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"
34 #include "intel_blit.h"
35 #include "intel_buffers.h"
36 #include "intel_context.h"
37 #include "intel_fbo.h"
38 #include "intel_reg.h"
39 #include "intel_regions.h"
40 #include "intel_batchbuffer.h"
42 #define FILE_DEBUG_FLAG DEBUG_BLIT
44 static GLuint
translate_raster_op(GLenum logicop
)
47 case GL_CLEAR
: return 0x00;
48 case GL_AND
: return 0x88;
49 case GL_AND_REVERSE
: return 0x44;
50 case GL_COPY
: return 0xCC;
51 case GL_AND_INVERTED
: return 0x22;
52 case GL_NOOP
: return 0xAA;
53 case GL_XOR
: return 0x66;
54 case GL_OR
: return 0xEE;
55 case GL_NOR
: return 0x11;
56 case GL_EQUIV
: return 0x99;
57 case GL_INVERT
: return 0x55;
58 case GL_OR_REVERSE
: return 0xDD;
59 case GL_COPY_INVERTED
: return 0x33;
60 case GL_OR_INVERTED
: return 0xBB;
61 case GL_NAND
: return 0x77;
62 case GL_SET
: return 0xFF;
89 intelEmitCopyBlit(struct intel_context
*intel
,
92 drm_intel_bo
*src_buffer
,
96 drm_intel_bo
*dst_buffer
,
99 GLshort src_x
, GLshort src_y
,
100 GLshort dst_x
, GLshort dst_y
,
101 GLshort w
, GLshort h
,
104 GLuint CMD
, BR13
, pass
= 0;
105 int dst_y2
= dst_y
+ h
;
106 int dst_x2
= dst_x
+ w
;
107 drm_intel_bo
*aper_array
[3];
110 if (dst_tiling
!= I915_TILING_NONE
) {
111 if (dst_offset
& 4095)
113 if (dst_tiling
== I915_TILING_Y
)
116 if (src_tiling
!= I915_TILING_NONE
) {
117 if (src_offset
& 4095)
119 if (src_tiling
== I915_TILING_Y
)
123 /* do space check before going any further */
125 aper_array
[0] = intel
->batch
->buf
;
126 aper_array
[1] = dst_buffer
;
127 aper_array
[2] = src_buffer
;
129 if (dri_bufmgr_check_aperture_space(aper_array
, 3) != 0) {
130 intel_batchbuffer_flush(intel
->batch
);
139 intel_batchbuffer_require_space(intel
->batch
, 8 * 4, true);
140 DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
142 src_buffer
, src_pitch
, src_offset
, src_x
, src_y
,
143 dst_buffer
, dst_pitch
, dst_offset
, dst_x
, dst_y
, w
, h
);
148 BR13
= br13_for_cpp(cpp
) | translate_raster_op(logic_op
) << 16;
153 CMD
= XY_SRC_COPY_BLT_CMD
;
156 CMD
= XY_SRC_COPY_BLT_CMD
| XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
163 if (dst_tiling
!= I915_TILING_NONE
) {
167 if (src_tiling
!= I915_TILING_NONE
) {
173 if (dst_y2
<= dst_y
|| dst_x2
<= dst_x
) {
177 assert(dst_x
< dst_x2
);
178 assert(dst_y
< dst_y2
);
182 OUT_BATCH(BR13
| (uint16_t)dst_pitch
);
183 OUT_BATCH((dst_y
<< 16) | dst_x
);
184 OUT_BATCH((dst_y2
<< 16) | dst_x2
);
185 OUT_RELOC_FENCED(dst_buffer
,
186 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
188 OUT_BATCH((src_y
<< 16) | src_x
);
189 OUT_BATCH((uint16_t)src_pitch
);
190 OUT_RELOC_FENCED(src_buffer
,
191 I915_GEM_DOMAIN_RENDER
, 0,
195 intel_batchbuffer_emit_mi_flush(intel
->batch
);
202 * Use blitting to clear the renderbuffers named by 'flags'.
203 * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field
204 * since that might include software renderbuffers or renderbuffers
205 * which we're clearing with triangles.
206 * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear
209 intelClearWithBlit(struct gl_context
*ctx
, GLbitfield mask
)
211 struct intel_context
*intel
= intel_context(ctx
);
212 struct gl_framebuffer
*fb
= ctx
->DrawBuffer
;
215 GLint cx
, cy
, cw
, ch
;
219 * Compute values for clearing the buffers.
222 if (mask
& BUFFER_BIT_DEPTH
) {
223 clear_depth
= (GLuint
) (fb
->_DepthMax
* ctx
->Depth
.Clear
);
225 if (mask
& BUFFER_BIT_STENCIL
) {
226 clear_depth
|= (ctx
->Stencil
.Clear
& 0xff) << 24;
231 cy
= ctx
->DrawBuffer
->Height
- fb
->_Ymax
;
234 cw
= fb
->_Xmax
- fb
->_Xmin
;
235 ch
= fb
->_Ymax
- fb
->_Ymin
;
237 if (cw
== 0 || ch
== 0)
241 all
= (cw
== fb
->Width
&& ch
== fb
->Height
);
243 /* Loop over all renderbuffers */
244 for (buf
= 0; buf
< BUFFER_COUNT
&& mask
; buf
++) {
245 const GLbitfield bufBit
= 1 << buf
;
246 struct intel_renderbuffer
*irb
;
247 drm_intel_bo
*write_buffer
;
252 drm_intel_bo
*aper_array
[2];
254 if (!(mask
& bufBit
))
257 /* OK, clear this renderbuffer */
258 irb
= intel_get_renderbuffer(fb
, buf
);
259 write_buffer
= intel_region_buffer(intel
, irb
->region
,
260 all
? INTEL_WRITE_FULL
:
262 x1
= cx
+ irb
->region
->draw_x
;
263 y1
= cy
+ irb
->region
->draw_y
;
264 x2
= cx
+ cw
+ irb
->region
->draw_x
;
265 y2
= cy
+ ch
+ irb
->region
->draw_y
;
267 pitch
= irb
->region
->pitch
;
268 cpp
= irb
->region
->cpp
;
270 DBG("%s dst:buf(%p)/%d %d,%d sz:%dx%d\n",
272 irb
->region
->buffer
, (pitch
* cpp
),
273 x1
, y1
, x2
- x1
, y2
- y1
);
275 BR13
= br13_for_cpp(cpp
) | 0xf0 << 16;
276 CMD
= XY_COLOR_BLT_CMD
;
278 /* Setup the blit command */
280 if (buf
== BUFFER_DEPTH
|| buf
== BUFFER_STENCIL
) {
281 if (mask
& BUFFER_BIT_DEPTH
)
282 CMD
|= XY_BLT_WRITE_RGB
;
283 if (mask
& BUFFER_BIT_STENCIL
)
284 CMD
|= XY_BLT_WRITE_ALPHA
;
287 CMD
|= XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
291 assert(irb
->region
->tiling
!= I915_TILING_Y
);
294 if (irb
->region
->tiling
!= I915_TILING_NONE
) {
299 BR13
|= (pitch
* cpp
);
301 if (buf
== BUFFER_DEPTH
|| buf
== BUFFER_STENCIL
) {
302 clear_val
= clear_depth
;
305 GLclampf
*color
= ctx
->Color
.ClearColor
;
307 CLAMPED_FLOAT_TO_UBYTE(clear
[0], color
[0]);
308 CLAMPED_FLOAT_TO_UBYTE(clear
[1], color
[1]);
309 CLAMPED_FLOAT_TO_UBYTE(clear
[2], color
[2]);
310 CLAMPED_FLOAT_TO_UBYTE(clear
[3], color
[3]);
312 switch (irb
->Base
.Format
) {
313 case MESA_FORMAT_ARGB8888
:
314 case MESA_FORMAT_XRGB8888
:
315 clear_val
= PACK_COLOR_8888(clear
[3], clear
[0],
318 case MESA_FORMAT_RGB565
:
319 clear_val
= PACK_COLOR_565(clear
[0], clear
[1], clear
[2]);
321 case MESA_FORMAT_ARGB4444
:
322 clear_val
= PACK_COLOR_4444(clear
[3], clear
[0],
325 case MESA_FORMAT_ARGB1555
:
326 clear_val
= PACK_COLOR_1555(clear
[3], clear
[0],
330 clear_val
= PACK_COLOR_8888(clear
[3], clear
[3],
334 _mesa_problem(ctx
, "Unexpected renderbuffer format: %d\n",
343 /* do space check before going any further */
344 aper_array
[0] = intel
->batch
->buf
;
345 aper_array
[1] = write_buffer
;
347 if (drm_intel_bufmgr_check_aperture_space(aper_array
,
348 ARRAY_SIZE(aper_array
)) != 0) {
349 intel_batchbuffer_flush(intel
->batch
);
355 OUT_BATCH((y1
<< 16) | x1
);
356 OUT_BATCH((y2
<< 16) | x2
);
357 OUT_RELOC_FENCED(write_buffer
,
358 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
360 OUT_BATCH(clear_val
);
363 if (intel
->always_flush_cache
)
364 intel_batchbuffer_emit_mi_flush(intel
->batch
);
366 if (buf
== BUFFER_DEPTH
|| buf
== BUFFER_STENCIL
)
367 mask
&= ~(BUFFER_BIT_DEPTH
| BUFFER_BIT_STENCIL
);
369 mask
&= ~bufBit
; /* turn off bit, for faster loop exit */
374 intelEmitImmediateColorExpandBlit(struct intel_context
*intel
,
376 GLubyte
*src_bits
, GLuint src_size
,
379 drm_intel_bo
*dst_buffer
,
382 GLshort x
, GLshort y
,
383 GLshort w
, GLshort h
,
386 int dwords
= ALIGN(src_size
, 8) / 4;
387 uint32_t opcode
, br13
, blit_cmd
;
389 if (dst_tiling
!= I915_TILING_NONE
) {
390 if (dst_offset
& 4095)
392 if (dst_tiling
== I915_TILING_Y
)
396 assert( logic_op
- GL_CLEAR
>= 0 );
397 assert( logic_op
- GL_CLEAR
< 0x10 );
398 assert(dst_pitch
> 0);
405 DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
407 dst_buffer
, dst_pitch
, dst_offset
, x
, y
, w
, h
, src_size
, dwords
);
409 intel_batchbuffer_require_space( intel
->batch
,
414 opcode
= XY_SETUP_BLT_CMD
;
416 opcode
|= XY_BLT_WRITE_ALPHA
| XY_BLT_WRITE_RGB
;
418 if (dst_tiling
!= I915_TILING_NONE
) {
419 opcode
|= XY_DST_TILED
;
424 br13
= dst_pitch
| (translate_raster_op(logic_op
) << 16) | (1 << 29);
425 br13
|= br13_for_cpp(cpp
);
427 blit_cmd
= XY_TEXT_IMMEDIATE_BLIT_CMD
| XY_TEXT_BYTE_PACKED
; /* packing? */
428 if (dst_tiling
!= I915_TILING_NONE
)
429 blit_cmd
|= XY_DST_TILED
;
431 BEGIN_BATCH_BLT(8 + 3);
434 OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
435 OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
436 OUT_RELOC_FENCED(dst_buffer
,
437 I915_GEM_DOMAIN_RENDER
, I915_GEM_DOMAIN_RENDER
,
439 OUT_BATCH(0); /* bg */
440 OUT_BATCH(fg_color
); /* fg */
441 OUT_BATCH(0); /* pattern base addr */
443 OUT_BATCH(blit_cmd
| ((3 - 2) + dwords
));
444 OUT_BATCH((y
<< 16) | x
);
445 OUT_BATCH(((y
+ h
) << 16) | (x
+ w
));
448 intel_batchbuffer_data(intel
->batch
,
452 intel_batchbuffer_emit_mi_flush(intel
->batch
);
457 /* We don't have a memmove-type blit like some other hardware, so we'll do a
458 * rectangular blit covering a large space, then emit 1-scanline blit at the
459 * end to cover the last if we need.
462 intel_emit_linear_blit(struct intel_context
*intel
,
463 drm_intel_bo
*dst_bo
,
464 unsigned int dst_offset
,
465 drm_intel_bo
*src_bo
,
466 unsigned int src_offset
,
469 GLuint pitch
, height
;
472 /* The pitch given to the GPU must be DWORD aligned, and
473 * we want width to match pitch. Max width is (1 << 15 - 1),
474 * rounding that down to the nearest DWORD is 1 << 15 - 4
476 pitch
= MIN2(size
, (1 << 15) - 4);
477 height
= size
/ pitch
;
478 ok
= intelEmitCopyBlit(intel
, 1,
479 pitch
, src_bo
, src_offset
, I915_TILING_NONE
,
480 pitch
, dst_bo
, dst_offset
, I915_TILING_NONE
,
483 pitch
, height
, /* w, h */
487 src_offset
+= pitch
* height
;
488 dst_offset
+= pitch
* height
;
489 size
-= pitch
* height
;
490 assert (size
< (1 << 15));
491 assert ((size
& 3) == 0); /* Pitch must be DWORD aligned */
493 ok
= intelEmitCopyBlit(intel
, 1,
494 size
, src_bo
, src_offset
, I915_TILING_NONE
,
495 size
, dst_bo
, dst_offset
, I915_TILING_NONE
,