intel_blit.c revision 0cade4de4f74f6b0e86fb6622e2fc370c73fd840
1/**************************************************************************
2 *
3 * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
4 * All Rights Reserved.
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a
7 * copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sub license, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice (including the
15 * next paragraph) shall be included in all copies or substantial portions
16 * of the Software.
17 *
18 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
19 * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
20 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
21 * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
22 * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
23 * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
24 * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
25 *
26 **************************************************************************/
27
28
29#include <stdio.h>
30#include <errno.h>
31
32#include "main/mtypes.h"
33#include "main/context.h"
34#include "main/enums.h"
35
36#include "intel_blit.h"
37#include "intel_buffers.h"
38#include "intel_context.h"
39#include "intel_fbo.h"
40#include "intel_reg.h"
41#include "intel_regions.h"
42#include "intel_batchbuffer.h"
43
44#define FILE_DEBUG_FLAG DEBUG_BLIT
45
46/**
47 * Copy the back color buffer to the front color buffer.
48 * Used for SwapBuffers().
49 */
50void
51intelCopyBuffer(const __DRIdrawablePrivate * dPriv,
52                const drm_clip_rect_t * rect)
53{
54
55   struct intel_context *intel;
56   const intelScreenPrivate *intelScreen;
57
58   DBG("%s\n", __FUNCTION__);
59
60   assert(dPriv);
61
62   intel = intelScreenContext(dPriv->driScreenPriv->private);
63   if (!intel)
64      return;
65
66   intelScreen = intel->intelScreen;
67
68   /* The LOCK_HARDWARE is required for the cliprects.  Buffer offsets
69    * should work regardless.
70    */
71   LOCK_HARDWARE(intel);
72
73   if (dPriv && dPriv->numClipRects) {
74      struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
75      struct intel_region *src, *dst;
76      int nbox = dPriv->numClipRects;
77      drm_clip_rect_t *pbox = dPriv->pClipRects;
78      int cpp;
79      int src_pitch, dst_pitch;
80      unsigned short src_x, src_y;
81      int BR13, CMD;
82      int i;
83      dri_bo *aper_array[3];
84
85      src = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT);
86      dst = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT);
87
88      src_pitch = src->pitch * src->cpp;
89      dst_pitch = dst->pitch * dst->cpp;
90
91      cpp = src->cpp;
92
93      ASSERT(intel_fb);
94      ASSERT(intel_fb->Base.Name == 0);    /* Not a user-created FBO */
95      ASSERT(src);
96      ASSERT(dst);
97      ASSERT(src->cpp == dst->cpp);
98
99      if (cpp == 2) {
100	 BR13 = (0xCC << 16) | (1 << 24);
101	 CMD = XY_SRC_COPY_BLT_CMD;
102      }
103      else {
104	 BR13 = (0xCC << 16) | (1 << 24) | (1 << 25);
105	 CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
106      }
107
108#ifndef I915
109      if (src->tiling != I915_TILING_NONE) {
110	 CMD |= XY_SRC_TILED;
111	 src_pitch /= 4;
112      }
113      if (dst->tiling != I915_TILING_NONE) {
114	 CMD |= XY_DST_TILED;
115	 dst_pitch /= 4;
116      }
117#endif
118      /* do space/cliprects check before going any further */
119      intel_batchbuffer_require_space(intel->batch, 8 * 4,
120				      REFERENCES_CLIPRECTS);
121   again:
122      aper_array[0] = intel->batch->buf;
123      aper_array[1] = dst->buffer;
124      aper_array[2] = src->buffer;
125
126      if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
127	intel_batchbuffer_flush(intel->batch);
128	goto again;
129      }
130
131      for (i = 0; i < nbox; i++, pbox++) {
132	 drm_clip_rect_t box = *pbox;
133
134	 if (rect) {
135	    if (!intel_intersect_cliprects(&box, &box, rect))
136	       continue;
137	 }
138
139	 if (box.x1 >= box.x2 ||
140	     box.y1 >= box.y2)
141	    continue;
142
143	 assert(box.x1 < box.x2);
144	 assert(box.y1 < box.y2);
145	 src_x = box.x1 - dPriv->x + dPriv->backX;
146	 src_y = box.y1 - dPriv->y + dPriv->backY;
147
148	 BEGIN_BATCH(8, REFERENCES_CLIPRECTS);
149	 OUT_BATCH(CMD);
150	 OUT_BATCH(BR13 | dst_pitch);
151	 OUT_BATCH((box.y1 << 16) | box.x1);
152	 OUT_BATCH((box.y2 << 16) | box.x2);
153
154	 OUT_RELOC(dst->buffer,
155		   I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
156		   0);
157	 OUT_BATCH((src_y << 16) | src_x);
158	 OUT_BATCH(src_pitch);
159	 OUT_RELOC(src->buffer,
160		   I915_GEM_DOMAIN_RENDER, 0,
161		   0);
162	 ADVANCE_BATCH();
163      }
164
165      /* Flush the rendering and the batch so that the results all land on the
166       * screen in a timely fashion.
167       */
168      intel_batchbuffer_emit_mi_flush(intel->batch);
169      intel_batchbuffer_flush(intel->batch);
170   }
171
172   UNLOCK_HARDWARE(intel);
173}
174
175
176
177
178void
179intelEmitFillBlit(struct intel_context *intel,
180		  GLuint cpp,
181		  GLshort dst_pitch,
182		  dri_bo *dst_buffer,
183		  GLuint dst_offset,
184		  uint32_t dst_tiling,
185		  GLshort x, GLshort y,
186		  GLshort w, GLshort h,
187		  GLuint color)
188{
189   GLuint BR13, CMD;
190   BATCH_LOCALS;
191
192   dst_pitch *= cpp;
193
194   switch (cpp) {
195   case 1:
196   case 2:
197   case 3:
198      BR13 = (0xF0 << 16) | (1 << 24);
199      CMD = XY_COLOR_BLT_CMD;
200      break;
201   case 4:
202      BR13 = (0xF0 << 16) | (1 << 24) | (1 << 25);
203      CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
204      break;
205   default:
206      return;
207   }
208#ifndef I915
209   if (dst_tiling != I915_TILING_NONE) {
210      CMD |= XY_DST_TILED;
211      dst_pitch /= 4;
212   }
213#endif
214
215   DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
216       __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h);
217
218   assert(w > 0);
219   assert(h > 0);
220
221   BEGIN_BATCH(6, NO_LOOP_CLIPRECTS);
222   OUT_BATCH(CMD);
223   OUT_BATCH(BR13 | dst_pitch);
224   OUT_BATCH((y << 16) | x);
225   OUT_BATCH(((y + h) << 16) | (x + w));
226   OUT_RELOC(dst_buffer,
227	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
228	     dst_offset);
229   OUT_BATCH(color);
230   ADVANCE_BATCH();
231}
232
233static GLuint translate_raster_op(GLenum logicop)
234{
235   switch(logicop) {
236   case GL_CLEAR: return 0x00;
237   case GL_AND: return 0x88;
238   case GL_AND_REVERSE: return 0x44;
239   case GL_COPY: return 0xCC;
240   case GL_AND_INVERTED: return 0x22;
241   case GL_NOOP: return 0xAA;
242   case GL_XOR: return 0x66;
243   case GL_OR: return 0xEE;
244   case GL_NOR: return 0x11;
245   case GL_EQUIV: return 0x99;
246   case GL_INVERT: return 0x55;
247   case GL_OR_REVERSE: return 0xDD;
248   case GL_COPY_INVERTED: return 0x33;
249   case GL_OR_INVERTED: return 0xBB;
250   case GL_NAND: return 0x77;
251   case GL_SET: return 0xFF;
252   default: return 0;
253   }
254}
255
256
257/* Copy BitBlt
258 */
259void
260intelEmitCopyBlit(struct intel_context *intel,
261		  GLuint cpp,
262		  GLshort src_pitch,
263		  dri_bo *src_buffer,
264		  GLuint src_offset,
265		  uint32_t src_tiling,
266		  GLshort dst_pitch,
267		  dri_bo *dst_buffer,
268		  GLuint dst_offset,
269		  uint32_t dst_tiling,
270		  GLshort src_x, GLshort src_y,
271		  GLshort dst_x, GLshort dst_y,
272		  GLshort w, GLshort h,
273		  GLenum logic_op)
274{
275   GLuint CMD, BR13, pass = 0;
276   int dst_y2 = dst_y + h;
277   int dst_x2 = dst_x + w;
278   dri_bo *aper_array[3];
279   BATCH_LOCALS;
280
281   /* do space/cliprects check before going any further */
282   do {
283       aper_array[0] = intel->batch->buf;
284       aper_array[1] = dst_buffer;
285       aper_array[2] = src_buffer;
286
287       if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
288           intel_batchbuffer_flush(intel->batch);
289           pass++;
290       } else
291           break;
292   } while (pass < 2);
293
294   if (pass >= 2) {
295       GLboolean locked = GL_FALSE;
296       if (!intel->locked) {
297           LOCK_HARDWARE(intel);
298           locked = GL_TRUE;
299       }
300
301       dri_bo_map(dst_buffer, GL_TRUE);
302       dri_bo_map(src_buffer, GL_FALSE);
303       _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
304                       cpp,
305                       dst_pitch,
306                       dst_x, dst_y,
307                       w, h,
308                       (GLubyte *)src_buffer->virtual + src_offset,
309                       src_pitch,
310                       src_x, src_y);
311
312       dri_bo_unmap(src_buffer);
313       dri_bo_unmap(dst_buffer);
314
315       if (locked)
316           UNLOCK_HARDWARE(intel);
317
318       return;
319   }
320
321   intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
322   DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
323       __FUNCTION__,
324       src_buffer, src_pitch, src_offset, src_x, src_y,
325       dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h);
326
327   src_pitch *= cpp;
328   dst_pitch *= cpp;
329
330   BR13 = translate_raster_op(logic_op) << 16;
331
332   switch (cpp) {
333   case 1:
334   case 2:
335   case 3:
336      BR13 |= (1 << 24);
337      CMD = XY_SRC_COPY_BLT_CMD;
338      break;
339   case 4:
340      BR13 |= (1 << 24) | (1 << 25);
341      CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
342      break;
343   default:
344      return;
345   }
346
347#ifndef I915
348   if (dst_tiling != I915_TILING_NONE) {
349      CMD |= XY_DST_TILED;
350      dst_pitch /= 4;
351   }
352   if (src_tiling != I915_TILING_NONE) {
353      CMD |= XY_SRC_TILED;
354      src_pitch /= 4;
355   }
356#endif
357
358   if (dst_y2 <= dst_y || dst_x2 <= dst_x) {
359      return;
360   }
361
362   /* Initial y values don't seem to work with negative pitches.  If
363    * we adjust the offsets manually (below), it seems to work fine.
364    *
365    * On the other hand, if we always adjust, the hardware doesn't
366    * know which blit directions to use, so overlapping copypixels get
367    * the wrong result.
368    */
369   if (dst_pitch > 0 && src_pitch > 0) {
370      assert(dst_x < dst_x2);
371      assert(dst_y < dst_y2);
372
373      BEGIN_BATCH(8, NO_LOOP_CLIPRECTS);
374      OUT_BATCH(CMD);
375      OUT_BATCH(BR13 | dst_pitch);
376      OUT_BATCH((dst_y << 16) | dst_x);
377      OUT_BATCH((dst_y2 << 16) | dst_x2);
378      OUT_RELOC(dst_buffer,
379		I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
380		dst_offset);
381      OUT_BATCH((src_y << 16) | src_x);
382      OUT_BATCH(src_pitch);
383      OUT_RELOC(src_buffer,
384		I915_GEM_DOMAIN_RENDER, 0,
385		src_offset);
386      ADVANCE_BATCH();
387   }
388   else {
389      assert(dst_x < dst_x2);
390      assert(h > 0);
391
392      BEGIN_BATCH(8, NO_LOOP_CLIPRECTS);
393      OUT_BATCH(CMD);
394      OUT_BATCH(BR13 | ((uint16_t)dst_pitch));
395      OUT_BATCH((0 << 16) | dst_x);
396      OUT_BATCH((h << 16) | dst_x2);
397      OUT_RELOC(dst_buffer,
398		I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
399		dst_offset + dst_y * dst_pitch);
400      OUT_BATCH((0 << 16) | src_x);
401      OUT_BATCH(src_pitch);
402      OUT_RELOC(src_buffer,
403		I915_GEM_DOMAIN_RENDER, 0,
404		src_offset + src_y * src_pitch);
405      ADVANCE_BATCH();
406   }
407   intel_batchbuffer_emit_mi_flush(intel->batch);
408}
409
410
411/**
412 * Use blitting to clear the renderbuffers named by 'flags'.
413 * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferIndexes field
414 * since that might include software renderbuffers or renderbuffers
415 * which we're clearing with triangles.
416 * \param mask  bitmask of BUFFER_BIT_* values indicating buffers to clear
417 */
418void
419intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
420{
421   struct intel_context *intel = intel_context(ctx);
422   struct gl_framebuffer *fb = ctx->DrawBuffer;
423   GLuint clear_depth;
424   GLbitfield skipBuffers = 0;
425   unsigned int num_cliprects;
426   struct drm_clip_rect *cliprects;
427   int x_off, y_off;
428   BATCH_LOCALS;
429
430   /*
431    * Compute values for clearing the buffers.
432    */
433   clear_depth = 0;
434   if (mask & BUFFER_BIT_DEPTH) {
435      clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear);
436   }
437   if (mask & BUFFER_BIT_STENCIL) {
438      clear_depth |= (ctx->Stencil.Clear & 0xff) << 24;
439   }
440
441   /* If clearing both depth and stencil, skip BUFFER_BIT_STENCIL in
442    * the loop below.
443    */
444   if ((mask & BUFFER_BIT_DEPTH) && (mask & BUFFER_BIT_STENCIL)) {
445      skipBuffers = BUFFER_BIT_STENCIL;
446   }
447
448   /* XXX Move this flush/lock into the following conditional? */
449   intelFlush(&intel->ctx);
450   LOCK_HARDWARE(intel);
451
452   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
453   if (num_cliprects) {
454      GLint cx, cy, cw, ch;
455      drm_clip_rect_t clear;
456      int i;
457
458      /* Get clear bounds after locking */
459      cx = fb->_Xmin;
460      cy = fb->_Ymin;
461      cw = fb->_Xmax - cx;
462      ch = fb->_Ymax - cy;
463
464      if (fb->Name == 0) {
465         /* clearing a window */
466
467         /* flip top to bottom */
468         clear.x1 = cx + x_off;
469         clear.y1 = intel->driDrawable->y + intel->driDrawable->h - cy - ch;
470         clear.x2 = clear.x1 + cw;
471         clear.y2 = clear.y1 + ch;
472      }
473      else {
474         /* clearing FBO */
475         assert(num_cliprects == 1);
476         assert(cliprects == &intel->fboRect);
477         clear.x1 = cx;
478         clear.y1 = cy;
479         clear.x2 = clear.x1 + cw;
480         clear.y2 = clear.y1 + ch;
481         /* no change to mask */
482      }
483
484      for (i = 0; i < num_cliprects; i++) {
485         const drm_clip_rect_t *box = &cliprects[i];
486         drm_clip_rect_t b;
487         GLuint buf;
488         GLuint clearMask = mask;      /* use copy, since we modify it below */
489         GLboolean all = (cw == fb->Width && ch == fb->Height);
490
491         if (!all) {
492            intel_intersect_cliprects(&b, &clear, box);
493         }
494         else {
495            b = *box;
496         }
497
498         if (b.x1 >= b.x2 || b.y1 >= b.y2)
499            continue;
500
501         if (0)
502            _mesa_printf("clear %d,%d..%d,%d, mask %x\n",
503                         b.x1, b.y1, b.x2, b.y2, mask);
504
505         /* Loop over all renderbuffers */
506         for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) {
507            const GLbitfield bufBit = 1 << buf;
508            if ((clearMask & bufBit) && !(bufBit & skipBuffers)) {
509               /* OK, clear this renderbuffer */
510               struct intel_region *irb_region =
511		  intel_get_rb_region(fb, buf);
512               dri_bo *write_buffer =
513                  intel_region_buffer(intel, irb_region,
514                                      all ? INTEL_WRITE_FULL :
515                                      INTEL_WRITE_PART);
516
517               GLuint clearVal;
518               GLint pitch, cpp;
519               GLuint BR13, CMD;
520
521               ASSERT(irb_region);
522
523               pitch = irb_region->pitch;
524               cpp = irb_region->cpp;
525
526               DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
527                   __FUNCTION__,
528                   irb_region->buffer, (pitch * cpp),
529                   irb_region->draw_offset,
530                   b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1);
531
532	       BR13 = 0xf0 << 16;
533	       CMD = XY_COLOR_BLT_CMD;
534
535               /* Setup the blit command */
536               if (cpp == 4) {
537                  BR13 |= (1 << 24) | (1 << 25);
538                  if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) {
539                     if (clearMask & BUFFER_BIT_DEPTH)
540                        CMD |= XY_BLT_WRITE_RGB;
541                     if (clearMask & BUFFER_BIT_STENCIL)
542                        CMD |= XY_BLT_WRITE_ALPHA;
543                  }
544                  else {
545                     /* clearing RGBA */
546                     CMD |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
547                  }
548               }
549               else {
550                  ASSERT(cpp == 2 || cpp == 0);
551                  BR13 |= (1 << 24);
552               }
553
554#ifndef I915
555	       if (irb_region->tiling != I915_TILING_NONE) {
556		  CMD |= XY_DST_TILED;
557		  pitch /= 4;
558	       }
559#endif
560	       BR13 |= (pitch * cpp);
561
562               if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) {
563                  clearVal = clear_depth;
564               }
565               else {
566                  clearVal = (cpp == 4)
567                     ? intel->ClearColor8888 : intel->ClearColor565;
568               }
569               /*
570                  _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n",
571                  buf, irb->Base.Name);
572                */
573	       intel_wait_flips(intel);
574
575               assert(b.x1 < b.x2);
576               assert(b.y1 < b.y2);
577
578               BEGIN_BATCH(6, REFERENCES_CLIPRECTS);
579               OUT_BATCH(CMD);
580               OUT_BATCH(BR13);
581               OUT_BATCH((b.y1 << 16) | b.x1);
582               OUT_BATCH((b.y2 << 16) | b.x2);
583               OUT_RELOC(write_buffer,
584			 I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
585                         irb_region->draw_offset);
586               OUT_BATCH(clearVal);
587               ADVANCE_BATCH();
588               clearMask &= ~bufBit;    /* turn off bit, for faster loop exit */
589            }
590         }
591      }
592      intel_batchbuffer_emit_mi_flush(intel->batch);
593   }
594
595   UNLOCK_HARDWARE(intel);
596}
597
598void
599intelEmitImmediateColorExpandBlit(struct intel_context *intel,
600				  GLuint cpp,
601				  GLubyte *src_bits, GLuint src_size,
602				  GLuint fg_color,
603				  GLshort dst_pitch,
604				  dri_bo *dst_buffer,
605				  GLuint dst_offset,
606				  uint32_t dst_tiling,
607				  GLshort x, GLshort y,
608				  GLshort w, GLshort h,
609				  GLenum logic_op)
610{
611   int dwords = ALIGN(src_size, 8) / 4;
612   uint32_t opcode, br13, blit_cmd;
613
614   assert( logic_op - GL_CLEAR >= 0 );
615   assert( logic_op - GL_CLEAR < 0x10 );
616
617   if (w < 0 || h < 0)
618      return;
619
620   dst_pitch *= cpp;
621
622   DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d, %d bytes %d dwords\n",
623       __FUNCTION__,
624       dst_buffer, dst_pitch, dst_offset, x, y, w, h, src_size, dwords);
625
626   intel_batchbuffer_require_space( intel->batch,
627				    (8 * 4) +
628				    (3 * 4) +
629				    dwords,
630				    REFERENCES_CLIPRECTS );
631
632   opcode = XY_SETUP_BLT_CMD;
633   if (cpp == 4)
634      opcode |= XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB;
635#ifndef I915
636   if (dst_tiling != I915_TILING_NONE) {
637      opcode |= XY_DST_TILED;
638      dst_pitch /= 4;
639   }
640#endif
641
642   br13 = dst_pitch | (translate_raster_op(logic_op) << 16) | (1 << 29);
643   if (cpp == 2)
644      br13 |= BR13_565;
645   else
646      br13 |= BR13_8888;
647
648   blit_cmd = XY_TEXT_IMMEDIATE_BLIT_CMD | XY_TEXT_BYTE_PACKED; /* packing? */
649   if (dst_tiling != I915_TILING_NONE)
650      blit_cmd |= XY_DST_TILED;
651
652   BEGIN_BATCH(8 + 3, REFERENCES_CLIPRECTS);
653   OUT_BATCH(opcode);
654   OUT_BATCH(br13);
655   OUT_BATCH((0 << 16) | 0); /* clip x1, y1 */
656   OUT_BATCH((100 << 16) | 100); /* clip x2, y2 */
657   OUT_RELOC(dst_buffer,
658	     I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
659	     dst_offset);
660   OUT_BATCH(0); /* bg */
661   OUT_BATCH(fg_color); /* fg */
662   OUT_BATCH(0); /* pattern base addr */
663
664   OUT_BATCH(blit_cmd | ((3 - 2) + dwords));
665   OUT_BATCH((y << 16) | x);
666   OUT_BATCH(((y + h) << 16) | (x + w));
667   ADVANCE_BATCH();
668
669   intel_batchbuffer_data( intel->batch,
670			   src_bits,
671			   dwords * 4,
672			   REFERENCES_CLIPRECTS );
673
674   intel_batchbuffer_emit_mi_flush(intel->batch);
675}
676