intel: avoid unnecessary front buffer flushing/updating
[mesa.git] / src / mesa / drivers / dri / intel / intel_span.c
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 #include "main/glheader.h"
29 #include "main/macros.h"
30 #include "main/mtypes.h"
31 #include "main/colormac.h"
32
33 #include "intel_buffers.h"
34 #include "intel_fbo.h"
35 #include "intel_screen.h"
36 #include "intel_span.h"
37 #include "intel_regions.h"
38 #include "intel_tex.h"
39
40 #include "swrast/swrast.h"
41
42 static void
43 intel_set_span_functions(struct intel_context *intel,
44 struct gl_renderbuffer *rb);
45
46 #define SPAN_CACHE_SIZE 4096
47
48 static void
49 get_span_cache(struct intel_renderbuffer *irb, uint32_t offset)
50 {
51 if (irb->span_cache == NULL) {
52 irb->span_cache = _mesa_malloc(SPAN_CACHE_SIZE);
53 irb->span_cache_offset = -1;
54 }
55
56 if ((offset & ~(SPAN_CACHE_SIZE - 1)) != irb->span_cache_offset) {
57 irb->span_cache_offset = offset & ~(SPAN_CACHE_SIZE - 1);
58 dri_bo_get_subdata(irb->region->buffer, irb->span_cache_offset,
59 SPAN_CACHE_SIZE, irb->span_cache);
60 }
61 }
62
63 static void
64 clear_span_cache(struct intel_renderbuffer *irb)
65 {
66 irb->span_cache_offset = -1;
67 }
68
69 static uint32_t
70 pread_32(struct intel_renderbuffer *irb, uint32_t offset)
71 {
72 get_span_cache(irb, offset);
73
74 return *(uint32_t *)(irb->span_cache + (offset & (SPAN_CACHE_SIZE - 1)));
75 }
76
77 static uint32_t
78 pread_xrgb8888(struct intel_renderbuffer *irb, uint32_t offset)
79 {
80 get_span_cache(irb, offset);
81
82 return *(uint32_t *)(irb->span_cache + (offset & (SPAN_CACHE_SIZE - 1))) |
83 0xff000000;
84 }
85
86 static uint16_t
87 pread_16(struct intel_renderbuffer *irb, uint32_t offset)
88 {
89 get_span_cache(irb, offset);
90
91 return *(uint16_t *)(irb->span_cache + (offset & (SPAN_CACHE_SIZE - 1)));
92 }
93
94 static uint8_t
95 pread_8(struct intel_renderbuffer *irb, uint32_t offset)
96 {
97 get_span_cache(irb, offset);
98
99 return *(uint8_t *)(irb->span_cache + (offset & (SPAN_CACHE_SIZE - 1)));
100 }
101
102 static void
103 pwrite_32(struct intel_renderbuffer *irb, uint32_t offset, uint32_t val)
104 {
105 clear_span_cache(irb);
106
107 dri_bo_subdata(irb->region->buffer, offset, 4, &val);
108 }
109
110 static void
111 pwrite_xrgb8888(struct intel_renderbuffer *irb, uint32_t offset, uint32_t val)
112 {
113 clear_span_cache(irb);
114
115 dri_bo_subdata(irb->region->buffer, offset, 3, &val);
116 }
117
118 static void
119 pwrite_16(struct intel_renderbuffer *irb, uint32_t offset, uint16_t val)
120 {
121 clear_span_cache(irb);
122
123 dri_bo_subdata(irb->region->buffer, offset, 2, &val);
124 }
125
126 static void
127 pwrite_8(struct intel_renderbuffer *irb, uint32_t offset, uint8_t val)
128 {
129 clear_span_cache(irb);
130
131 dri_bo_subdata(irb->region->buffer, offset, 1, &val);
132 }
133
134 static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb,
135 int x, int y)
136 {
137 return (y * irb->region->pitch + x) * irb->region->cpp;
138 }
139
140 /*
141 * Deal with tiled surfaces
142 */
143
144 static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
145 int x, int y)
146 {
147 int tile_stride;
148 int xbyte;
149 int x_tile_off, y_tile_off;
150 int x_tile_number, y_tile_number;
151 int tile_off, tile_base;
152
153 x += irb->region->draw_x;
154 y += irb->region->draw_y;
155
156 tile_stride = (irb->region->pitch * irb->region->cpp) << 3;
157
158 xbyte = x * irb->region->cpp;
159
160 x_tile_off = xbyte & 0x1ff;
161 y_tile_off = y & 7;
162
163 x_tile_number = xbyte >> 9;
164 y_tile_number = y >> 3;
165
166 tile_off = (y_tile_off << 9) + x_tile_off;
167
168 switch (irb->region->bit_6_swizzle) {
169 case I915_BIT_6_SWIZZLE_NONE:
170 break;
171 case I915_BIT_6_SWIZZLE_9:
172 tile_off ^= ((tile_off >> 3) & 64);
173 break;
174 case I915_BIT_6_SWIZZLE_9_10:
175 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64);
176 break;
177 case I915_BIT_6_SWIZZLE_9_11:
178 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 5) & 64);
179 break;
180 case I915_BIT_6_SWIZZLE_9_10_11:
181 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64) ^
182 ((tile_off >> 5) & 64);
183 break;
184 default:
185 fprintf(stderr, "Unknown tile swizzling mode %d\n",
186 irb->region->bit_6_swizzle);
187 exit(1);
188 }
189
190 tile_base = (x_tile_number << 12) + y_tile_number * tile_stride;
191
192 #if 0
193 printf("(%d,%d) -> %d + %d = %d (pitch = %d, tstride = %d)\n",
194 x, y, tile_off, tile_base,
195 tile_off + tile_base,
196 irb->region->pitch, tile_stride);
197 #endif
198
199 return tile_base + tile_off;
200 }
201
202 static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
203 int x, int y)
204 {
205 int tile_stride;
206 int xbyte;
207 int x_tile_off, y_tile_off;
208 int x_tile_number, y_tile_number;
209 int tile_off, tile_base;
210
211 x += irb->region->draw_x;
212 y += irb->region->draw_y;
213
214 tile_stride = (irb->region->pitch * irb->region->cpp) << 5;
215
216 xbyte = x * irb->region->cpp;
217
218 x_tile_off = xbyte & 0x7f;
219 y_tile_off = y & 0x1f;
220
221 x_tile_number = xbyte >> 7;
222 y_tile_number = y >> 5;
223
224 tile_off = ((x_tile_off & ~0xf) << 5) + (y_tile_off << 4) +
225 (x_tile_off & 0xf);
226
227 switch (irb->region->bit_6_swizzle) {
228 case I915_BIT_6_SWIZZLE_NONE:
229 break;
230 case I915_BIT_6_SWIZZLE_9:
231 tile_off ^= ((tile_off >> 3) & 64);
232 break;
233 case I915_BIT_6_SWIZZLE_9_10:
234 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64);
235 break;
236 case I915_BIT_6_SWIZZLE_9_11:
237 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 5) & 64);
238 break;
239 case I915_BIT_6_SWIZZLE_9_10_11:
240 tile_off ^= ((tile_off >> 3) & 64) ^ ((tile_off >> 4) & 64) ^
241 ((tile_off >> 5) & 64);
242 break;
243 default:
244 fprintf(stderr, "Unknown tile swizzling mode %d\n",
245 irb->region->bit_6_swizzle);
246 exit(1);
247 }
248
249 tile_base = (x_tile_number << 12) + y_tile_number * tile_stride;
250
251 return tile_base + tile_off;
252 }
253
254 /*
255 break intelWriteRGBASpan_ARGB8888
256 */
257
258 #undef DBG
259 #define DBG 0
260
261 #define LOCAL_VARS \
262 struct intel_context *intel = intel_context(ctx); \
263 struct intel_renderbuffer *irb = intel_renderbuffer(rb); \
264 const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1; \
265 const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
266 unsigned int num_cliprects; \
267 struct drm_clip_rect *cliprects; \
268 int x_off, y_off; \
269 int pitch = irb->region->pitch * irb->region->cpp; \
270 void *buf = irb->region->buffer->virtual; \
271 GLuint p; \
272 (void) p; \
273 (void)buf; (void)pitch; /* unused for non-gttmap. */ \
274 intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
275
276 /* XXX FBO: this is identical to the macro in spantmp2.h except we get
277 * the cliprect info from the context, not the driDrawable.
278 * Move this into spantmp2.h someday.
279 */
280 #define HW_CLIPLOOP() \
281 do { \
282 int _nc = num_cliprects; \
283 while ( _nc-- ) { \
284 int minx = cliprects[_nc].x1 - x_off; \
285 int miny = cliprects[_nc].y1 - y_off; \
286 int maxx = cliprects[_nc].x2 - x_off; \
287 int maxy = cliprects[_nc].y2 - y_off;
288
289 #if 0
290 }}
291 #endif
292
293 #define Y_FLIP(_y) ((_y) * yScale + yBias)
294
295 /* XXX with GEM, these need to tell the kernel */
296 #define HW_LOCK()
297
298 #define HW_UNLOCK()
299
300 /* Convenience macros to avoid typing the swizzle argument over and over */
301 #define NO_TILE(_X, _Y) no_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
302 #define X_TILE(_X, _Y) x_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
303 #define Y_TILE(_X, _Y) y_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
304
305 /* r5g6b5 color span and pixel functions */
306 #define INTEL_PIXEL_FMT GL_RGB
307 #define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
308 #define INTEL_READ_VALUE(offset) pread_16(irb, offset)
309 #define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
310 #define INTEL_TAG(x) x##_RGB565
311 #include "intel_spantmp.h"
312
313 /* a4r4g4b4 color span and pixel functions */
314 #define INTEL_PIXEL_FMT GL_BGRA
315 #define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_4_4_4_4_REV
316 #define INTEL_READ_VALUE(offset) pread_16(irb, offset)
317 #define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
318 #define INTEL_TAG(x) x##_ARGB4444
319 #include "intel_spantmp.h"
320
321 /* a1r5g5b5 color span and pixel functions */
322 #define INTEL_PIXEL_FMT GL_BGRA
323 #define INTEL_PIXEL_TYPE GL_UNSIGNED_SHORT_1_5_5_5_REV
324 #define INTEL_READ_VALUE(offset) pread_16(irb, offset)
325 #define INTEL_WRITE_VALUE(offset, v) pwrite_16(irb, offset, v)
326 #define INTEL_TAG(x) x##_ARGB1555
327 #include "intel_spantmp.h"
328
329 /* a8r8g8b8 color span and pixel functions */
330 #define INTEL_PIXEL_FMT GL_BGRA
331 #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
332 #define INTEL_READ_VALUE(offset) pread_32(irb, offset)
333 #define INTEL_WRITE_VALUE(offset, v) pwrite_32(irb, offset, v)
334 #define INTEL_TAG(x) x##_ARGB8888
335 #include "intel_spantmp.h"
336
337 /* x8r8g8b8 color span and pixel functions */
338 #define INTEL_PIXEL_FMT GL_BGRA
339 #define INTEL_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
340 #define INTEL_READ_VALUE(offset) pread_xrgb8888(irb, offset)
341 #define INTEL_WRITE_VALUE(offset, v) pwrite_xrgb8888(irb, offset, v)
342 #define INTEL_TAG(x) x##_xRGB8888
343 #include "intel_spantmp.h"
344
345 #define LOCAL_DEPTH_VARS \
346 struct intel_context *intel = intel_context(ctx); \
347 struct intel_renderbuffer *irb = intel_renderbuffer(rb); \
348 const GLint yScale = ctx->DrawBuffer->Name ? 1 : -1; \
349 const GLint yBias = ctx->DrawBuffer->Name ? 0 : irb->Base.Height - 1;\
350 unsigned int num_cliprects; \
351 struct drm_clip_rect *cliprects; \
352 int x_off, y_off; \
353 int pitch = irb->region->pitch * irb->region->cpp; \
354 void *buf = irb->region->buffer->virtual; \
355 (void)buf; (void)pitch; /* unused for non-gttmap. */ \
356 intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
357
358
359 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
360
361 /* z16 depthbuffer functions. */
362 #define INTEL_VALUE_TYPE GLushort
363 #define INTEL_WRITE_DEPTH(offset, d) pwrite_16(irb, offset, d)
364 #define INTEL_READ_DEPTH(offset) pread_16(irb, offset)
365 #define INTEL_TAG(name) name##_z16
366 #include "intel_depthtmp.h"
367
368 /* z24x8 depthbuffer functions. */
369 #define INTEL_VALUE_TYPE GLuint
370 #define INTEL_WRITE_DEPTH(offset, d) pwrite_32(irb, offset, d)
371 #define INTEL_READ_DEPTH(offset) pread_32(irb, offset)
372 #define INTEL_TAG(name) name##_z24_x8
373 #include "intel_depthtmp.h"
374
375
376 /**
377 ** 8-bit stencil function (XXX FBO: This is obsolete)
378 **/
379 /* XXX */
380 #define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, NO_TILE(_x, _y) + 3, d)
381 #define READ_STENCIL(d, _x, _y) d = pread_8(irb, NO_TILE(_x, _y) + 3);
382 #define TAG(x) intel_gttmap_##x##_z24_s8
383 #include "stenciltmp.h"
384
385 /**
386 ** 8-bit stencil function (XXX FBO: This is obsolete)
387 **/
388 #define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, NO_TILE(_x, _y) + 3, d)
389 #define READ_STENCIL(d, _x, _y) d = pread_8(irb, NO_TILE(_x, _y) + 3);
390 #define TAG(x) intel##x##_z24_s8
391 #include "stenciltmp.h"
392
393 /**
394 ** 8-bit x-tile stencil function (XXX FBO: This is obsolete)
395 **/
396 #define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, X_TILE(_x, _y) + 3, d)
397 #define READ_STENCIL(d, _x, _y) d = pread_8(irb, X_TILE(_x, _y) + 3);
398 #define TAG(x) intel_XTile_##x##_z24_s8
399 #include "stenciltmp.h"
400
401 /**
402 ** 8-bit y-tile stencil function (XXX FBO: This is obsolete)
403 **/
404 #define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, Y_TILE(_x, _y) + 3, d)
405 #define READ_STENCIL(d, _x, _y) d = pread_8(irb, Y_TILE(_x, _y) + 3)
406 #define TAG(x) intel_YTile_##x##_z24_s8
407 #include "stenciltmp.h"
408
409 void
410 intel_renderbuffer_map(struct intel_context *intel, struct gl_renderbuffer *rb)
411 {
412 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
413
414 if (irb == NULL || irb->region == NULL)
415 return;
416
417 if (intel->intelScreen->kernel_exec_fencing)
418 drm_intel_gem_bo_map_gtt(irb->region->buffer);
419
420 intel_set_span_functions(intel, rb);
421 }
422
423 void
424 intel_renderbuffer_unmap(struct intel_context *intel,
425 struct gl_renderbuffer *rb)
426 {
427 struct intel_renderbuffer *irb = intel_renderbuffer(rb);
428
429 if (irb == NULL || irb->region == NULL)
430 return;
431
432 if (intel->intelScreen->kernel_exec_fencing)
433 drm_intel_gem_bo_unmap_gtt(irb->region->buffer);
434 else
435 clear_span_cache(irb);
436
437 rb->GetRow = NULL;
438 rb->PutRow = NULL;
439 }
440
441 /**
442 * Map or unmap all the renderbuffers which we may need during
443 * software rendering.
444 * XXX in the future, we could probably convey extra information to
445 * reduce the number of mappings needed. I.e. if doing a glReadPixels
446 * from the depth buffer, we really only need one mapping.
447 *
448 * XXX Rewrite this function someday.
449 * We can probably just loop over all the renderbuffer attachments,
450 * map/unmap all of them, and not worry about the _ColorDrawBuffers
451 * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields.
452 */
453 static void
454 intel_map_unmap_framebuffer(struct intel_context *intel,
455 struct gl_framebuffer *fb,
456 GLboolean map)
457 {
458 GLuint i;
459
460 /* color draw buffers */
461 for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
462 if (map)
463 intel_renderbuffer_map(intel, fb->_ColorDrawBuffers[i]);
464 else
465 intel_renderbuffer_unmap(intel, fb->_ColorDrawBuffers[i]);
466 }
467
468 /* color read buffer */
469 if (map)
470 intel_renderbuffer_map(intel, fb->_ColorReadBuffer);
471 else
472 intel_renderbuffer_unmap(intel, fb->_ColorReadBuffer);
473
474 /* check for render to textures */
475 for (i = 0; i < BUFFER_COUNT; i++) {
476 struct gl_renderbuffer_attachment *att =
477 fb->Attachment + i;
478 struct gl_texture_object *tex = att->Texture;
479 if (tex) {
480 /* render to texture */
481 ASSERT(att->Renderbuffer);
482 if (map)
483 intel_tex_map_images(intel, intel_texture_object(tex));
484 else
485 intel_tex_unmap_images(intel, intel_texture_object(tex));
486 }
487 }
488
489 /* depth buffer (Note wrapper!) */
490 if (fb->_DepthBuffer) {
491 if (map)
492 intel_renderbuffer_map(intel, fb->_DepthBuffer->Wrapped);
493 else
494 intel_renderbuffer_unmap(intel, fb->_DepthBuffer->Wrapped);
495 }
496
497 /* stencil buffer (Note wrapper!) */
498 if (fb->_StencilBuffer) {
499 if (map)
500 intel_renderbuffer_map(intel, fb->_StencilBuffer->Wrapped);
501 else
502 intel_renderbuffer_unmap(intel, fb->_StencilBuffer->Wrapped);
503 }
504
505 intel_check_front_buffer_rendering(intel);
506 }
507
508 /**
509 * Prepare for software rendering. Map current read/draw framebuffers'
510 * renderbuffes and all currently bound texture objects.
511 *
512 * Old note: Moved locking out to get reasonable span performance.
513 */
514 void
515 intelSpanRenderStart(GLcontext * ctx)
516 {
517 struct intel_context *intel = intel_context(ctx);
518 GLuint i;
519
520 intelFlush(&intel->ctx);
521 LOCK_HARDWARE(intel);
522
523 for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
524 if (ctx->Texture.Unit[i]._ReallyEnabled) {
525 struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
526 intel_tex_map_images(intel, intel_texture_object(texObj));
527 }
528 }
529
530 intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_TRUE);
531 if (ctx->ReadBuffer != ctx->DrawBuffer)
532 intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_TRUE);
533 }
534
535 /**
536 * Called when done software rendering. Unmap the buffers we mapped in
537 * the above function.
538 */
539 void
540 intelSpanRenderFinish(GLcontext * ctx)
541 {
542 struct intel_context *intel = intel_context(ctx);
543 GLuint i;
544
545 _swrast_flush(ctx);
546
547 for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
548 if (ctx->Texture.Unit[i]._ReallyEnabled) {
549 struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
550 intel_tex_unmap_images(intel, intel_texture_object(texObj));
551 }
552 }
553
554 intel_map_unmap_framebuffer(intel, ctx->DrawBuffer, GL_FALSE);
555 if (ctx->ReadBuffer != ctx->DrawBuffer)
556 intel_map_unmap_framebuffer(intel, ctx->ReadBuffer, GL_FALSE);
557
558 UNLOCK_HARDWARE(intel);
559 }
560
561
562 void
563 intelInitSpanFuncs(GLcontext * ctx)
564 {
565 struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
566 swdd->SpanRenderStart = intelSpanRenderStart;
567 swdd->SpanRenderFinish = intelSpanRenderFinish;
568 }
569
570 void
571 intel_map_vertex_shader_textures(GLcontext *ctx)
572 {
573 struct intel_context *intel = intel_context(ctx);
574 int i;
575
576 if (ctx->VertexProgram._Current == NULL)
577 return;
578
579 for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
580 if (ctx->Texture.Unit[i]._ReallyEnabled &&
581 ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) {
582 struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
583
584 intel_tex_map_images(intel, intel_texture_object(texObj));
585 }
586 }
587 }
588
589 void
590 intel_unmap_vertex_shader_textures(GLcontext *ctx)
591 {
592 struct intel_context *intel = intel_context(ctx);
593 int i;
594
595 if (ctx->VertexProgram._Current == NULL)
596 return;
597
598 for (i = 0; i < ctx->Const.MaxTextureImageUnits; i++) {
599 if (ctx->Texture.Unit[i]._ReallyEnabled &&
600 ctx->VertexProgram._Current->Base.TexturesUsed[i] != 0) {
601 struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current;
602
603 intel_tex_unmap_images(intel, intel_texture_object(texObj));
604 }
605 }
606 }
607
608 /**
609 * Plug in appropriate span read/write functions for the given renderbuffer.
610 * These are used for the software fallbacks.
611 */
612 static void
613 intel_set_span_functions(struct intel_context *intel,
614 struct gl_renderbuffer *rb)
615 {
616 struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb;
617 uint32_t tiling;
618
619 /* If in GEM mode, we need to do the tile address swizzling ourselves,
620 * instead of the fence registers handling it.
621 */
622 if (intel->ttm)
623 tiling = irb->region->tiling;
624 else
625 tiling = I915_TILING_NONE;
626
627 if (intel->intelScreen->kernel_exec_fencing) {
628 switch (irb->texformat) {
629 case MESA_FORMAT_RGB565:
630 intel_gttmap_InitPointers_RGB565(rb);
631 break;
632 case MESA_FORMAT_ARGB4444:
633 intel_gttmap_InitPointers_ARGB4444(rb);
634 break;
635 case MESA_FORMAT_ARGB1555:
636 intel_gttmap_InitPointers_ARGB1555(rb);
637 break;
638 case MESA_FORMAT_XRGB8888:
639 intel_gttmap_InitPointers_xRGB8888(rb);
640 break;
641 case MESA_FORMAT_ARGB8888:
642 if (rb->_BaseFormat == GL_RGB) {
643 /* XXX remove this code someday when we enable XRGB surfaces */
644 /* 8888 RGBx */
645 intel_gttmap_InitPointers_xRGB8888(rb);
646 } else {
647 intel_gttmap_InitPointers_ARGB8888(rb);
648 }
649 break;
650 case MESA_FORMAT_Z16:
651 intel_gttmap_InitDepthPointers_z16(rb);
652 break;
653 case MESA_FORMAT_X8_Z24:
654 intel_gttmap_InitDepthPointers_z24_x8(rb);
655 break;
656 case MESA_FORMAT_S8_Z24:
657 /* There are a few different ways SW asks us to access the S8Z24 data:
658 * Z24 depth-only depth reads
659 * S8Z24 depth reads
660 * S8Z24 stencil reads.
661 */
662 if (rb->Format == MESA_FORMAT_S8_Z24) {
663 intel_gttmap_InitDepthPointers_z24_x8(rb);
664 } else if (rb->Format == MESA_FORMAT_S8) {
665 intel_gttmap_InitStencilPointers_z24_s8(rb);
666 }
667 break;
668 default:
669 _mesa_problem(NULL,
670 "Unexpected MesaFormat %d in intelSetSpanFunctions",
671 irb->texformat);
672 break;
673 }
674 return;
675 }
676
677 switch (irb->texformat) {
678 case MESA_FORMAT_RGB565:
679 switch (tiling) {
680 case I915_TILING_NONE:
681 default:
682 intelInitPointers_RGB565(rb);
683 break;
684 case I915_TILING_X:
685 intel_XTile_InitPointers_RGB565(rb);
686 break;
687 case I915_TILING_Y:
688 intel_YTile_InitPointers_RGB565(rb);
689 break;
690 }
691 break;
692 case MESA_FORMAT_ARGB4444:
693 switch (tiling) {
694 case I915_TILING_NONE:
695 default:
696 intelInitPointers_ARGB4444(rb);
697 break;
698 case I915_TILING_X:
699 intel_XTile_InitPointers_ARGB4444(rb);
700 break;
701 case I915_TILING_Y:
702 intel_YTile_InitPointers_ARGB4444(rb);
703 break;
704 }
705 break;
706 case MESA_FORMAT_ARGB1555:
707 switch (tiling) {
708 case I915_TILING_NONE:
709 default:
710 intelInitPointers_ARGB1555(rb);
711 break;
712 case I915_TILING_X:
713 intel_XTile_InitPointers_ARGB1555(rb);
714 break;
715 case I915_TILING_Y:
716 intel_YTile_InitPointers_ARGB1555(rb);
717 break;
718 }
719 break;
720 case MESA_FORMAT_XRGB8888:
721 switch (tiling) {
722 case I915_TILING_NONE:
723 default:
724 intelInitPointers_xRGB8888(rb);
725 break;
726 case I915_TILING_X:
727 intel_XTile_InitPointers_xRGB8888(rb);
728 break;
729 case I915_TILING_Y:
730 intel_YTile_InitPointers_xRGB8888(rb);
731 break;
732 }
733 break;
734 case MESA_FORMAT_ARGB8888:
735 if (rb->_BaseFormat == GL_RGB) {
736 /* XXX remove this code someday when we enable XRGB surfaces */
737 /* 8888 RGBx */
738 switch (tiling) {
739 case I915_TILING_NONE:
740 default:
741 intelInitPointers_xRGB8888(rb);
742 break;
743 case I915_TILING_X:
744 intel_XTile_InitPointers_xRGB8888(rb);
745 break;
746 case I915_TILING_Y:
747 intel_YTile_InitPointers_xRGB8888(rb);
748 break;
749 }
750 } else {
751 /* 8888 RGBA */
752 switch (tiling) {
753 case I915_TILING_NONE:
754 default:
755 intelInitPointers_ARGB8888(rb);
756 break;
757 case I915_TILING_X:
758 intel_XTile_InitPointers_ARGB8888(rb);
759 break;
760 case I915_TILING_Y:
761 intel_YTile_InitPointers_ARGB8888(rb);
762 break;
763 }
764 }
765 break;
766 case MESA_FORMAT_Z16:
767 switch (tiling) {
768 case I915_TILING_NONE:
769 default:
770 intelInitDepthPointers_z16(rb);
771 break;
772 case I915_TILING_X:
773 intel_XTile_InitDepthPointers_z16(rb);
774 break;
775 case I915_TILING_Y:
776 intel_YTile_InitDepthPointers_z16(rb);
777 break;
778 }
779 break;
780 case MESA_FORMAT_X8_Z24:
781 case MESA_FORMAT_S8_Z24:
782 /* There are a few different ways SW asks us to access the S8Z24 data:
783 * Z24 depth-only depth reads
784 * S8Z24 depth reads
785 * S8Z24 stencil reads.
786 */
787 if (rb->Format == MESA_FORMAT_S8_Z24) {
788 switch (tiling) {
789 case I915_TILING_NONE:
790 default:
791 intelInitDepthPointers_z24_x8(rb);
792 break;
793 case I915_TILING_X:
794 intel_XTile_InitDepthPointers_z24_x8(rb);
795 break;
796 case I915_TILING_Y:
797 intel_YTile_InitDepthPointers_z24_x8(rb);
798 break;
799 }
800 } else if (rb->Format == MESA_FORMAT_S8) {
801 switch (tiling) {
802 case I915_TILING_NONE:
803 default:
804 intelInitStencilPointers_z24_s8(rb);
805 break;
806 case I915_TILING_X:
807 intel_XTile_InitStencilPointers_z24_s8(rb);
808 break;
809 case I915_TILING_Y:
810 intel_YTile_InitStencilPointers_z24_s8(rb);
811 break;
812 }
813 } else {
814 _mesa_problem(NULL,
815 "Unexpected ActualFormat in intelSetSpanFunctions");
816 }
817 break;
818 default:
819 _mesa_problem(NULL,
820 "Unexpected MesaFormat in intelSetSpanFunctions");
821 break;
822 }
823 }