transform.DepthClampNear;
786 rasterizer.depth_clip_far = !ctx->Transform.DepthClampFar;
787 rasterizer.depth_clamp = !rasterizer.depth_clip_far;
788 rasterizer.scissor = ctx->Scissor.EnableFlags;
789 cso_set_rasterizer(cso, &rasterizer);
790 }
791
792 if (write_stencil) {
793 /* Stencil writing bypasses the normal fragment pipeline to
794 * disable color writing and set stencil test to always pass.
795 */
796 struct pipe_depth_stencil_alpha_state dsa;
797 struct pipe_blend_state blend;
798
799 /* depth/stencil */
800 memset(&dsa, 0, sizeof(dsa));
801 dsa.stencil[0].enabled = 1;
802 dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
803 dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
804 dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
805 if (write_depth) {
806 /* writing depth+stencil: depth test always passes */
807 dsa.depth_enabled = 1;
808 dsa.depth_writemask = ctx->Depth.Mask;
809 dsa.depth_func = PIPE_FUNC_ALWAYS;
810 }
811 cso_set_depth_stencil_alpha(cso, &dsa);
812
813 /* blend (colormask) */
814 memset(&blend, 0, sizeof(blend));
815 cso_set_blend(cso, &blend);
816 }
817
818 /* fragment shader state: TEX lookup program */
819 cso_set_fragment_shader_handle(cso, driver_fp);
820
821 /* vertex shader state: position + texcoord pass-through */
822 cso_set_vertex_shader_handle(cso, driver_vp);
823
824 /* disable other shaders */
825 cso_set_tessctrl_shader_handle(cso, NULL);
826 cso_set_tesseval_shader_handle(cso, NULL);
827 cso_set_geometry_shader_handle(cso, NULL);
828
829 /* user samplers, plus the drawpix samplers */
830 {
831 struct pipe_sampler_state sampler;
832
833 memset(&sampler, 0, sizeof(sampler));
834 sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
835 sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
836 sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
837 sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
838 sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
839 sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
840 sampler.normalized_coords = normalized;
841
842 if (fpv) {
843 /* drawing a color image */
844 const struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
845 uint num = MAX3(fpv->drawpix_sampler + 1,
846 fpv->pixelmap_sampler + 1,
847 st->state.num_frag_samplers);
848 uint i;
849
850 for (i = 0; i < st->state.num_frag_samplers; i++)
851 samplers[i] = &st->state.frag_samplers[i];
852
853 samplers[fpv->drawpix_sampler] = &sampler;
854 if (sv[1])
855 samplers[fpv->pixelmap_sampler] = &sampler;
856
857 cso_set_samplers(cso, PIPE_SHADER_FRAGMENT, num, samplers);
858 } else {
859 /* drawing a depth/stencil image */
860 const struct pipe_sampler_state *samplers[2] = {&sampler, &sampler};
861
862 cso_set_samplers(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, samplers);
863 }
864 }
865
866 unsigned tex_width = sv[0]->texture->width0;
867 unsigned tex_height = sv[0]->texture->height0;
868
869 /* user textures, plus the drawpix textures */
870 if (fpv) {
871 /* drawing a color image */
872 struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
873 unsigned num_views =
874 st_get_sampler_views(st, PIPE_SHADER_FRAGMENT,
875 ctx->FragmentProgram._Current, sampler_views);
876
877 num_views = MAX3(fpv->drawpix_sampler + 1, fpv->pixelmap_sampler + 1,
878 num_views);
879
880 sampler_views[fpv->drawpix_sampler] = sv[0];
881 if (sv[1])
882 sampler_views[fpv->pixelmap_sampler] = sv[1];
883 pipe->set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0, num_views, 0,
884 true, sampler_views);
885 st->state.num_sampler_views[PIPE_SHADER_FRAGMENT] = num_views;
886 } else {
887 /* drawing a depth/stencil image */
888 pipe->set_sampler_views(pipe, PIPE_SHADER_FRAGMENT, 0, num_sampler_view,
889 0, false, sv);
890 st->state.num_sampler_views[PIPE_SHADER_FRAGMENT] =
891 MAX2(st->state.num_sampler_views[PIPE_SHADER_FRAGMENT], num_sampler_view);
892
893 for (unsigned i = 0; i < num_sampler_view; i++)
894 pipe_sampler_view_reference(&sv[i], NULL);
895 }
896
897 /* viewport state: viewport matching window dims */
898 cso_set_viewport_dims(cso, fb_width, fb_height, TRUE);
899
900 st->util_velems.count = 3;
901 cso_set_vertex_elements(cso, &st->util_velems);
902 cso_set_stream_outputs(cso, 0, NULL, NULL);
903
904 /* Compute Gallium window coords (y=0=top) with pixel zoom.
905 * Recall that these coords are transformed by the current
906 * vertex shader and viewport transformation.
907 */
908 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_BOTTOM) {
909 y = fb_height - (int) (y + height * ctx->Pixel.ZoomY);
910 invertTex = !invertTex;
911 }
912
913 x0 = (GLfloat) x;
914 x1 = x + width * ctx->Pixel.ZoomX;
915 y0 = (GLfloat) y;
916 y1 = y + height * ctx->Pixel.ZoomY;
917
918 /* convert Z from [0,1] to [-1,-1] to match viewport Z scale/bias */
919 z = z * 2.0f - 1.0f;
920
921 {
922 const float clip_x0 = x0 / (float) fb_width * 2.0f - 1.0f;
923 const float clip_y0 = y0 / (float) fb_height * 2.0f - 1.0f;
924 const float clip_x1 = x1 / (float) fb_width * 2.0f - 1.0f;
925 const float clip_y1 = y1 / (float) fb_height * 2.0f - 1.0f;
926 const float maxXcoord = normalized ?
927 ((float) width / tex_width) : (float) width;
928 const float maxYcoord = normalized
929 ? ((float) height / tex_height) : (float) height;
930 const float sLeft = 0.0f, sRight = maxXcoord;
931 const float tTop = invertTex ? maxYcoord : 0.0f;
932 const float tBot = invertTex ? 0.0f : maxYcoord;
933
934 if (!st_draw_quad(st, clip_x0, clip_y0, clip_x1, clip_y1, z,
935 sLeft, tBot, sRight, tTop, color, 0)) {
936 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
937 }
938 }
939
940 /* restore state */
941 /* Unbind all because st/mesa won't do it if the current shader doesn't
942 * use them.
943 */
944 cso_restore_state(cso, CSO_UNBIND_FS_SAMPLERVIEWS);
945 st->state.num_sampler_views[PIPE_SHADER_FRAGMENT] = 0;
946
947 ctx->Array.NewVertexElements = true;
948 st->dirty |= ST_NEW_VERTEX_ARRAYS |
949 ST_NEW_FS_SAMPLER_VIEWS;
950 }
951
952
953 /**
954 * Software fallback to do glDrawPixels(GL_STENCIL_INDEX) when we
955 * can't use a fragment shader to write stencil values.
956 */
957 static void
draw_stencil_pixels(struct gl_context * ctx,GLint x,GLint y,GLsizei width,GLsizei height,GLenum format,GLenum type,const struct gl_pixelstore_attrib * unpack,const void * pixels)958 draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
959 GLsizei width, GLsizei height, GLenum format, GLenum type,
960 const struct gl_pixelstore_attrib *unpack,
961 const void *pixels)
962 {
963 struct st_context *st = st_context(ctx);
964 struct pipe_context *pipe = st->pipe;
965 struct gl_renderbuffer *rb;
966 enum pipe_map_flags usage;
967 struct pipe_transfer *pt;
968 const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0;
969 ubyte *stmap;
970 struct gl_pixelstore_attrib clippedUnpack = *unpack;
971 GLubyte *sValues;
972 GLuint *zValues;
973
974 rb = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
975
976 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
977 y = ctx->DrawBuffer->Height - y - height;
978 }
979
980 if (format == GL_STENCIL_INDEX &&
981 _mesa_is_format_packed_depth_stencil(rb->Format)) {
982 /* writing stencil to a combined depth+stencil buffer */
983 usage = PIPE_MAP_READ_WRITE;
984 }
985 else {
986 usage = PIPE_MAP_WRITE;
987 }
988
989 stmap = pipe_texture_map(pipe, rb->texture,
990 rb->surface->u.tex.level,
991 rb->surface->u.tex.first_layer,
992 usage, x, y,
993 width, height, &pt);
994
995 pixels = _mesa_map_pbo_source(ctx, &clippedUnpack, pixels);
996 assert(pixels);
997
998 sValues = malloc(width * sizeof(GLubyte));
999 zValues = malloc(width * sizeof(GLuint));
1000
1001 if (sValues && zValues) {
1002 GLint row;
1003 for (row = 0; row < height; row++) {
1004 GLfloat *zValuesFloat = (GLfloat*)zValues;
1005 GLenum destType = GL_UNSIGNED_BYTE;
1006 const void *source = _mesa_image_address2d(&clippedUnpack, pixels,
1007 width, height,
1008 format, type,
1009 row, 0);
1010 _mesa_unpack_stencil_span(ctx, width, destType, sValues,
1011 type, source, &clippedUnpack,
1012 ctx->_ImageTransferState);
1013
1014 if (format == GL_DEPTH_STENCIL) {
1015 GLenum ztype =
1016 pt->resource->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT ?
1017 GL_FLOAT : GL_UNSIGNED_INT;
1018
1019 _mesa_unpack_depth_span(ctx, width, ztype, zValues,
1020 (1 << 24) - 1, type, source,
1021 &clippedUnpack);
1022 }
1023
1024 if (zoom) {
1025 _mesa_problem(ctx, "Gallium glDrawPixels(GL_STENCIL) with "
1026 "zoom not complete");
1027 }
1028
1029 {
1030 GLint spanY;
1031
1032 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1033 spanY = height - row - 1;
1034 }
1035 else {
1036 spanY = row;
1037 }
1038
1039 /* now pack the stencil (and Z) values in the dest format */
1040 switch (pt->resource->format) {
1041 case PIPE_FORMAT_S8_UINT:
1042 {
1043 ubyte *dest = stmap + spanY * pt->stride;
1044 assert(usage == PIPE_MAP_WRITE);
1045 memcpy(dest, sValues, width);
1046 }
1047 break;
1048 case PIPE_FORMAT_Z24_UNORM_S8_UINT:
1049 if (format == GL_DEPTH_STENCIL) {
1050 uint *dest = (uint *) (stmap + spanY * pt->stride);
1051 GLint k;
1052 assert(usage == PIPE_MAP_WRITE);
1053 for (k = 0; k < width; k++) {
1054 dest[k] = zValues[k] | (sValues[k] << 24);
1055 }
1056 }
1057 else {
1058 uint *dest = (uint *) (stmap + spanY * pt->stride);
1059 GLint k;
1060 assert(usage == PIPE_MAP_READ_WRITE);
1061 for (k = 0; k < width; k++) {
1062 dest[k] = (dest[k] & 0xffffff) | (sValues[k] << 24);
1063 }
1064 }
1065 break;
1066 case PIPE_FORMAT_S8_UINT_Z24_UNORM:
1067 if (format == GL_DEPTH_STENCIL) {
1068 uint *dest = (uint *) (stmap + spanY * pt->stride);
1069 GLint k;
1070 assert(usage == PIPE_MAP_WRITE);
1071 for (k = 0; k < width; k++) {
1072 dest[k] = (zValues[k] << 8) | (sValues[k] & 0xff);
1073 }
1074 }
1075 else {
1076 uint *dest = (uint *) (stmap + spanY * pt->stride);
1077 GLint k;
1078 assert(usage == PIPE_MAP_READ_WRITE);
1079 for (k = 0; k < width; k++) {
1080 dest[k] = (dest[k] & 0xffffff00) | (sValues[k] & 0xff);
1081 }
1082 }
1083 break;
1084 case PIPE_FORMAT_Z32_FLOAT_S8X24_UINT:
1085 if (format == GL_DEPTH_STENCIL) {
1086 uint *dest = (uint *) (stmap + spanY * pt->stride);
1087 GLfloat *destf = (GLfloat*)dest;
1088 GLint k;
1089 assert(usage == PIPE_MAP_WRITE);
1090 for (k = 0; k < width; k++) {
1091 destf[k*2] = zValuesFloat[k];
1092 dest[k*2+1] = sValues[k] & 0xff;
1093 }
1094 }
1095 else {
1096 uint *dest = (uint *) (stmap + spanY * pt->stride);
1097 GLint k;
1098 assert(usage == PIPE_MAP_READ_WRITE);
1099 for (k = 0; k < width; k++) {
1100 dest[k*2+1] = sValues[k] & 0xff;
1101 }
1102 }
1103 break;
1104 default:
1105 assert(0);
1106 }
1107 }
1108 }
1109 }
1110 else {
1111 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels()");
1112 }
1113
1114 free(sValues);
1115 free(zValues);
1116
1117 _mesa_unmap_pbo_source(ctx, &clippedUnpack);
1118
1119 /* unmap the stencil buffer */
1120 pipe_texture_unmap(pipe, pt);
1121 }
1122
1123
1124 /**
1125 * Get fragment program variant for a glDrawPixels or glCopyPixels
1126 * command for RGBA data.
1127 */
1128 static struct st_fp_variant *
get_color_fp_variant(struct st_context * st)1129 get_color_fp_variant(struct st_context *st)
1130 {
1131 struct gl_context *ctx = st->ctx;
1132 struct st_fp_variant_key key;
1133 struct st_fp_variant *fpv;
1134
1135 memset(&key, 0, sizeof(key));
1136
1137 key.st = st->has_shareable_shaders ? NULL : st;
1138 key.drawpixels = 1;
1139 key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
1140 ctx->Pixel.RedScale != 1.0 ||
1141 ctx->Pixel.GreenBias != 0.0 ||
1142 ctx->Pixel.GreenScale != 1.0 ||
1143 ctx->Pixel.BlueBias != 0.0 ||
1144 ctx->Pixel.BlueScale != 1.0 ||
1145 ctx->Pixel.AlphaBias != 0.0 ||
1146 ctx->Pixel.AlphaScale != 1.0);
1147 key.pixelMaps = ctx->Pixel.MapColorFlag;
1148 key.clamp_color = st->clamp_frag_color_in_shader &&
1149 ctx->Color._ClampFragmentColor;
1150 key.lower_alpha_func = COMPARE_FUNC_ALWAYS;
1151
1152 fpv = st_get_fp_variant(st, st->fp, &key);
1153
1154 return fpv;
1155 }
1156
1157 /**
1158 * Get fragment program variant for a glDrawPixels command
1159 * for COLOR_INDEX data
1160 */
1161 static struct st_fp_variant *
get_color_index_fp_variant(struct st_context * st)1162 get_color_index_fp_variant(struct st_context *st)
1163 {
1164 struct gl_context *ctx = st->ctx;
1165 struct st_fp_variant_key key;
1166 struct st_fp_variant *fpv;
1167
1168 memset(&key, 0, sizeof(key));
1169
1170 key.st = st->has_shareable_shaders ? NULL : st;
1171 key.drawpixels = 1;
1172 /* Since GL is always in RGBA mode MapColorFlag does not
1173 * affect GL_COLOR_INDEX format.
1174 * Scale and bias also never affect GL_COLOR_INDEX format.
1175 */
1176 key.scaleAndBias = 0;
1177 key.pixelMaps = 0;
1178 key.clamp_color = st->clamp_frag_color_in_shader &&
1179 ctx->Color._ClampFragmentColor;
1180 key.lower_alpha_func = COMPARE_FUNC_ALWAYS;
1181
1182 fpv = st_get_fp_variant(st, st->fp, &key);
1183
1184 return fpv;
1185 }
1186
1187
1188 /**
1189 * Clamp glDrawPixels width and height to the maximum texture size.
1190 */
1191 static void
clamp_size(struct st_context * st,GLsizei * width,GLsizei * height,struct gl_pixelstore_attrib * unpack)1192 clamp_size(struct st_context *st, GLsizei *width, GLsizei *height,
1193 struct gl_pixelstore_attrib *unpack)
1194 {
1195 const int maxSize = st->screen->get_param(st->screen,
1196 PIPE_CAP_MAX_TEXTURE_2D_SIZE);
1197
1198 if (*width > maxSize) {
1199 if (unpack->RowLength == 0)
1200 unpack->RowLength = *width;
1201 *width = maxSize;
1202 }
1203 if (*height > maxSize) {
1204 *height = maxSize;
1205 }
1206 }
1207
1208
1209 /**
1210 * Search the array of 4 swizzle components for the named component and return
1211 * its position.
1212 */
1213 static unsigned
search_swizzle(const unsigned char swizzle[4],unsigned component)1214 search_swizzle(const unsigned char swizzle[4], unsigned component)
1215 {
1216 unsigned i;
1217 for (i = 0; i < 4; i++) {
1218 if (swizzle[i] == component)
1219 return i;
1220 }
1221 assert(!"search_swizzle() failed");
1222 return 0;
1223 }
1224
1225
1226 /**
1227 * Set the sampler view's swizzle terms. This is used to handle RGBA
1228 * swizzling when the incoming image format isn't an exact match for
1229 * the actual texture format. For example, if we have glDrawPixels(
1230 * GL_RGBA, GL_UNSIGNED_BYTE) and we chose the texture format
1231 * PIPE_FORMAT_B8G8R8A8 then we can do use the sampler view swizzle to
1232 * avoid swizzling all the pixels in software in the texstore code.
1233 */
1234 static void
setup_sampler_swizzle(struct pipe_sampler_view * sv,GLenum format,GLenum type)1235 setup_sampler_swizzle(struct pipe_sampler_view *sv, GLenum format, GLenum type)
1236 {
1237 if ((format == GL_RGBA || format == GL_BGRA) && type == GL_UNSIGNED_BYTE) {
1238 const struct util_format_description *desc =
1239 util_format_description(sv->format);
1240 unsigned c0, c1, c2, c3;
1241
1242 /* Every gallium driver supports at least one 32-bit packed RGBA format.
1243 * We must have chosen one for (GL_RGBA, GL_UNSIGNED_BYTE).
1244 */
1245 assert(desc->block.bits == 32);
1246
1247 /* invert the format's swizzle to setup the sampler's swizzle */
1248 if (format == GL_RGBA) {
1249 c0 = PIPE_SWIZZLE_X;
1250 c1 = PIPE_SWIZZLE_Y;
1251 c2 = PIPE_SWIZZLE_Z;
1252 c3 = PIPE_SWIZZLE_W;
1253 }
1254 else {
1255 assert(format == GL_BGRA);
1256 c0 = PIPE_SWIZZLE_Z;
1257 c1 = PIPE_SWIZZLE_Y;
1258 c2 = PIPE_SWIZZLE_X;
1259 c3 = PIPE_SWIZZLE_W;
1260 }
1261 sv->swizzle_r = search_swizzle(desc->swizzle, c0);
1262 sv->swizzle_g = search_swizzle(desc->swizzle, c1);
1263 sv->swizzle_b = search_swizzle(desc->swizzle, c2);
1264 sv->swizzle_a = search_swizzle(desc->swizzle, c3);
1265 }
1266 else {
1267 /* use the default sampler swizzle */
1268 }
1269 }
1270
1271 void
st_DrawPixels(struct gl_context * ctx,GLint x,GLint y,GLsizei width,GLsizei height,GLenum format,GLenum type,const struct gl_pixelstore_attrib * unpack,const void * pixels)1272 st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
1273 GLsizei width, GLsizei height,
1274 GLenum format, GLenum type,
1275 const struct gl_pixelstore_attrib *unpack, const void *pixels)
1276 {
1277 void *driver_fp;
1278 struct st_context *st = st_context(ctx);
1279 GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
1280 struct pipe_sampler_view *sv[2] = { NULL };
1281 int num_sampler_view = 1;
1282 struct gl_pixelstore_attrib clippedUnpack;
1283 struct st_fp_variant *fpv = NULL;
1284 struct pipe_resource *pt;
1285
1286 /* Mesa state should be up to date by now */
1287 assert(ctx->NewState == 0x0);
1288
1289 _mesa_update_draw_buffer_bounds(ctx, ctx->DrawBuffer);
1290
1291 st_flush_bitmap_cache(st);
1292 st_invalidate_readpix_cache(st);
1293
1294 st_validate_state(st, ST_PIPELINE_META);
1295
1296 clippedUnpack = *unpack;
1297 unpack = &clippedUnpack;
1298
1299 /* Skip totally clipped DrawPixels. */
1300 if (ctx->Pixel.ZoomX == 1 && ctx->Pixel.ZoomY == 1 &&
1301 !_mesa_clip_drawpixels(ctx, &x, &y, &width, &height, &clippedUnpack))
1302 return;
1303
1304 /* Limit the size of the glDrawPixels to the max texture size.
1305 * Strictly speaking, that's not correct but since we don't handle
1306 * larger images yet, this is better than crashing.
1307 */
1308 clamp_size(st, &width, &height, &clippedUnpack);
1309
1310 if (format == GL_DEPTH_STENCIL)
1311 write_stencil = write_depth = GL_TRUE;
1312 else if (format == GL_STENCIL_INDEX)
1313 write_stencil = GL_TRUE;
1314 else if (format == GL_DEPTH_COMPONENT)
1315 write_depth = GL_TRUE;
1316
1317 if (write_stencil &&
1318 !st->has_stencil_export) {
1319 /* software fallback */
1320 draw_stencil_pixels(ctx, x, y, width, height, format, type,
1321 unpack, pixels);
1322 return;
1323 }
1324
1325 /* Put glDrawPixels image into a texture */
1326 pt = make_texture(st, width, height, format, type, unpack, pixels);
1327 if (!pt) {
1328 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
1329 return;
1330 }
1331
1332 st_make_passthrough_vertex_shader(st);
1333
1334 /*
1335 * Get vertex/fragment shaders
1336 */
1337 if (write_depth || write_stencil) {
1338 driver_fp = get_drawpix_z_stencil_program(st, write_depth,
1339 write_stencil);
1340 }
1341 else {
1342 fpv = (format != GL_COLOR_INDEX) ? get_color_fp_variant(st) :
1343 get_color_index_fp_variant(st);
1344
1345 driver_fp = fpv->base.driver_shader;
1346
1347 if (ctx->Pixel.MapColorFlag && format != GL_COLOR_INDEX) {
1348 pipe_sampler_view_reference(&sv[1],
1349 st->pixel_xfer.pixelmap_sampler_view);
1350 num_sampler_view++;
1351 }
1352
1353 /* compiling a new fragment shader variant added new state constants
1354 * into the constant buffer, we need to update them
1355 */
1356 st_upload_constants(st, st->fp, MESA_SHADER_FRAGMENT);
1357 }
1358
1359 {
1360 /* create sampler view for the image */
1361 struct pipe_sampler_view templ;
1362
1363 u_sampler_view_default_template(&templ, pt, pt->format);
1364 /* Set up the sampler view's swizzle */
1365 setup_sampler_swizzle(&templ, format, type);
1366
1367 sv[0] = st->pipe->create_sampler_view(st->pipe, pt, &templ);
1368 }
1369 if (!sv[0]) {
1370 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
1371 pipe_resource_reference(&pt, NULL);
1372 return;
1373 }
1374
1375 /* Create a second sampler view to read stencil. The stencil is
1376 * written using the shader stencil export functionality.
1377 */
1378 if (write_stencil) {
1379 enum pipe_format stencil_format =
1380 util_format_stencil_only(pt->format);
1381 /* we should not be doing pixel map/transfer (see above) */
1382 assert(num_sampler_view == 1);
1383 sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
1384 stencil_format);
1385 if (!sv[1]) {
1386 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
1387 pipe_resource_reference(&pt, NULL);
1388 pipe_sampler_view_reference(&sv[0], NULL);
1389 return;
1390 }
1391 num_sampler_view++;
1392 }
1393
1394 draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
1395 width, height,
1396 ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1397 sv,
1398 num_sampler_view,
1399 st->passthrough_vs,
1400 driver_fp, fpv,
1401 ctx->Current.RasterColor,
1402 GL_FALSE, write_depth, write_stencil);
1403
1404 /* free the texture (but may persist in the cache) */
1405 pipe_resource_reference(&pt, NULL);
1406 }
1407
1408
1409
1410 /**
1411 * Software fallback for glCopyPixels(GL_STENCIL).
1412 */
1413 static void
copy_stencil_pixels(struct gl_context * ctx,GLint srcx,GLint srcy,GLsizei width,GLsizei height,GLint dstx,GLint dsty)1414 copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1415 GLsizei width, GLsizei height,
1416 GLint dstx, GLint dsty)
1417 {
1418 struct gl_renderbuffer *rbDraw;
1419 struct pipe_context *pipe = st_context(ctx)->pipe;
1420 enum pipe_map_flags usage;
1421 struct pipe_transfer *ptDraw;
1422 ubyte *drawMap;
1423 ubyte *buffer;
1424 int i;
1425
1426 buffer = malloc(width * height * sizeof(ubyte));
1427 if (!buffer) {
1428 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)");
1429 return;
1430 }
1431
1432 /* Get the dest renderbuffer */
1433 rbDraw = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
1434
1435 /* this will do stencil pixel transfer ops */
1436 _mesa_readpixels(ctx, srcx, srcy, width, height,
1437 GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
1438 &ctx->DefaultPacking, buffer);
1439
1440 if (0) {
1441 /* debug code: dump stencil values */
1442 GLint row, col;
1443 for (row = 0; row < height; row++) {
1444 printf("%3d: ", row);
1445 for (col = 0; col < width; col++) {
1446 printf("%02x ", buffer[col + row * width]);
1447 }
1448 printf("\n");
1449 }
1450 }
1451
1452 if (_mesa_is_format_packed_depth_stencil(rbDraw->Format))
1453 usage = PIPE_MAP_READ_WRITE;
1454 else
1455 usage = PIPE_MAP_WRITE;
1456
1457 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1458 dsty = rbDraw->Height - dsty - height;
1459 }
1460
1461 assert(util_format_get_blockwidth(rbDraw->texture->format) == 1);
1462 assert(util_format_get_blockheight(rbDraw->texture->format) == 1);
1463
1464 /* map the stencil buffer */
1465 drawMap = pipe_texture_map(pipe,
1466 rbDraw->texture,
1467 rbDraw->surface->u.tex.level,
1468 rbDraw->surface->u.tex.first_layer,
1469 usage, dstx, dsty,
1470 width, height, &ptDraw);
1471
1472 /* draw */
1473 /* XXX PixelZoom not handled yet */
1474 for (i = 0; i < height; i++) {
1475 ubyte *dst;
1476 const ubyte *src;
1477 int y;
1478
1479 y = i;
1480
1481 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1482 y = height - y - 1;
1483 }
1484
1485 dst = drawMap + y * ptDraw->stride;
1486 src = buffer + i * width;
1487
1488 _mesa_pack_ubyte_stencil_row(rbDraw->Format, width, src, dst);
1489 }
1490
1491 free(buffer);
1492
1493 /* unmap the stencil buffer */
1494 pipe_texture_unmap(pipe, ptDraw);
1495 }
1496
1497
1498 /**
1499 * Return renderbuffer to use for reading color pixels for glCopyPixels
1500 */
1501 static struct gl_renderbuffer *
st_get_color_read_renderbuffer(struct gl_context * ctx)1502 st_get_color_read_renderbuffer(struct gl_context *ctx)
1503 {
1504 struct gl_framebuffer *fb = ctx->ReadBuffer;
1505 return fb->_ColorReadBuffer;
1506 }
1507
1508
1509 /**
1510 * Try to do a glCopyPixels for simple cases with a blit by calling
1511 * pipe->blit().
1512 *
1513 * We can do this when we're copying color pixels (depth/stencil
1514 * eventually) with no pixel zoom, no pixel transfer ops, no
1515 * per-fragment ops, and the src/dest regions don't overlap.
1516 */
1517 static GLboolean
blit_copy_pixels(struct gl_context * ctx,GLint srcx,GLint srcy,GLsizei width,GLsizei height,GLint dstx,GLint dsty,GLenum type)1518 blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1519 GLsizei width, GLsizei height,
1520 GLint dstx, GLint dsty, GLenum type)
1521 {
1522 struct st_context *st = st_context(ctx);
1523 struct pipe_context *pipe = st->pipe;
1524 struct pipe_screen *screen = st->screen;
1525 struct gl_pixelstore_attrib pack, unpack;
1526 GLint readX, readY, readW, readH, drawX, drawY, drawW, drawH;
1527
1528 if (type == GL_DEPTH_STENCIL_TO_RGBA_NV || type == GL_DEPTH_STENCIL_TO_BGRA_NV)
1529 return GL_FALSE;
1530
1531 if (ctx->Pixel.ZoomX == 1.0 &&
1532 ctx->Pixel.ZoomY == 1.0 &&
1533 (type != GL_COLOR ||
1534 (ctx->_ImageTransferState == 0x0 &&
1535 !ctx->Color.BlendEnabled &&
1536 !ctx->Color.AlphaEnabled &&
1537 (!ctx->Color.ColorLogicOpEnabled || ctx->Color.LogicOp == GL_COPY) &&
1538 !ctx->Depth.BoundsTest &&
1539 (!ctx->Depth.Test || (ctx->Depth.Func == GL_ALWAYS && !ctx->Depth.Mask)) &&
1540 !ctx->Fog.Enabled &&
1541 (!ctx->Stencil.Enabled ||
1542 (ctx->Stencil.FailFunc[0] == GL_KEEP &&
1543 ctx->Stencil.ZPassFunc[0] == GL_KEEP &&
1544 ctx->Stencil.ZFailFunc[0] == GL_KEEP)) &&
1545 !ctx->FragmentProgram.Enabled &&
1546 !ctx->_Shader->CurrentProgram[MESA_SHADER_FRAGMENT] &&
1547 !_mesa_ati_fragment_shader_enabled(ctx) &&
1548 ctx->DrawBuffer->_NumColorDrawBuffers == 1)) &&
1549 !ctx->Query.CurrentOcclusionObject) {
1550 struct gl_renderbuffer *rbRead, *rbDraw;
1551
1552 /*
1553 * Clip the read region against the src buffer bounds.
1554 * We'll still allocate a temporary buffer/texture for the original
1555 * src region size but we'll only read the region which is on-screen.
1556 * This may mean that we draw garbage pixels into the dest region, but
1557 * that's expected.
1558 */
1559 readX = srcx;
1560 readY = srcy;
1561 readW = width;
1562 readH = height;
1563 pack = ctx->DefaultPacking;
1564 if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
1565 return GL_TRUE; /* all done */
1566
1567 /* clip against dest buffer bounds and scissor box */
1568 drawX = dstx + pack.SkipPixels;
1569 drawY = dsty + pack.SkipRows;
1570 unpack = pack;
1571 if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
1572 return GL_TRUE; /* all done */
1573
1574 readX = readX - pack.SkipPixels + unpack.SkipPixels;
1575 readY = readY - pack.SkipRows + unpack.SkipRows;
1576
1577 drawW = readW;
1578 drawH = readH;
1579
1580 if (type == GL_COLOR) {
1581 rbRead = st_get_color_read_renderbuffer(ctx);
1582 rbDraw = ctx->DrawBuffer->_ColorDrawBuffers[0];
1583 } else if (type == GL_DEPTH || type == GL_DEPTH_STENCIL) {
1584 rbRead = ctx->ReadBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
1585 rbDraw = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
1586 } else if (type == GL_STENCIL) {
1587 rbRead = ctx->ReadBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
1588 rbDraw = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
1589 } else {
1590 return false;
1591 }
1592
1593 /* Flip src/dst position depending on the orientation of buffers. */
1594 if (_mesa_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1595 readY = rbRead->Height - readY;
1596 readH = -readH;
1597 }
1598
1599 if (_mesa_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
1600 /* We can't flip the destination for pipe->blit, so we only adjust
1601 * its position and flip the source.
1602 */
1603 drawY = rbDraw->Height - drawY - drawH;
1604 readY += readH;
1605 readH = -readH;
1606 }
1607
1608 if (rbRead != rbDraw ||
1609 !_mesa_regions_overlap(readX, readY, readX + readW, readY + readH,
1610 drawX, drawY, drawX + drawW, drawY + drawH)) {
1611 struct pipe_blit_info blit;
1612
1613 memset(&blit, 0, sizeof(blit));
1614 blit.src.resource = rbRead->texture;
1615 blit.src.level = rbRead->surface->u.tex.level;
1616 blit.src.format = rbRead->texture->format;
1617 blit.src.box.x = readX;
1618 blit.src.box.y = readY;
1619 blit.src.box.z = rbRead->surface->u.tex.first_layer;
1620 blit.src.box.width = readW;
1621 blit.src.box.height = readH;
1622 blit.src.box.depth = 1;
1623 blit.dst.resource = rbDraw->texture;
1624 blit.dst.level = rbDraw->surface->u.tex.level;
1625 blit.dst.format = rbDraw->texture->format;
1626 blit.dst.box.x = drawX;
1627 blit.dst.box.y = drawY;
1628 blit.dst.box.z = rbDraw->surface->u.tex.first_layer;
1629 blit.dst.box.width = drawW;
1630 blit.dst.box.height = drawH;
1631 blit.dst.box.depth = 1;
1632 blit.filter = PIPE_TEX_FILTER_NEAREST;
1633 blit.render_condition_enable = ctx->Query.CondRenderQuery != NULL;
1634
1635 if (type == GL_COLOR)
1636 blit.mask |= PIPE_MASK_RGBA;
1637 if (type == GL_DEPTH)
1638 blit.mask |= PIPE_MASK_Z;
1639 if (type == GL_STENCIL)
1640 blit.mask |= PIPE_MASK_S;
1641 if (type == GL_DEPTH_STENCIL)
1642 blit.mask |= PIPE_MASK_ZS;
1643
1644 if (ctx->DrawBuffer != ctx->WinSysDrawBuffer)
1645 st_window_rectangles_to_blit(ctx, &blit);
1646
1647 if (screen->is_format_supported(screen, blit.src.format,
1648 blit.src.resource->target,
1649 blit.src.resource->nr_samples,
1650 blit.src.resource->nr_storage_samples,
1651 PIPE_BIND_SAMPLER_VIEW) &&
1652 screen->is_format_supported(screen, blit.dst.format,
1653 blit.dst.resource->target,
1654 blit.dst.resource->nr_samples,
1655 blit.dst.resource->nr_storage_samples,
1656 PIPE_BIND_RENDER_TARGET)) {
1657 pipe->blit(pipe, &blit);
1658 return GL_TRUE;
1659 }
1660 }
1661 }
1662
1663 return GL_FALSE;
1664 }
1665
1666 void
st_CopyPixels(struct gl_context * ctx,GLint srcx,GLint srcy,GLsizei width,GLsizei height,GLint dstx,GLint dsty,GLenum type)1667 st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
1668 GLsizei width, GLsizei height,
1669 GLint dstx, GLint dsty, GLenum type)
1670 {
1671 struct st_context *st = st_context(ctx);
1672 struct pipe_context *pipe = st->pipe;
1673 struct pipe_screen *screen = st->screen;
1674 struct gl_renderbuffer *rbRead;
1675 void *driver_fp;
1676 struct pipe_resource *pt;
1677 struct pipe_sampler_view *sv[2] = { NULL };
1678 struct st_fp_variant *fpv = NULL;
1679 int num_sampler_view = 1;
1680 enum pipe_format srcFormat;
1681 unsigned srcBind;
1682 GLboolean invertTex = GL_FALSE;
1683 GLint readX, readY, readW, readH;
1684 struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
1685 GLboolean write_stencil = GL_FALSE;
1686 GLboolean write_depth = GL_FALSE;
1687
1688 _mesa_update_draw_buffer_bounds(ctx, ctx->DrawBuffer);
1689
1690 st_flush_bitmap_cache(st);
1691 st_invalidate_readpix_cache(st);
1692
1693 st_validate_state(st, ST_PIPELINE_META);
1694
1695 if (blit_copy_pixels(ctx, srcx, srcy, width, height, dstx, dsty, type))
1696 return;
1697
1698 /* fallback if the driver can't do stencil exports */
1699 if (type == GL_DEPTH_STENCIL &&
1700 !st->has_stencil_export) {
1701 st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_STENCIL);
1702 st_CopyPixels(ctx, srcx, srcy, width, height, dstx, dsty, GL_DEPTH);
1703 return;
1704 }
1705
1706 /* fallback if the driver can't do stencil exports */
1707 if (type == GL_STENCIL &&
1708 !st->has_stencil_export) {
1709 copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
1710 return;
1711 }
1712
1713 /*
1714 * The subsequent code implements glCopyPixels by copying the source
1715 * pixels into a temporary texture that's then applied to a textured quad.
1716 * When we draw the textured quad, all the usual per-fragment operations
1717 * are handled.
1718 */
1719
1720 st_make_passthrough_vertex_shader(st);
1721
1722 /*
1723 * Get vertex/fragment shaders
1724 */
1725 if (type == GL_COLOR) {
1726 fpv = get_color_fp_variant(st);
1727
1728 rbRead = st_get_color_read_renderbuffer(ctx);
1729
1730 driver_fp = fpv->base.driver_shader;
1731
1732 if (ctx->Pixel.MapColorFlag) {
1733 pipe_sampler_view_reference(&sv[1],
1734 st->pixel_xfer.pixelmap_sampler_view);
1735 num_sampler_view++;
1736 }
1737
1738 /* compiling a new fragment shader variant added new state constants
1739 * into the constant buffer, we need to update them
1740 */
1741 st_upload_constants(st, st->fp, MESA_SHADER_FRAGMENT);
1742 } else if (type == GL_DEPTH) {
1743 rbRead = ctx->ReadBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
1744 driver_fp = get_drawpix_z_stencil_program(st, GL_TRUE, GL_FALSE);
1745 } else if (type == GL_STENCIL) {
1746 rbRead = ctx->ReadBuffer->Attachment[BUFFER_STENCIL].Renderbuffer;
1747 driver_fp = get_drawpix_z_stencil_program(st, GL_FALSE, GL_TRUE);
1748 } else if (type == GL_DEPTH_STENCIL) {
1749 rbRead = ctx->ReadBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
1750 driver_fp = get_drawpix_z_stencil_program(st, GL_TRUE, GL_TRUE);
1751 } else {
1752 assert(type == GL_DEPTH_STENCIL_TO_RGBA_NV || type == GL_DEPTH_STENCIL_TO_BGRA_NV);
1753 rbRead = ctx->ReadBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
1754 if (type == GL_DEPTH_STENCIL_TO_RGBA_NV)
1755 driver_fp = get_drawpix_zs_to_color_program(st, GL_TRUE);
1756 else
1757 driver_fp = get_drawpix_zs_to_color_program(st, GL_FALSE);
1758 if (!driver_fp) {
1759 assert(0 && "operation not supported by CopyPixels implemetation");
1760 return;
1761 }
1762 }
1763
1764
1765 /* Choose the format for the temporary texture. */
1766 srcFormat = rbRead->texture->format;
1767 srcBind = PIPE_BIND_SAMPLER_VIEW |
1768 (type == GL_COLOR ? PIPE_BIND_RENDER_TARGET : PIPE_BIND_DEPTH_STENCIL);
1769
1770 if (!screen->is_format_supported(screen, srcFormat, st->internal_target, 0,
1771 0, srcBind)) {
1772 /* srcFormat is non-renderable. Find a compatible renderable format. */
1773 if (type == GL_DEPTH) {
1774 srcFormat = st_choose_format(st, GL_DEPTH_COMPONENT, GL_NONE,
1775 GL_NONE, st->internal_target, 0, 0,
1776 srcBind, false, false);
1777 }
1778 else if (type == GL_STENCIL) {
1779 /* can't use texturing, fallback to copy */
1780 copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty);
1781 return;
1782 }
1783 else {
1784 assert(type == GL_COLOR);
1785
1786 if (util_format_is_float(srcFormat)) {
1787 srcFormat = st_choose_format(st, GL_RGBA32F, GL_NONE,
1788 GL_NONE, st->internal_target, 0, 0,
1789 srcBind, false, false);
1790 }
1791 else if (util_format_is_pure_sint(srcFormat)) {
1792 srcFormat = st_choose_format(st, GL_RGBA32I, GL_NONE,
1793 GL_NONE, st->internal_target, 0, 0,
1794 srcBind, false, false);
1795 }
1796 else if (util_format_is_pure_uint(srcFormat)) {
1797 srcFormat = st_choose_format(st, GL_RGBA32UI, GL_NONE,
1798 GL_NONE, st->internal_target, 0, 0,
1799 srcBind, false, false);
1800 }
1801 else if (util_format_is_snorm(srcFormat)) {
1802 srcFormat = st_choose_format(st, GL_RGBA16_SNORM, GL_NONE,
1803 GL_NONE, st->internal_target, 0, 0,
1804 srcBind, false, false);
1805 }
1806 else {
1807 srcFormat = st_choose_format(st, GL_RGBA, GL_NONE,
1808 GL_NONE, st->internal_target, 0, 0,
1809 srcBind, false, false);
1810 }
1811 }
1812
1813 if (srcFormat == PIPE_FORMAT_NONE) {
1814 assert(0 && "cannot choose a format for src of CopyPixels");
1815 return;
1816 }
1817 }
1818
1819 /* Invert src region if needed */
1820 if (_mesa_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
1821 srcy = ctx->ReadBuffer->Height - srcy - height;
1822 invertTex = !invertTex;
1823 }
1824
1825 /* Clip the read region against the src buffer bounds.
1826 * We'll still allocate a temporary buffer/texture for the original
1827 * src region size but we'll only read the region which is on-screen.
1828 * This may mean that we draw garbage pixels into the dest region, but
1829 * that's expected.
1830 */
1831 readX = srcx;
1832 readY = srcy;
1833 readW = width;
1834 readH = height;
1835 if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack)) {
1836 /* The source region is completely out of bounds. Do nothing.
1837 * The GL spec says "Results of copies from outside the window,
1838 * or from regions of the window that are not exposed, are
1839 * hardware dependent and undefined."
1840 */
1841 return;
1842 }
1843
1844 readW = MAX2(0, readW);
1845 readH = MAX2(0, readH);
1846
1847 /* Allocate the temporary texture. */
1848 pt = alloc_texture(st, width, height, srcFormat, srcBind);
1849 if (!pt)
1850 return;
1851
1852 sv[0] = st_create_texture_sampler_view(st->pipe, pt);
1853 if (!sv[0]) {
1854 pipe_resource_reference(&pt, NULL);
1855 return;
1856 }
1857
1858 /* Create a second sampler view to read stencil */
1859 if (type == GL_STENCIL || type == GL_DEPTH_STENCIL ||
1860 type == GL_DEPTH_STENCIL_TO_RGBA_NV || type == GL_DEPTH_STENCIL_TO_BGRA_NV) {
1861 write_stencil = GL_TRUE;
1862 if (type == GL_DEPTH_STENCIL)
1863 write_depth = GL_TRUE;
1864 if (type == GL_DEPTH_STENCIL_TO_RGBA_NV || type == GL_DEPTH_STENCIL_TO_BGRA_NV) {
1865 write_depth = FALSE;
1866 write_stencil = FALSE;
1867 }
1868
1869 enum pipe_format stencil_format =
1870 util_format_stencil_only(pt->format);
1871 /* we should not be doing pixel map/transfer (see above) */
1872 assert(num_sampler_view == 1);
1873 sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
1874 stencil_format);
1875 if (!sv[1]) {
1876 _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels");
1877 pipe_resource_reference(&pt, NULL);
1878 pipe_sampler_view_reference(&sv[0], NULL);
1879 return;
1880 }
1881 num_sampler_view++;
1882 }
1883 /* Copy the src region to the temporary texture. */
1884 {
1885 struct pipe_blit_info blit;
1886
1887 memset(&blit, 0, sizeof(blit));
1888 blit.src.resource = rbRead->texture;
1889 blit.src.level = rbRead->surface->u.tex.level;
1890 blit.src.format = rbRead->texture->format;
1891 blit.src.box.x = readX;
1892 blit.src.box.y = readY;
1893 blit.src.box.z = rbRead->surface->u.tex.first_layer;
1894 blit.src.box.width = readW;
1895 blit.src.box.height = readH;
1896 blit.src.box.depth = 1;
1897 blit.dst.resource = pt;
1898 blit.dst.level = 0;
1899 blit.dst.format = pt->format;
1900 blit.dst.box.x = pack.SkipPixels;
1901 blit.dst.box.y = pack.SkipRows;
1902 blit.dst.box.z = 0;
1903 blit.dst.box.width = readW;
1904 blit.dst.box.height = readH;
1905 blit.dst.box.depth = 1;
1906 if (type == GL_DEPTH)
1907 blit.mask = util_format_get_mask(pt->format) & ~PIPE_MASK_S;
1908 else if (type == GL_STENCIL)
1909 blit.mask = util_format_get_mask(pt->format) & ~PIPE_MASK_Z;
1910 else
1911 blit.mask = util_format_get_mask(pt->format);
1912 blit.filter = PIPE_TEX_FILTER_NEAREST;
1913
1914 pipe->blit(pipe, &blit);
1915 }
1916
1917 /* OK, the texture 'pt' contains the src image/pixels. Now draw a
1918 * textured quad with that texture.
1919 */
1920
1921 draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
1922 width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
1923 sv,
1924 num_sampler_view,
1925 st->passthrough_vs,
1926 driver_fp, fpv,
1927 ctx->Current.Attrib[VERT_ATTRIB_COLOR0],
1928 invertTex, write_depth, write_stencil);
1929
1930 pipe_resource_reference(&pt, NULL);
1931 }
1932
1933 void
st_destroy_drawpix(struct st_context * st)1934 st_destroy_drawpix(struct st_context *st)
1935 {
1936 GLuint i;
1937
1938 for (i = 0; i < ARRAY_SIZE(st->drawpix.zs_shaders); i++) {
1939 if (st->drawpix.zs_shaders[i])
1940 st->pipe->delete_fs_state(st->pipe, st->drawpix.zs_shaders[i]);
1941 }
1942
1943 if (st->passthrough_vs)
1944 st->pipe->delete_vs_state(st->pipe, st->passthrough_vs);
1945
1946 /* Free cache data */
1947 for (i = 0; i < ARRAY_SIZE(st->drawpix_cache.entries); i++) {
1948 struct drawpix_cache_entry *entry = &st->drawpix_cache.entries[i];
1949 free(entry->image);
1950 pipe_resource_reference(&entry->texture, NULL);
1951 }
1952 }
1953