1 /**************************************************************************
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
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:
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
26 **************************************************************************/
29 #include "main/mtypes.h"
30 #include "main/context.h"
31 #include "main/enums.h"
32 #include "main/colormac.h"
33 #include "main/fbobject.h"
35 #include "intel_blit.h"
36 #include "intel_buffers.h"
37 #include "intel_context.h"
38 #include "intel_fbo.h"
39 #include "intel_reg.h"
40 #include "intel_regions.h"
41 #include "intel_batchbuffer.h"
42 #include "intel_mipmap_tree.h"
44 #define FILE_DEBUG_FLAG DEBUG_BLIT
46 static GLuint
translate_raster_op(GLenum logicop
)
49 case GL_CLEAR
: return 0x00;
50 case GL_AND
: return 0x88;
51 case GL_AND_REVERSE
: return 0x44;
52 case GL_COPY
: return 0xCC;
53 case GL_AND_INVERTED
: return 0x22;
54 case GL_NOOP
: return 0xAA;
55 case GL_XOR
: return 0x66;
56 case GL_OR
: return 0xEE;
57 case GL_NOR
: return 0x11;
58 case GL_EQUIV
: return 0x99;
59 case GL_INVERT
: return 0x55;
60 case GL_OR_REVERSE
: return 0xDD;
61 case GL_COPY_INVERTED
: return 0x33;
62 case GL_OR_INVERTED
: return 0xBB;
63 case GL_NAND
: return 0x77;
64 case GL_SET
: return 0xFF;
91 intelEmitCopyBlit(struct intel_context
*intel
,
94 drm_intel_bo
*src_buffer
,
98 drm_intel_bo
*dst_buffer
,
101 GLshort src_x
, GLshort src_y
,
102 GLshort dst_x
, GLshort dst_y
,
103 GLshort w
, GLshort h
,
106 GLuint CMD
, BR13
, pass
= 0;
107 int dst_y2
= dst_y
+ h
;
108 int dst_x2
= dst_x
+ w
;
109 drm_intel_bo
*aper_array
[3];
112 if (dst_tiling
!= I915_TILING_NONE
) {
113 if (dst_offset
& 4095)
115 if (dst_tiling
== I915_TILING_Y
)
118 if (src_tiling
!= I915_TILING_NONE
) {
119 if (src_offset
& 4095)
121 if (src_tiling
== I915_TILING_Y
)
125 /* do space check before going any further */
127 aper_array
[0] = intel
->batch
.bo
;
128 aper_array
[1] = dst_buffer
;
129 aper_array
[2] = src_buffer
;
131 if (dri_bufmgr_check_aperture_space(aper_array
, 3) != 0) {
132 intel_batchbuffer_flush(intel
);
141 intel_batchbuffer_require_space(intel
, 8 * 4, true);
142 DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
144 src_buffer
, src_pitch
, src_offset
, src_x
, src_y
,
145 dst_buffer
, dst_pitch
, dst_offset
, dst_x
, dst_y
, w
, h
);
150 /* Blit pitch must be dword-aligned. Otherwise, the hardware appears to drop
153 assert(src_pitch
% 4 == 0);
154 assert(dst_pitch
% 4 == 0);
156 /* For big formats (such as floating point), do the copy using 32bpp and
157 * multiply the coordinates.
160 assert(cpp
% 4 == 0);
167 BR13
= br13_for_cpp(cpp
) | translate_raster_op(logic_op
) << 16;
172 CMD
= XY_SRC_COPY_BLT_CMD
;
175 CMD
= XY_SRC_COPY_BLT_CMD
| XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
182 if (dst_tiling
!= I915_TILING_NONE
) {
186 if (src_tiling
!= I915_TILING_NONE
) {
192 if (dst_y2
<= dst_y
|| dst_x2
<= dst_x
) {
196 assert(dst_x
< dst_x2
);
197 assert(dst_y
< dst_y2
);
201 OUT_BATCH(BR13
| (uint16_t)dst_pitch
);
202 OUT_BATCH((dst_y
<< 16) | dst_x
);
203 OUT_BATCH((dst_y2
<< 16) | dst_x2
);
204 OUT_RELOC_FENCED(dst_buffer
,
205 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
207 OUT_BATCH((src_y
<< 16) | src_x
);
208 OUT_BATCH((uint16_t)src_pitch
);
209 OUT_RELOC_FENCED(src_buffer
,
210 I915_GEM_DOMAIN_RENDER
, 0,
214 intel_batchbuffer_emit_mi_flush(intel
);
221 * Use blitting to clear the renderbuffers named by 'flags'.
222 * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field
223 * since that might include software renderbuffers or renderbuffers
224 * which we're clearing with triangles.
225 * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear
228 intelClearWithBlit(struct gl_context
*ctx
, GLbitfield mask
)
230 struct intel_context
*intel
= intel_context(ctx
);
231 struct gl_framebuffer
*fb
= ctx
->DrawBuffer
;
232 GLuint clear_depth_value
, clear_depth_mask
;
233 GLint cx
, cy
, cw
, ch
;
234 GLbitfield fail_mask
= 0;
238 * Compute values for clearing the buffers.
240 clear_depth_value
= 0;
241 clear_depth_mask
= 0;
242 if (mask
& BUFFER_BIT_DEPTH
) {
243 clear_depth_value
= (GLuint
) (fb
->_DepthMax
* ctx
->Depth
.Clear
);
244 clear_depth_mask
= XY_BLT_WRITE_RGB
;
246 if (mask
& BUFFER_BIT_STENCIL
) {
247 clear_depth_value
|= (ctx
->Stencil
.Clear
& 0xff) << 24;
248 clear_depth_mask
|= XY_BLT_WRITE_ALPHA
;
252 if (_mesa_is_winsys_fbo(fb
))
253 cy
= ctx
->DrawBuffer
->Height
- fb
->_Ymax
;
256 cw
= fb
->_Xmax
- fb
->_Xmin
;
257 ch
= fb
->_Ymax
- fb
->_Ymin
;
259 if (cw
== 0 || ch
== 0)
262 /* Loop over all renderbuffers */
263 mask
&= (1 << BUFFER_COUNT
) - 1;
265 GLuint buf
= ffs(mask
) - 1;
266 bool is_depth_stencil
= buf
== BUFFER_DEPTH
|| buf
== BUFFER_STENCIL
;
267 struct intel_renderbuffer
*irb
;
271 struct intel_region
*region
;
273 drm_intel_bo
*aper_array
[2];
277 irb
= intel_get_renderbuffer(fb
, buf
);
278 if (irb
&& irb
->mt
) {
279 region
= irb
->mt
->region
;
283 fail_mask
|= 1 << buf
;
287 /* OK, clear this renderbuffer */
288 x1
= cx
+ irb
->draw_x
;
289 y1
= cy
+ irb
->draw_y
;
290 x2
= cx
+ cw
+ irb
->draw_x
;
291 y2
= cy
+ ch
+ irb
->draw_y
;
293 pitch
= region
->pitch
;
296 DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
298 region
->bo
, (pitch
* cpp
),
299 x1
, y1
, x2
- x1
, y2
- y1
);
302 CMD
= XY_COLOR_BLT_CMD
;
304 /* Setup the blit command */
306 if (is_depth_stencil
) {
307 CMD
|= clear_depth_mask
;
310 CMD
|= XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
314 assert(region
->tiling
!= I915_TILING_Y
);
317 if (region
->tiling
!= I915_TILING_NONE
) {
322 BR13
|= (pitch
* cpp
);
324 if (is_depth_stencil
) {
325 clear_val
= clear_depth_value
;
328 GLfloat
*color
= ctx
->Color
.ClearColor
.f
;
330 _mesa_unclamped_float_rgba_to_ubyte(clear
, color
);
332 switch (intel_rb_format(irb
)) {
333 case MESA_FORMAT_ARGB8888
:
334 case MESA_FORMAT_XRGB8888
:
335 clear_val
= PACK_COLOR_8888(clear
[3], clear
[0],
338 case MESA_FORMAT_RGB565
:
339 clear_val
= PACK_COLOR_565(clear
[0], clear
[1], clear
[2]);
341 case MESA_FORMAT_ARGB4444
:
342 clear_val
= PACK_COLOR_4444(clear
[3], clear
[0],
345 case MESA_FORMAT_ARGB1555
:
346 clear_val
= PACK_COLOR_1555(clear
[3], clear
[0],
350 clear_val
= PACK_COLOR_8888(clear
[3], clear
[3],
354 fail_mask
|= 1 << buf
;
359 BR13
|= br13_for_cpp(cpp
);
364 /* do space check before going any further */
365 aper_array
[0] = intel
->batch
.bo
;
366 aper_array
[1] = region
->bo
;
368 if (drm_intel_bufmgr_check_aperture_space(aper_array
,
369 ARRAY_SIZE(aper_array
)) != 0) {
370 intel_batchbuffer_flush(intel
);
376 OUT_BATCH((y1
<< 16) | x1
);
377 OUT_BATCH((y2
<< 16) | x2
);
378 OUT_RELOC_FENCED(region
->bo
,
379 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
381 OUT_BATCH(clear_val
);
384 if (intel
->always_flush_cache
)
385 intel_batchbuffer_emit_mi_flush(intel
);
387 if (buf
== BUFFER_DEPTH
|| buf
== BUFFER_STENCIL
)
388 mask
&= ~(BUFFER_BIT_DEPTH
| BUFFER_BIT_STENCIL
);
395 intelEmitImmediateColorExpandBlit(struct intel_context
*intel
,
397 GLubyte
*src_bits
, GLuint src_size
,
400 drm_intel_bo
*dst_buffer
,
403 GLshort x
, GLshort y
,
404 GLshort w
, GLshort h
,
407 int dwords
= ALIGN(src_size
, 8) / 4;
408 uint32_t opcode
, br13
, blit_cmd
;
410 if (dst_tiling
!= I915_TILING_NONE
) {
411 if (dst_offset
& 4095)
413 if (dst_tiling
== I915_TILING_Y
)
417 assert( logic_op
- GL_CLEAR
>= 0 );
418 assert( logic_op
- GL_CLEAR
< 0x10 );
419 assert(dst_pitch
> 0);
426 DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
428 dst_buffer
, dst_pitch
, dst_offset
, x
, y
, w
, h
, src_size
, dwords
);
430 intel_batchbuffer_require_space(intel
,
435 opcode
= XY_SETUP_BLT_CMD
;
437 opcode
|= XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
439 if (dst_tiling
!= I915_TILING_NONE
) {
440 opcode
|= XY_DST_TILED
;
445 br13
= dst_pitch
| (translate_raster_op(logic_op
) << 16) | (1 << 29);
446 br13
|= br13_for_cpp(cpp
);
448 blit_cmd
= XY_TEXT_IMMEDIATE_BLIT_CMD
| XY_TEXT_BYTE_PACKED
; /* packing? */
449 if (dst_tiling
!= I915_TILING_NONE
)
450 blit_cmd
|= XY_DST_TILED
;
452 BEGIN_BATCH_BLT(8 + 3);
455 OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
456 OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
457 OUT_RELOC_FENCED(dst_buffer
,
458 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
460 OUT_BATCH(0); /* bg */
461 OUT_BATCH(fg_color
); /* fg */
462 OUT_BATCH(0); /* pattern base addr */
464 OUT_BATCH(blit_cmd
| ((3 - 2) + dwords
));
465 OUT_BATCH((y
<< 16) | x
);
466 OUT_BATCH(((y
+ h
) << 16) | (x
+ w
));
469 intel_batchbuffer_data(intel
, src_bits
, dwords
* 4, true);
471 intel_batchbuffer_emit_mi_flush(intel
);
476 /* We don't have a memmove-type blit like some other hardware, so we'll do a
477 * rectangular blit covering a large space, then emit 1-scanline blit at the
478 * end to cover the last if we need.
481 intel_emit_linear_blit(struct intel_context
*intel
,
482 drm_intel_bo
*dst_bo
,
483 unsigned int dst_offset
,
484 drm_intel_bo
*src_bo
,
485 unsigned int src_offset
,
488 GLuint pitch
, height
;
491 /* The pitch given to the GPU must be DWORD aligned, and
492 * we want width to match pitch. Max width is (1 << 15 - 1),
493 * rounding that down to the nearest DWORD is 1 << 15 - 4
495 pitch
= ROUND_DOWN_TO(MIN2(size
, (1 << 15) - 1), 4);
496 height
= (pitch
== 0) ? 1 : size
/ pitch
;
497 ok
= intelEmitCopyBlit(intel
, 1,
498 pitch
, src_bo
, src_offset
, I915_TILING_NONE
,
499 pitch
, dst_bo
, dst_offset
, I915_TILING_NONE
,
502 pitch
, height
, /* w, h */
506 src_offset
+= pitch
* height
;
507 dst_offset
+= pitch
* height
;
508 size
-= pitch
* height
;
509 assert (size
< (1 << 15));
510 pitch
= ALIGN(size
, 4);
512 ok
= intelEmitCopyBlit(intel
, 1,
513 pitch
, src_bo
, src_offset
, I915_TILING_NONE
,
514 pitch
, dst_bo
, dst_offset
, I915_TILING_NONE
,
524 * Used to initialize the alpha value of an ARGB8888 teximage after
525 * loading it from an XRGB8888 source.
527 * This is very common with glCopyTexImage2D().
530 intel_set_teximage_alpha_to_one(struct gl_context
*ctx
,
531 struct intel_texture_image
*intel_image
)
533 struct intel_context
*intel
= intel_context(ctx
);
534 unsigned int image_x
, image_y
;
535 uint32_t x1
, y1
, x2
, y2
;
538 drm_intel_bo
*aper_array
[2];
539 struct intel_region
*region
= intel_image
->mt
->region
;
540 int width
, height
, depth
;
543 intel_miptree_get_dimensions_for_image(&intel_image
->base
.Base
,
544 &width
, &height
, &depth
);
547 assert(intel_image
->base
.Base
.TexFormat
== MESA_FORMAT_ARGB8888
);
549 /* get dest x/y in destination texture */
550 intel_miptree_get_image_offset(intel_image
->mt
,
551 intel_image
->base
.Base
.Level
,
552 intel_image
->base
.Base
.Face
,
558 x2
= image_x
+ width
;
559 y2
= image_y
+ height
;
561 pitch
= region
->pitch
;
564 DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
566 intel_image
->mt
->region
->bo
, (pitch
* cpp
),
567 x1
, y1
, x2
- x1
, y2
- y1
);
569 BR13
= br13_for_cpp(cpp
) | 0xf0 << 16;
570 CMD
= XY_COLOR_BLT_CMD
;
571 CMD
|= XY_BLT_WRITE_ALPHA
;
573 assert(region
->tiling
!= I915_TILING_Y
);
576 if (region
->tiling
!= I915_TILING_NONE
) {
581 BR13
|= (pitch
* cpp
);
583 /* do space check before going any further */
584 aper_array
[0] = intel
->batch
.bo
;
585 aper_array
[1] = region
->bo
;
587 if (drm_intel_bufmgr_check_aperture_space(aper_array
,
588 ARRAY_SIZE(aper_array
)) != 0) {
589 intel_batchbuffer_flush(intel
);
595 OUT_BATCH((y1
<< 16) | x1
);
596 OUT_BATCH((y2
<< 16) | x2
);
597 OUT_RELOC_FENCED(region
->bo
,
598 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
600 OUT_BATCH(0xffffffff); /* white, but only alpha gets written */
603 intel_batchbuffer_emit_mi_flush(intel
);