2 * Mesa 3-D graphics library
4 * Copyright (C) 2009 VMware, Inc. All Rights Reserved.
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the "Software"),
8 * to deal in the Software without restriction, including without limitation
9 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
10 * and/or sell copies of the Software, and to permit persons to whom the
11 * Software is furnished to do so, subject to the following conditions:
13 * The above copyright notice and this permission notice shall be included
14 * in all copies or substantial portions of the Software.
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
17 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
20 * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
21 * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
22 * OTHER DEALINGS IN THE SOFTWARE.
25 #include "main/glheader.h"
26 #include "main/mtypes.h"
27 #include "main/imports.h"
28 #include "main/arbprogram.h"
29 #include "main/arrayobj.h"
30 #include "main/blend.h"
31 #include "main/condrender.h"
32 #include "main/depth.h"
33 #include "main/enable.h"
34 #include "main/enums.h"
35 #include "main/fbobject.h"
36 #include "main/image.h"
37 #include "main/macros.h"
38 #include "main/matrix.h"
39 #include "main/multisample.h"
40 #include "main/objectlabel.h"
41 #include "main/readpix.h"
42 #include "main/scissor.h"
43 #include "main/shaderapi.h"
44 #include "main/texobj.h"
45 #include "main/texenv.h"
46 #include "main/teximage.h"
47 #include "main/texparam.h"
48 #include "main/uniforms.h"
49 #include "main/varray.h"
50 #include "main/viewport.h"
51 #include "swrast/swrast.h"
52 #include "drivers/common/meta.h"
53 #include "util/ralloc.h"
55 /** Return offset in bytes of the field within a vertex struct */
56 #define OFFSET(FIELD) ((void *) offsetof(struct vertex, FIELD))
59 setup_glsl_msaa_blit_scaled_shader(struct gl_context
*ctx
,
60 struct blit_state
*blit
,
61 struct gl_renderbuffer
*src_rb
,
62 GLenum target
, GLenum filter
)
64 GLint loc_src_width
, loc_src_height
;
66 int shader_offset
= 0;
67 void *mem_ctx
= ralloc_context(NULL
);
69 char *name
, *sample_number
;
70 const uint8_t *sample_map
;
71 char *sample_map_str
= rzalloc_size(mem_ctx
, 1);
72 char *sample_map_expr
= rzalloc_size(mem_ctx
, 1);
73 char *texel_fetch_macro
= rzalloc_size(mem_ctx
, 1);
74 const char *sampler_array_suffix
= "";
75 float x_scale
, y_scale
;
76 enum blit_msaa_shader shader_index
;
79 samples
= MAX2(src_rb
->NumSamples
, 1);
85 y_scale
= samples
/ x_scale
;
87 /* We expect only power of 2 samples in source multisample buffer. */
88 assert(samples
> 0 && _mesa_is_pow_two(samples
));
89 while (samples
>> (shader_offset
+ 1)) {
92 /* Update the assert if we plan to support more than 16X MSAA. */
93 assert(shader_offset
> 0 && shader_offset
<= 4);
95 assert(target
== GL_TEXTURE_2D_MULTISAMPLE
||
96 target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
);
98 shader_index
= BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE
+
101 if (target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
) {
102 shader_index
+= BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_SCALED_RESOLVE
-
103 BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_SCALED_RESOLVE
;
104 sampler_array_suffix
= "Array";
107 if (blit
->msaa_shaders
[shader_index
]) {
108 _mesa_UseProgram(blit
->msaa_shaders
[shader_index
]);
109 /* Update the uniform values. */
111 _mesa_GetUniformLocation(blit
->msaa_shaders
[shader_index
], "src_width");
113 _mesa_GetUniformLocation(blit
->msaa_shaders
[shader_index
], "src_height");
114 _mesa_Uniform1f(loc_src_width
, src_rb
->Width
);
115 _mesa_Uniform1f(loc_src_height
, src_rb
->Height
);
119 name
= ralloc_asprintf(mem_ctx
, "vec4 MSAA scaled resolve");
121 /* Below switch is used to setup the shader expression, which computes
122 * sample index and map it to to a sample number on hardware.
126 sample_number
= "sample_map[int(2 * fract(coord.x))]";
127 sample_map
= ctx
->Const
.SampleMap2x
;
130 sample_number
= "sample_map[int(2 * fract(coord.x) + 4 * fract(coord.y))]";
131 sample_map
= ctx
->Const
.SampleMap4x
;
134 sample_number
= "sample_map[int(2 * fract(coord.x) + 8 * fract(coord.y))]";
135 sample_map
= ctx
->Const
.SampleMap8x
;
138 sample_number
= "sample_map[int(4 * fract(coord.x) + 16 * fract(coord.y))]";
139 sample_map
= ctx
->Const
.SampleMap16x
;
142 sample_number
= NULL
;
144 _mesa_problem(ctx
, "Unsupported sample count %d\n", samples
);
145 unreachable("Unsupported sample count");
148 /* Create sample map string. */
149 for (i
= 0 ; i
< samples
- 1; i
++) {
150 ralloc_asprintf_append(&sample_map_str
, "%d, ", sample_map
[i
]);
152 ralloc_asprintf_append(&sample_map_str
, "%d", sample_map
[samples
- 1]);
154 /* Create sample map expression using above string. */
155 ralloc_asprintf_append(&sample_map_expr
,
156 " const int sample_map[%d] = int[%d](%s);\n",
157 samples
, samples
, sample_map_str
);
159 if (target
== GL_TEXTURE_2D_MULTISAMPLE
) {
160 ralloc_asprintf_append(&texel_fetch_macro
,
161 "#define TEXEL_FETCH(coord) texelFetch(texSampler, ivec2(coord), %s);\n",
164 ralloc_asprintf_append(&texel_fetch_macro
,
165 "#define TEXEL_FETCH(coord) texelFetch(texSampler, ivec3(coord, layer), %s);\n",
169 static const char vs_source
[] =
171 "in vec2 position;\n"
172 "in vec3 textureCoords;\n"
173 "out vec2 texCoords;\n"
174 "flat out int layer;\n"
177 " texCoords = textureCoords.xy;\n"
178 " layer = int(textureCoords.z);\n"
179 " gl_Position = vec4(position, 0.0, 1.0);\n"
183 fs_source
= ralloc_asprintf(mem_ctx
,
185 "#extension GL_ARB_texture_multisample : enable\n"
186 "uniform sampler2DMS%s texSampler;\n"
187 "uniform float src_width, src_height;\n"
188 "in vec2 texCoords;\n"
189 "flat in int layer;\n"
190 "out vec4 out_color;\n"
196 " const vec2 scale = vec2(%ff, %ff);\n"
197 " const vec2 scale_inv = vec2(%ff, %ff);\n"
198 " const vec2 s_0_offset = vec2(%ff, %ff);\n"
199 " vec2 s_0_coord, s_1_coord, s_2_coord, s_3_coord;\n"
200 " vec4 s_0_color, s_1_color, s_2_color, s_3_color;\n"
201 " vec4 x_0_color, x_1_color;\n"
202 " vec2 tex_coord = texCoords - s_0_offset;\n"
204 " tex_coord *= scale;\n"
205 " tex_coord.x = clamp(tex_coord.x, 0.0f, scale.x * src_width - 1.0f);\n"
206 " tex_coord.y = clamp(tex_coord.y, 0.0f, scale.y * src_height - 1.0f);\n"
207 " interp = fract(tex_coord);\n"
208 " tex_coord = ivec2(tex_coord) * scale_inv;\n"
210 " /* Compute the sample coordinates used for filtering. */\n"
211 " s_0_coord = tex_coord;\n"
212 " s_1_coord = tex_coord + vec2(scale_inv.x, 0.0f);\n"
213 " s_2_coord = tex_coord + vec2(0.0f, scale_inv.y);\n"
214 " s_3_coord = tex_coord + vec2(scale_inv.x, scale_inv.y);\n"
216 " /* Fetch sample color values. */\n"
218 " s_0_color = TEXEL_FETCH(s_0_coord)\n"
219 " s_1_color = TEXEL_FETCH(s_1_coord)\n"
220 " s_2_color = TEXEL_FETCH(s_2_coord)\n"
221 " s_3_color = TEXEL_FETCH(s_3_coord)\n"
222 "#undef TEXEL_FETCH\n"
224 " /* Do bilinear filtering on sample colors. */\n"
225 " x_0_color = mix(s_0_color, s_1_color, interp.x);\n"
226 " x_1_color = mix(s_2_color, s_3_color, interp.x);\n"
227 " out_color = mix(x_0_color, x_1_color, interp.y);\n"
229 sampler_array_suffix
,
232 1.0f
/ x_scale
, 1.0f
/ y_scale
,
233 0.5f
/ x_scale
, 0.5f
/ y_scale
,
236 _mesa_meta_compile_and_link_program(ctx
, vs_source
, fs_source
, name
,
237 &blit
->msaa_shaders
[shader_index
]);
239 _mesa_GetUniformLocation(blit
->msaa_shaders
[shader_index
], "src_width");
241 _mesa_GetUniformLocation(blit
->msaa_shaders
[shader_index
], "src_height");
242 _mesa_Uniform1f(loc_src_width
, src_rb
->Width
);
243 _mesa_Uniform1f(loc_src_height
, src_rb
->Height
);
245 ralloc_free(mem_ctx
);
249 setup_glsl_msaa_blit_shader(struct gl_context
*ctx
,
250 struct blit_state
*blit
,
251 const struct gl_framebuffer
*drawFb
,
252 struct gl_renderbuffer
*src_rb
,
255 const char *vs_source
;
258 enum blit_msaa_shader shader_index
;
259 bool dst_is_msaa
= false;
261 const char *vec4_prefix
;
262 const char *sampler_array_suffix
= "";
264 const char *texcoord_type
= "vec2";
266 int shader_offset
= 0;
269 samples
= MAX2(src_rb
->NumSamples
, 1);
270 src_datatype
= _mesa_get_format_datatype(src_rb
->Format
);
272 /* depth-or-color glCopyTexImage fallback path that passes a NULL rb and
273 * doesn't handle integer.
276 src_datatype
= GL_UNSIGNED_NORMALIZED
;
279 /* We expect only power of 2 samples in source multisample buffer. */
280 assert(samples
> 0 && _mesa_is_pow_two(samples
));
281 while (samples
>> (shader_offset
+ 1)) {
284 /* Update the assert if we plan to support more than 16X MSAA. */
285 assert(shader_offset
>= 0 && shader_offset
<= 4);
287 if (drawFb
->Visual
.samples
> 1) {
288 /* If you're calling meta_BlitFramebuffer with the destination
289 * multisampled, this is the only path that will work -- swrast and
290 * CopyTexImage won't work on it either.
292 assert(ctx
->Extensions
.ARB_sample_shading
);
296 /* We need shader invocation per sample, not per pixel */
297 _mesa_set_enable(ctx
, GL_MULTISAMPLE
, GL_TRUE
);
298 _mesa_set_enable(ctx
, GL_SAMPLE_SHADING
, GL_TRUE
);
299 _mesa_MinSampleShading(1.0);
303 case GL_TEXTURE_2D_MULTISAMPLE
:
304 case GL_TEXTURE_2D_MULTISAMPLE_ARRAY
:
305 if (src_rb
&& (src_rb
->_BaseFormat
== GL_DEPTH_COMPONENT
||
306 src_rb
->_BaseFormat
== GL_DEPTH_STENCIL
)) {
308 shader_index
= BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_COPY
;
310 shader_index
= BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_RESOLVE
;
313 shader_index
= BLIT_MSAA_SHADER_2D_MULTISAMPLE_COPY
;
315 shader_index
= BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE
+
320 if (target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
) {
321 shader_index
+= (BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_RESOLVE
-
322 BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE
);
323 sampler_array_suffix
= "Array";
324 texcoord_type
= "vec3";
328 _mesa_problem(ctx
, "Unknown texture target %s\n",
329 _mesa_enum_to_string(target
));
330 shader_index
= BLIT_2X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE
;
333 /* We rely on the enum being sorted this way. */
334 STATIC_ASSERT(BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE_INT
==
335 BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE
+ 5);
336 STATIC_ASSERT(BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE_UINT
==
337 BLIT_1X_MSAA_SHADER_2D_MULTISAMPLE_RESOLVE
+ 10);
338 if (src_datatype
== GL_INT
) {
341 } else if (src_datatype
== GL_UNSIGNED_INT
) {
348 if (blit
->msaa_shaders
[shader_index
]) {
349 _mesa_UseProgram(blit
->msaa_shaders
[shader_index
]);
353 mem_ctx
= ralloc_context(NULL
);
355 if (shader_index
== BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_RESOLVE
||
356 shader_index
== BLIT_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_DEPTH_RESOLVE
||
357 shader_index
== BLIT_MSAA_SHADER_2D_MULTISAMPLE_ARRAY_DEPTH_COPY
||
358 shader_index
== BLIT_MSAA_SHADER_2D_MULTISAMPLE_DEPTH_COPY
) {
360 const char *tex_coords
= "texCoords";
363 sample_index
= "gl_SampleID";
364 name
= "depth MSAA copy";
366 if (ctx
->Extensions
.ARB_gpu_shader5
&& samples
>= 16) {
367 /* See comment below for the color copy */
368 tex_coords
= "interpolateAtOffset(texCoords, vec2(0.0))";
371 /* From the GL 4.3 spec:
373 * "If there is a multisample buffer (the value of SAMPLE_BUFFERS
374 * is one), then values are obtained from the depth samples in
375 * this buffer. It is recommended that the depth value of the
376 * centermost sample be used, though implementations may choose
377 * any function of the depth sample values at each pixel.
379 * We're slacking and instead of choosing centermost, we've got 0.
382 name
= "depth MSAA resolve";
385 vs_source
= ralloc_asprintf(mem_ctx
,
387 "in vec2 position;\n"
388 "in %s textureCoords;\n"
389 "out %s texCoords;\n"
392 " texCoords = textureCoords;\n"
393 " gl_Position = vec4(position, 0.0, 1.0);\n"
397 fs_source
= ralloc_asprintf(mem_ctx
,
399 "#extension GL_ARB_texture_multisample : enable\n"
400 "#extension GL_ARB_sample_shading : enable\n"
401 "#extension GL_ARB_gpu_shader5 : enable\n"
402 "uniform sampler2DMS%s texSampler;\n"
404 "out vec4 out_color;\n"
408 " gl_FragDepth = texelFetch(texSampler, i%s(%s), %s).r;\n"
410 sampler_array_suffix
,
416 /* You can create 2D_MULTISAMPLE textures with 0 sample count (meaning 1
417 * sample). Yes, this is ridiculous.
419 char *sample_resolve
;
420 const char *merge_function
;
421 name
= ralloc_asprintf(mem_ctx
, "%svec4 MSAA %s",
423 dst_is_msaa
? "copy" : "resolve");
426 const char *tex_coords
;
428 if (ctx
->Extensions
.ARB_gpu_shader5
&& samples
>= 16) {
429 /* If interpolateAtOffset is available then it will be used to
430 * force the interpolation to the center. This is required at
431 * least on Intel hardware because it is possible to have a sample
432 * position on the 0 x or y axis which means it will lie exactly
433 * on the pixel boundary. If we let the hardware interpolate the
434 * coordinates at one of these positions then it is possible for
435 * it to jump to a neighboring texel when converting to ints due
436 * to rounding errors. This is only done for >= 16x MSAA because
437 * it probably has some overhead. It is more likely that some
438 * hardware will use one of these problematic positions at 16x
439 * MSAA because in that case in D3D they are defined to be at
442 tex_coords
= "interpolateAtOffset(texCoords, vec2(0.0))";
444 tex_coords
= "texCoords";
448 ralloc_asprintf(mem_ctx
,
449 " out_color = texelFetch(texSampler, "
450 "i%s(%s), gl_SampleID);",
451 texcoord_type
, tex_coords
);
458 if (src_datatype
== GL_INT
|| src_datatype
== GL_UNSIGNED_INT
) {
460 "gvec4 merge(gvec4 a, gvec4 b) { return (a >> gvec4(1)) + (b >> gvec4(1)) + (a & b & gvec4(1)); }\n";
462 /* The divide will happen at the end for floats. */
464 "vec4 merge(vec4 a, vec4 b) { return (a + b); }\n";
467 /* We're assuming power of two samples for this resolution procedure.
469 * To avoid losing any floating point precision if the samples all
470 * happen to have the same value, we merge pairs of values at a time
471 * (so the floating point exponent just gets increased), rather than
472 * doing a naive sum and dividing.
474 assert(_mesa_is_pow_two(samples
));
475 /* Fetch each individual sample. */
476 sample_resolve
= rzalloc_size(mem_ctx
, 1);
477 for (i
= 0; i
< samples
; i
++) {
478 ralloc_asprintf_append(&sample_resolve
,
479 " gvec4 sample_1_%d = texelFetch(texSampler, i%s(texCoords), %d);\n",
480 i
, texcoord_type
, i
);
482 /* Now, merge each pair of samples, then merge each pair of those,
485 for (step
= 2; step
<= samples
; step
*= 2) {
486 for (i
= 0; i
< samples
; i
+= step
) {
487 ralloc_asprintf_append(&sample_resolve
,
488 " gvec4 sample_%d_%d = merge(sample_%d_%d, sample_%d_%d);\n",
491 step
/ 2, i
+ step
/ 2);
495 /* Scale the final result. */
496 if (src_datatype
== GL_UNSIGNED_INT
|| src_datatype
== GL_INT
) {
497 ralloc_asprintf_append(&sample_resolve
,
498 " out_color = sample_%d_0;\n",
501 ralloc_asprintf_append(&sample_resolve
,
502 " gl_FragColor = sample_%d_0 / %f;\n",
503 samples
, (float)samples
);
507 vs_source
= ralloc_asprintf(mem_ctx
,
509 "in vec2 position;\n"
510 "in %s textureCoords;\n"
511 "out %s texCoords;\n"
514 " texCoords = textureCoords;\n"
515 " gl_Position = vec4(position, 0.0, 1.0);\n"
519 fs_source
= ralloc_asprintf(mem_ctx
,
521 "#extension GL_ARB_texture_multisample : enable\n"
522 "#extension GL_ARB_sample_shading : enable\n"
523 "#extension GL_ARB_gpu_shader5 : enable\n"
524 "#define gvec4 %svec4\n"
525 "uniform %ssampler2DMS%s texSampler;\n"
527 "out gvec4 out_color;\n"
529 "%s" /* merge_function */
532 "%s\n" /* sample_resolve */
536 sampler_array_suffix
,
542 _mesa_meta_compile_and_link_program(ctx
, vs_source
, fs_source
, name
,
543 &blit
->msaa_shaders
[shader_index
]);
545 ralloc_free(mem_ctx
);
549 setup_glsl_blit_framebuffer(struct gl_context
*ctx
,
550 struct blit_state
*blit
,
551 const struct gl_framebuffer
*drawFb
,
552 struct gl_renderbuffer
*src_rb
,
553 GLenum target
, GLenum filter
,
557 unsigned texcoord_size
;
558 bool is_target_multisample
= target
== GL_TEXTURE_2D_MULTISAMPLE
||
559 target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
;
560 bool is_filter_scaled_resolve
= filter
== GL_SCALED_RESOLVE_FASTEST_EXT
||
561 filter
== GL_SCALED_RESOLVE_NICEST_EXT
;
563 /* target = GL_TEXTURE_RECTANGLE is not supported in GLES 3.0 */
564 assert(_mesa_is_desktop_gl(ctx
) || target
== GL_TEXTURE_2D
);
566 texcoord_size
= 2 + (src_rb
->Depth
> 1 ? 1 : 0);
568 _mesa_meta_setup_vertex_objects(ctx
, &blit
->VAO
, &blit
->buf_obj
, true,
569 2, texcoord_size
, 0);
571 if (is_target_multisample
&& is_filter_scaled_resolve
&& is_scaled_blit
) {
572 setup_glsl_msaa_blit_scaled_shader(ctx
, blit
, src_rb
, target
, filter
);
573 } else if (is_target_multisample
) {
574 setup_glsl_msaa_blit_shader(ctx
, blit
, drawFb
, src_rb
, target
);
576 _mesa_meta_setup_blit_shader(ctx
, target
, do_depth
,
577 do_depth
? &blit
->shaders_with_depth
578 : &blit
->shaders_without_depth
);
583 * Try to do a color or depth glBlitFramebuffer using texturing.
585 * We can do this when the src renderbuffer is actually a texture, or when the
586 * driver exposes BindRenderbufferTexImage().
589 blitframebuffer_texture(struct gl_context
*ctx
,
590 const struct gl_framebuffer
*readFb
,
591 const struct gl_framebuffer
*drawFb
,
592 GLint srcX0
, GLint srcY0
, GLint srcX1
, GLint srcY1
,
593 GLint dstX0
, GLint dstY0
, GLint dstX1
, GLint dstY1
,
594 GLenum filter
, GLint flipX
, GLint flipY
,
595 GLboolean glsl_version
, GLboolean do_depth
)
597 int att_index
= do_depth
? BUFFER_DEPTH
: readFb
->_ColorReadBufferIndex
;
598 const struct gl_renderbuffer_attachment
*readAtt
=
599 &readFb
->Attachment
[att_index
];
600 struct blit_state
*blit
= &ctx
->Meta
->Blit
;
601 struct fb_tex_blit_state fb_tex_blit
;
602 const GLint dstX
= MIN2(dstX0
, dstX1
);
603 const GLint dstY
= MIN2(dstY0
, dstY1
);
604 const GLint dstW
= abs(dstX1
- dstX0
);
605 const GLint dstH
= abs(dstY1
- dstY0
);
606 const int srcW
= abs(srcX1
- srcX0
);
607 const int srcH
= abs(srcY1
- srcY0
);
608 bool scaled_blit
= false;
609 struct gl_texture_object
*texObj
;
612 struct gl_renderbuffer
*rb
= readAtt
->Renderbuffer
;
613 struct temp_texture
*meta_temp_texture
;
615 if (rb
->NumSamples
&& !ctx
->Extensions
.ARB_texture_multisample
)
618 _mesa_meta_fb_tex_blit_begin(ctx
, &fb_tex_blit
);
620 if (readAtt
->Texture
&&
621 (readAtt
->Texture
->Target
== GL_TEXTURE_2D
||
622 readAtt
->Texture
->Target
== GL_TEXTURE_RECTANGLE
||
623 readAtt
->Texture
->Target
== GL_TEXTURE_2D_MULTISAMPLE
||
624 readAtt
->Texture
->Target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
)) {
625 /* If there's a texture attached of a type we can handle, then just use
628 srcLevel
= readAtt
->TextureLevel
;
629 texObj
= readAtt
->Texture
;
630 target
= texObj
->Target
;
631 } else if (!readAtt
->Texture
&& ctx
->Driver
.BindRenderbufferTexImage
) {
632 if (!_mesa_meta_bind_rb_as_tex_image(ctx
, rb
, &fb_tex_blit
.tempTex
,
637 if (_mesa_is_winsys_fbo(readFb
)) {
639 srcY0
= rb
->Height
- srcY1
;
640 srcY1
= rb
->Height
- temp
;
644 GLenum tex_base_format
;
645 /* Fall back to doing a CopyTexSubImage to get the destination
646 * renderbuffer into a texture.
648 if (ctx
->Meta
->Blit
.no_ctsi_fallback
)
651 if (rb
->NumSamples
> 1)
655 meta_temp_texture
= _mesa_meta_get_temp_depth_texture(ctx
);
656 tex_base_format
= GL_DEPTH_COMPONENT
;
658 meta_temp_texture
= _mesa_meta_get_temp_texture(ctx
);
660 _mesa_base_tex_format(ctx
, rb
->InternalFormat
);
664 target
= meta_temp_texture
->Target
;
665 texObj
= _mesa_lookup_texture(ctx
, meta_temp_texture
->TexObj
);
666 if (texObj
== NULL
) {
670 _mesa_meta_setup_copypix_texture(ctx
, meta_temp_texture
,
683 fb_tex_blit
.baseLevelSave
= texObj
->BaseLevel
;
684 fb_tex_blit
.maxLevelSave
= texObj
->MaxLevel
;
685 fb_tex_blit
.stencilSamplingSave
= texObj
->StencilSampling
;
687 scaled_blit
= dstW
!= srcW
|| dstH
!= srcH
;
690 setup_glsl_blit_framebuffer(ctx
, blit
, drawFb
, rb
, target
, filter
, scaled_blit
,
694 _mesa_meta_setup_ff_tnl_for_blit(ctx
,
695 &ctx
->Meta
->Blit
.VAO
,
696 &ctx
->Meta
->Blit
.buf_obj
,
701 printf("Blit from texture!\n");
702 printf(" srcAtt %p dstAtt %p\n", readAtt, drawAtt);
703 printf(" srcTex %p dstText %p\n", texObj, drawAtt->Texture);
706 fb_tex_blit
.samp_obj
= _mesa_meta_setup_sampler(ctx
, texObj
, target
, filter
,
709 /* Always do our blits with no net sRGB decode or encode.
711 * However, if both the src and dst can be srgb decode/encoded, enable them
712 * so that we do any blending (from scaling or from MSAA resolves) in the
715 * Our choice of not doing any net encode/decode is from the GL 3.0
718 * "Blit operations bypass the fragment pipeline. The only fragment
719 * operations which affect a blit are the pixel ownership test and the
722 * The GL 4.4 specification disagrees and says that the sRGB part of the
723 * fragment pipeline applies, but this was found to break applications.
725 if (ctx
->Extensions
.EXT_texture_sRGB_decode
) {
726 if (_mesa_get_format_color_encoding(rb
->Format
) == GL_SRGB
&&
727 drawFb
->Visual
.sRGBCapable
) {
728 _mesa_set_sampler_srgb_decode(ctx
, fb_tex_blit
.samp_obj
,
730 _mesa_set_framebuffer_srgb(ctx
, GL_TRUE
);
732 _mesa_set_sampler_srgb_decode(ctx
, fb_tex_blit
.samp_obj
,
734 /* set_framebuffer_srgb was set by _mesa_meta_begin(). */
739 _mesa_TexEnvi(GL_TEXTURE_ENV
, GL_TEXTURE_ENV_MODE
, GL_REPLACE
);
740 _mesa_set_enable(ctx
, target
, GL_TRUE
);
743 /* Prepare vertex data (the VBO was previously created and bound) */
745 struct vertex verts
[4];
746 GLfloat s0
, t0
, s1
, t1
;
748 if (target
== GL_TEXTURE_2D
) {
749 const struct gl_texture_image
*texImage
750 = _mesa_select_tex_image(texObj
, target
, srcLevel
);
751 s0
= srcX0
/ (float) texImage
->Width
;
752 s1
= srcX1
/ (float) texImage
->Width
;
753 t0
= srcY0
/ (float) texImage
->Height
;
754 t1
= srcY1
/ (float) texImage
->Height
;
757 assert(target
== GL_TEXTURE_RECTANGLE_ARB
||
758 target
== GL_TEXTURE_2D_MULTISAMPLE
||
759 target
== GL_TEXTURE_2D_MULTISAMPLE_ARRAY
);
766 /* Silence valgrind warnings about reading uninitialized stack. */
767 memset(verts
, 0, sizeof(verts
));
769 /* setup vertex positions */
770 verts
[0].x
= -1.0F
* flipX
;
771 verts
[0].y
= -1.0F
* flipY
;
772 verts
[1].x
= 1.0F
* flipX
;
773 verts
[1].y
= -1.0F
* flipY
;
774 verts
[2].x
= 1.0F
* flipX
;
775 verts
[2].y
= 1.0F
* flipY
;
776 verts
[3].x
= -1.0F
* flipX
;
777 verts
[3].y
= 1.0F
* flipY
;
779 verts
[0].tex
[0] = s0
;
780 verts
[0].tex
[1] = t0
;
781 verts
[0].tex
[2] = readAtt
->Zoffset
;
782 verts
[1].tex
[0] = s1
;
783 verts
[1].tex
[1] = t0
;
784 verts
[1].tex
[2] = readAtt
->Zoffset
;
785 verts
[2].tex
[0] = s1
;
786 verts
[2].tex
[1] = t1
;
787 verts
[2].tex
[2] = readAtt
->Zoffset
;
788 verts
[3].tex
[0] = s0
;
789 verts
[3].tex
[1] = t1
;
790 verts
[3].tex
[2] = readAtt
->Zoffset
;
792 _mesa_buffer_sub_data(ctx
, blit
->buf_obj
, 0, sizeof(verts
), verts
,
797 _mesa_set_viewport(ctx
, 0, dstX
, dstY
, dstW
, dstH
);
798 _mesa_ColorMask(!do_depth
, !do_depth
, !do_depth
, !do_depth
);
799 _mesa_set_enable(ctx
, GL_DEPTH_TEST
, do_depth
);
800 _mesa_DepthMask(do_depth
);
801 _mesa_DepthFunc(GL_ALWAYS
);
803 _mesa_DrawArrays(GL_TRIANGLE_FAN
, 0, 4);
804 _mesa_meta_fb_tex_blit_end(ctx
, target
, &fb_tex_blit
);
810 _mesa_meta_fb_tex_blit_begin(struct gl_context
*ctx
,
811 struct fb_tex_blit_state
*blit
)
813 /* None of the existing callers preinitialize fb_tex_blit_state to zeros,
814 * and both use stack variables. If samp_obj_save is not NULL,
815 * _mesa_reference_sampler_object will try to dereference it. Leaving
816 * random garbage in samp_obj_save can only lead to crashes.
818 * Since the state isn't persistent across calls, we won't catch ref
821 blit
->samp_obj_save
= NULL
;
822 _mesa_reference_sampler_object(ctx
, &blit
->samp_obj_save
,
823 ctx
->Texture
.Unit
[ctx
->Texture
.CurrentUnit
].Sampler
);
828 _mesa_meta_fb_tex_blit_end(struct gl_context
*ctx
, GLenum target
,
829 struct fb_tex_blit_state
*blit
)
831 struct gl_texture_object
*const texObj
=
832 _mesa_get_current_tex_object(ctx
, target
);
834 /* Restore texture object state, the texture binding will
835 * be restored by _mesa_meta_end().
837 if (target
!= GL_TEXTURE_RECTANGLE_ARB
) {
838 _mesa_texture_parameteriv(ctx
, texObj
, GL_TEXTURE_BASE_LEVEL
,
839 &blit
->baseLevelSave
, false);
840 _mesa_texture_parameteriv(ctx
, texObj
, GL_TEXTURE_MAX_LEVEL
,
841 &blit
->maxLevelSave
, false);
844 /* If ARB_stencil_texturing is not supported, the mode won't have changed. */
845 if (texObj
->StencilSampling
!= blit
->stencilSamplingSave
) {
846 /* GLint so the compiler won't complain about type signedness mismatch
847 * in the call to _mesa_texture_parameteriv below.
849 const GLint param
= blit
->stencilSamplingSave
?
850 GL_STENCIL_INDEX
: GL_DEPTH_COMPONENT
;
852 _mesa_texture_parameteriv(ctx
, texObj
, GL_DEPTH_STENCIL_TEXTURE_MODE
,
856 _mesa_bind_sampler(ctx
, ctx
->Texture
.CurrentUnit
, blit
->samp_obj_save
);
857 _mesa_reference_sampler_object(ctx
, &blit
->samp_obj_save
, NULL
);
858 _mesa_reference_sampler_object(ctx
, &blit
->samp_obj
, NULL
);
861 _mesa_DeleteTextures(1, &blit
->tempTex
);
865 _mesa_meta_bind_rb_as_tex_image(struct gl_context
*ctx
,
866 struct gl_renderbuffer
*rb
,
868 struct gl_texture_object
**texObj
,
871 struct gl_texture_image
*texImage
;
874 if (rb
->NumSamples
> 1)
875 *target
= GL_TEXTURE_2D_MULTISAMPLE
;
877 *target
= GL_TEXTURE_2D
;
880 _mesa_GenTextures(1, &tempTex
);
886 _mesa_BindTexture(*target
, *tex
);
887 *texObj
= _mesa_lookup_texture(ctx
, *tex
);
888 texImage
= _mesa_get_tex_image(ctx
, *texObj
, *target
, 0);
890 if (!ctx
->Driver
.BindRenderbufferTexImage(ctx
, rb
, texImage
)) {
891 _mesa_DeleteTextures(1, tex
);
895 if (ctx
->Driver
.FinishRenderTexture
&& !rb
->NeedsFinishRenderTexture
) {
896 rb
->NeedsFinishRenderTexture
= true;
897 ctx
->Driver
.FinishRenderTexture(ctx
, rb
);
903 struct gl_sampler_object
*
904 _mesa_meta_setup_sampler(struct gl_context
*ctx
,
905 struct gl_texture_object
*texObj
,
906 GLenum target
, GLenum filter
, GLuint srcLevel
)
908 struct gl_sampler_object
*samp_obj
;
909 GLenum tex_filter
= (filter
== GL_SCALED_RESOLVE_FASTEST_EXT
||
910 filter
== GL_SCALED_RESOLVE_NICEST_EXT
) ?
913 samp_obj
= ctx
->Driver
.NewSamplerObject(ctx
, 0xDEADBEEF);
914 if (samp_obj
== NULL
)
917 _mesa_bind_sampler(ctx
, ctx
->Texture
.CurrentUnit
, samp_obj
);
918 _mesa_set_sampler_filters(ctx
, samp_obj
, tex_filter
, tex_filter
);
919 _mesa_set_sampler_wrap(ctx
, samp_obj
, GL_CLAMP_TO_EDGE
, GL_CLAMP_TO_EDGE
,
922 /* Prepare src texture state */
923 _mesa_BindTexture(target
, texObj
->Name
);
924 if (target
!= GL_TEXTURE_RECTANGLE_ARB
) {
925 _mesa_texture_parameteriv(ctx
, texObj
, GL_TEXTURE_BASE_LEVEL
,
926 (GLint
*) &srcLevel
, false);
927 _mesa_texture_parameteriv(ctx
, texObj
, GL_TEXTURE_MAX_LEVEL
,
928 (GLint
*) &srcLevel
, false);
935 * Meta implementation of ctx->Driver.BlitFramebuffer() in terms
936 * of texture mapping and polygon rendering.
939 _mesa_meta_BlitFramebuffer(struct gl_context
*ctx
,
940 const struct gl_framebuffer
*readFb
,
941 const struct gl_framebuffer
*drawFb
,
942 GLint srcX0
, GLint srcY0
, GLint srcX1
, GLint srcY1
,
943 GLint dstX0
, GLint dstY0
, GLint dstX1
, GLint dstY1
,
944 GLbitfield mask
, GLenum filter
)
946 const GLint dstW
= abs(dstX1
- dstX0
);
947 const GLint dstH
= abs(dstY1
- dstY0
);
948 const GLint dstFlipX
= (dstX1
- dstX0
) / dstW
;
949 const GLint dstFlipY
= (dstY1
- dstY0
) / dstH
;
952 GLint srcX0
, srcY0
, srcX1
, srcY1
;
953 GLint dstX0
, dstY0
, dstX1
, dstY1
;
955 srcX0
, srcY0
, srcX1
, srcY1
,
956 dstX0
, dstY0
, dstX1
, dstY1
959 const GLboolean use_glsl_version
= ctx
->Extensions
.ARB_vertex_shader
&&
960 ctx
->Extensions
.ARB_fragment_shader
;
962 /* Multisample texture blit support requires texture multisample. */
963 if (readFb
->Visual
.samples
> 0 &&
964 !ctx
->Extensions
.ARB_texture_multisample
) {
968 /* Clip a copy of the blit coordinates. If these differ from the input
969 * coordinates, then we'll set the scissor.
971 if (!_mesa_clip_blit(ctx
, readFb
, drawFb
,
972 &clip
.srcX0
, &clip
.srcY0
, &clip
.srcX1
, &clip
.srcY1
,
973 &clip
.dstX0
, &clip
.dstY0
, &clip
.dstX1
, &clip
.dstY1
)) {
974 /* clipped/scissored everything away */
978 /* Only scissor affects blit, but we're doing to set a custom scissor if
979 * necessary anyway, so save/clear state.
981 _mesa_meta_begin(ctx
, MESA_META_ALL
& ~MESA_META_DRAW_BUFFERS
);
983 /* Dithering shouldn't be performed for glBlitFramebuffer */
984 _mesa_set_enable(ctx
, GL_DITHER
, GL_FALSE
);
986 /* If the clipping earlier changed the destination rect at all, then
987 * enable the scissor to clip to it.
989 if (clip
.dstX0
!= dstX0
|| clip
.dstY0
!= dstY0
||
990 clip
.dstX1
!= dstX1
|| clip
.dstY1
!= dstY1
) {
991 _mesa_set_enable(ctx
, GL_SCISSOR_TEST
, GL_TRUE
);
992 _mesa_Scissor(MIN2(clip
.dstX0
, clip
.dstX1
),
993 MIN2(clip
.dstY0
, clip
.dstY1
),
994 abs(clip
.dstX0
- clip
.dstX1
),
995 abs(clip
.dstY0
- clip
.dstY1
));
998 /* Try faster, direct texture approach first */
999 if (mask
& GL_COLOR_BUFFER_BIT
) {
1000 if (blitframebuffer_texture(ctx
, readFb
, drawFb
,
1001 srcX0
, srcY0
, srcX1
, srcY1
,
1002 dstX0
, dstY0
, dstX1
, dstY1
,
1003 filter
, dstFlipX
, dstFlipY
,
1004 use_glsl_version
, false)) {
1005 mask
&= ~GL_COLOR_BUFFER_BIT
;
1009 if (mask
& GL_DEPTH_BUFFER_BIT
&& use_glsl_version
) {
1010 if (blitframebuffer_texture(ctx
, readFb
, drawFb
,
1011 srcX0
, srcY0
, srcX1
, srcY1
,
1012 dstX0
, dstY0
, dstX1
, dstY1
,
1013 filter
, dstFlipX
, dstFlipY
,
1014 use_glsl_version
, true)) {
1015 mask
&= ~GL_DEPTH_BUFFER_BIT
;
1019 if (mask
& GL_STENCIL_BUFFER_BIT
) {
1020 /* XXX can't easily do stencil */
1023 _mesa_meta_end(ctx
);
1029 _mesa_meta_glsl_blit_cleanup(struct gl_context
*ctx
, struct blit_state
*blit
)
1032 _mesa_DeleteVertexArrays(1, &blit
->VAO
);
1034 _mesa_reference_buffer_object(ctx
, &blit
->buf_obj
, NULL
);
1037 _mesa_meta_blit_shader_table_cleanup(&blit
->shaders_with_depth
);
1038 _mesa_meta_blit_shader_table_cleanup(&blit
->shaders_without_depth
);
1040 _mesa_DeleteTextures(1, &blit
->depthTex
.TexObj
);
1041 blit
->depthTex
.TexObj
= 0;
1045 _mesa_meta_and_swrast_BlitFramebuffer(struct gl_context
*ctx
,
1046 struct gl_framebuffer
*readFb
,
1047 struct gl_framebuffer
*drawFb
,
1048 GLint srcX0
, GLint srcY0
,
1049 GLint srcX1
, GLint srcY1
,
1050 GLint dstX0
, GLint dstY0
,
1051 GLint dstX1
, GLint dstY1
,
1052 GLbitfield mask
, GLenum filter
)
1054 mask
= _mesa_meta_BlitFramebuffer(ctx
, readFb
, drawFb
,
1055 srcX0
, srcY0
, srcX1
, srcY1
,
1056 dstX0
, dstY0
, dstX1
, dstY1
,
1061 _swrast_BlitFramebuffer(ctx
, readFb
, drawFb
,
1062 srcX0
, srcY0
, srcX1
, srcY1
,
1063 dstX0
, dstY0
, dstX1
, dstY1
,