summaryrefslogtreecommitdiff
path: root/src/mesa/drivers/dri/r300/r300_texmem.c
blob: 78c077fa40f58f555aa184e42dd2350f6da93d4c (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
/* $XFree86: xc/lib/GL/mesa/src/drv/r300/r300_texmem.c,v 1.5 2002/12/17 00:32:56 dawes Exp $ */
/**************************************************************************

Copyright (C) Tungsten Graphics 2002.  All Rights Reserved.
The Weather Channel, Inc. funded Tungsten Graphics to develop the
initial release of the Radeon 8500 driver under the XFree86
license. This notice must be preserved.

Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
"Software"), to deal in the Software without restriction, including
without limitation on the rights to use, copy, modify, merge, publish,
distribute, sub license, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:

The above copyright notice and this permission notice (including the
next paragraph) shall be included in all copies or substantial
portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
NON-INFRINGEMENT. IN NO EVENT SHALL ATI, VA LINUX SYSTEMS AND/OR THEIR
SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

**************************************************************************/

/*
 * Authors:
 *   Kevin E. Martin <martin@valinux.com>
 *   Gareth Hughes <gareth@valinux.com>
 *
 */

#include <errno.h>

#include "glheader.h"
#include "imports.h"
#include "context.h"
#include "colormac.h"
#include "macros.h"
#include "simple_list.h"
#include "radeon_reg.h"		/* gets definition for usleep */
#include "r300_context.h"
#include "r300_state.h"
#include "radeon_ioctl.h"
/*
#include "r300_swtcl.h"
*/
#include "r300_tex.h"

#include <unistd.h>		/* for usleep() */

/**
 * Destroy any device-dependent state associated with the texture.  This may
 * include NULLing out hardware state that points to the texture.
 */
void r300DestroyTexObj(r300ContextPtr rmesa, r300TexObjPtr t)
{
	if (RADEON_DEBUG & DEBUG_TEXTURE) {
		fprintf(stderr, "%s( %p, %p )\n", __FUNCTION__,
			(void *)t, (void *)t->base.tObj);
	}

	if (rmesa != NULL) {
		unsigned i;

		for (i = 0; i < rmesa->radeon.glCtx->Const.MaxTextureUnits; i++) {
			if (t == rmesa->state.texture.unit[i].texobj) {
				rmesa->state.texture.unit[i].texobj = NULL;
				/* This code below is meant to shorten state
				   pushed to the hardware by not programming 
				   unneeded units.
				   
				   This does not appear to be worthwhile on R300 */
				#if 0
				remove_from_list(&rmesa->hw.tex[i]);
				make_empty_list(&rmesa->hw.tex[i]);
				remove_from_list(&rmesa->hw.cube[i]);
				make_empty_list(&rmesa->hw.cube[i]);
				#endif
			}
		}
	}
}

/* ------------------------------------------------------------
 * Texture image conversions
 */

static void r300UploadGARTClientSubImage(r300ContextPtr rmesa,
					 r300TexObjPtr t,
					 struct gl_texture_image *texImage,
					 GLint hwlevel,
					 GLint x, GLint y,
					 GLint width, GLint height)
{
	const struct gl_texture_format *texFormat = texImage->TexFormat;
	GLuint srcPitch, dstPitch;
	int blit_format;
	int srcOffset;

	/*
	 * XXX it appears that we always upload the full image, not a subimage.
	 * I.e. x==0, y==0, width=texWidth, height=texWidth.  If this is ever
	 * changed, the src pitch will have to change.
	 */
	switch (texFormat->TexelBytes) {
	case 1:
		blit_format = R200_CP_COLOR_FORMAT_CI8;
		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
		break;
	case 2:
		blit_format = R200_CP_COLOR_FORMAT_RGB565;
		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
		break;
	case 4:
		blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
		srcPitch = t->image[0][0].width * texFormat->TexelBytes;
		dstPitch = t->image[0][0].width * texFormat->TexelBytes;
		break;
	default:
		return;
	}

	t->image[0][hwlevel].data = texImage->Data;
	srcOffset = r300GartOffsetFromVirtual(rmesa, texImage->Data);

	assert(srcOffset != ~0);

	/* Don't currently need to cope with small pitches?
	 */
	width = texImage->Width;
	height = texImage->Height;

	r300EmitWait(rmesa, RADEON_WAIT_3D);

	r300EmitBlit(rmesa, blit_format,
		     srcPitch,
		     srcOffset,
		     dstPitch,
		     t->bufAddr,
		     x,
		     y,
		     t->image[0][hwlevel].x + x,
		     t->image[0][hwlevel].y + y, width, height);

	r300EmitWait(rmesa, RADEON_WAIT_2D);
}

static void r300UploadRectSubImage(r300ContextPtr rmesa,
				   r300TexObjPtr t,
				   struct gl_texture_image *texImage,
				   GLint x, GLint y, GLint width, GLint height)
{
	const struct gl_texture_format *texFormat = texImage->TexFormat;
	int blit_format, dstPitch, done;

	switch (texFormat->TexelBytes) {
	case 1:
		blit_format = R200_CP_COLOR_FORMAT_CI8;
		break;
	case 2:
		blit_format = R200_CP_COLOR_FORMAT_RGB565;
		break;
	case 4:
		blit_format = R200_CP_COLOR_FORMAT_ARGB8888;
		break;
	default:
		return;
	}

	t->image[0][0].data = texImage->Data;

	/* Currently don't need to cope with small pitches.
	 */
	width = texImage->Width;
	height = texImage->Height;
	dstPitch = t->pitch + 32;

	if (rmesa->prefer_gart_client_texturing && texImage->IsClientData) {
		/* In this case, could also use GART texturing.  This is
		 * currently disabled, but has been tested & works.
		 */
		t->offset =
		    r300GartOffsetFromVirtual(rmesa, texImage->Data);
		t->pitch =
		    texImage->RowStride * texFormat->TexelBytes - 32;

		if (RADEON_DEBUG & DEBUG_TEXTURE)
			fprintf(stderr,
				"Using GART texturing for rectangular client texture\n");

		/* Release FB memory allocated for this image:
		 */
		/* FIXME This may not be correct as driSwapOutTextureObject sets
		 * FIXME dirty_images.  It may be fine, though.
		 */
		if (t->base.memBlock) {
			driSwapOutTextureObject((driTextureObject *) t);
		}
	} else if (texImage->IsClientData) {
		/* Data already in GART memory, with usable pitch.
		 */
		GLuint srcPitch;
		srcPitch = texImage->RowStride * texFormat->TexelBytes;
		r300EmitBlit(rmesa,
			     blit_format,
			     srcPitch,
			     r300GartOffsetFromVirtual(rmesa, texImage->Data),
			     dstPitch, t->bufAddr, 0, 0, 0, 0, width, height);
	} else {
		/* Data not in GART memory, or bad pitch.
		 */
		for (done = 0; done < height;) {
			struct r300_dma_region region;
			int lines =
			    MIN2(height - done, RADEON_BUFFER_SIZE / dstPitch);
			int src_pitch;
			char *tex;

			src_pitch = texImage->RowStride * texFormat->TexelBytes;

			tex = (char *)texImage->Data + done * src_pitch;

			memset(&region, 0, sizeof(region));
			r300AllocDmaRegion(rmesa, &region, lines * dstPitch,
					   1024);

			/* Copy texdata to dma:
			 */
			if (0)
				fprintf(stderr,
					"%s: src_pitch %d dst_pitch %d\n",
					__FUNCTION__, src_pitch, dstPitch);

			if (src_pitch == dstPitch) {
				memcpy(region.address + region.start, tex,
				       lines * src_pitch);
			} else {
				char *buf = region.address + region.start;
				int i;
				for (i = 0; i < lines; i++) {
					memcpy(buf, tex, src_pitch);
					buf += dstPitch;
					tex += src_pitch;
				}
			}

			r300EmitWait(rmesa, RADEON_WAIT_3D);

			/* Blit to framebuffer
			 */
			r300EmitBlit(rmesa,
				     blit_format,
				     dstPitch, GET_START(&region),
				     dstPitch, t->bufAddr,
				     0, 0, 0, done, width, lines);

			r300EmitWait(rmesa, RADEON_WAIT_2D);

			r300ReleaseDmaRegion(rmesa, &region, __FUNCTION__);
			done += lines;
		}
	}
}

/**
 * Upload the texture image associated with texture \a t at the specified
 * level at the address relative to \a start.
 */
static void uploadSubImage(r300ContextPtr rmesa, r300TexObjPtr t,
			   GLint hwlevel,
			   GLint x, GLint y, GLint width, GLint height,
			   GLuint face)
{
	struct gl_texture_image *texImage = NULL;
	GLuint offset;
	GLint imageWidth, imageHeight;
	GLint ret;
	drm_radeon_texture_t tex;
	drm_radeon_tex_image_t tmp;
	const int level = hwlevel + t->base.firstLevel;

	if (RADEON_DEBUG & DEBUG_TEXTURE) {
		fprintf(stderr,
			"%s( %p, %p ) level/width/height/face = %d/%d/%d/%u\n",
			__FUNCTION__, (void *)t, (void *)t->base.tObj, level,
			width, height, face);
	}

	ASSERT(face < 6);

	/* Ensure we have a valid texture to upload */
	if ((hwlevel < 0) || (hwlevel >= RADEON_MAX_TEXTURE_LEVELS)) {
		_mesa_problem(NULL, "bad texture level in %s", __FUNCTION__);
		return;
	}

	texImage = t->base.tObj->Image[face][level];

	if (!texImage) {
		if (RADEON_DEBUG & DEBUG_TEXTURE)
			fprintf(stderr, "%s: texImage %d is NULL!\n",
				__FUNCTION__, level);
		return;
	}
	if (!texImage->Data) {
		if (RADEON_DEBUG & DEBUG_TEXTURE)
			fprintf(stderr, "%s: image data is NULL!\n",
				__FUNCTION__);
		return;
	}

	if (t->base.tObj->Target == GL_TEXTURE_RECTANGLE_NV) {
		assert(level == 0);
		assert(hwlevel == 0);
		if (RADEON_DEBUG & DEBUG_TEXTURE)
			fprintf(stderr, "%s: image data is rectangular\n",
				__FUNCTION__);
		r300UploadRectSubImage(rmesa, t, texImage, x, y, width, height);
		return;
	} else if (texImage->IsClientData) {
		if (RADEON_DEBUG & DEBUG_TEXTURE)
			fprintf(stderr,
				"%s: image data is in GART client storage\n",
				__FUNCTION__);
		r300UploadGARTClientSubImage(rmesa, t, texImage, hwlevel, x, y,
					     width, height);
		return;
	} else if (RADEON_DEBUG & DEBUG_TEXTURE)
		fprintf(stderr, "%s: image data is in normal memory\n",
			__FUNCTION__);

	imageWidth = texImage->Width;
	imageHeight = texImage->Height;

	offset = t->bufAddr;

	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
		GLint imageX = 0;
		GLint imageY = 0;
		GLint blitX = t->image[face][hwlevel].x;
		GLint blitY = t->image[face][hwlevel].y;
		GLint blitWidth = t->image[face][hwlevel].width;
		GLint blitHeight = t->image[face][hwlevel].height;
		fprintf(stderr, "   upload image: %d,%d at %d,%d\n",
			imageWidth, imageHeight, imageX, imageY);
		fprintf(stderr, "   upload  blit: %d,%d at %d,%d\n",
			blitWidth, blitHeight, blitX, blitY);
		fprintf(stderr, "       blit ofs: 0x%07x level: %d/%d\n",
			(GLuint) offset, hwlevel, level);
	}

	t->image[face][hwlevel].data = texImage->Data;

	/* Init the DRM_RADEON_TEXTURE command / drm_radeon_texture_t struct.
	 * NOTE: we're always use a 1KB-wide blit and I8 texture format.
	 * We used to use 1, 2 and 4-byte texels and used to use the texture
	 * width to dictate the blit width - but that won't work for compressed
	 * textures. (Brian)
	 */
	tex.offset = offset;
	tex.pitch = BLIT_WIDTH_BYTES / 64;
	tex.format = R200_TXFORMAT_I8;	/* any 1-byte texel format */
	if (texImage->TexFormat->TexelBytes) {
		tex.width = imageWidth * texImage->TexFormat->TexelBytes;	/* in bytes */
		tex.height = imageHeight;
	} else {
		tex.width = imageWidth;	/* compressed */
		tex.height = imageHeight;
		if (tex.height < 4)
			tex.height = 4;
	}
	tex.image = &tmp;

	/* copy (x,y,width,height,data) */
	memcpy(&tmp, &t->image[face][hwlevel], sizeof(tmp));

	LOCK_HARDWARE(&rmesa->radeon);
	do {
		ret = drmCommandWriteRead(rmesa->radeon.dri.fd, DRM_RADEON_TEXTURE,
					  &tex, sizeof(drm_radeon_texture_t));
		if (ret) {
			if (RADEON_DEBUG & DEBUG_IOCTL)
				fprintf(stderr,
					"DRM_RADEON_TEXTURE:  again!\n");
			usleep(1);
		}
	} while (ret && errno == EAGAIN);

	UNLOCK_HARDWARE(&rmesa->radeon);

	if (ret) {
		fprintf(stderr, "DRM_RADEON_TEXTURE: return = %d\n", ret);
		fprintf(stderr, "   offset=0x%08x\n", offset);
		fprintf(stderr, "   image width=%d height=%d\n",
			imageWidth, imageHeight);
		fprintf(stderr, "    blit width=%d height=%d data=%p\n",
			t->image[face][hwlevel].width,
			t->image[face][hwlevel].height,
			t->image[face][hwlevel].data);
		exit(1);
	}
}

/**
 * Upload the texture images associated with texture \a t.  This might
 * require the allocation of texture memory.
 *
 * \param rmesa Context pointer
 * \param t Texture to be uploaded
 * \param face Cube map face to be uploaded.  Zero for non-cube maps.
 */

int r300UploadTexImages(r300ContextPtr rmesa, r300TexObjPtr t, GLuint face)
{
	const int numLevels = t->base.lastLevel - t->base.firstLevel + 1;

	if (RADEON_DEBUG & (DEBUG_TEXTURE | DEBUG_IOCTL)) {
		fprintf(stderr, "%s( %p, %p ) sz=%d lvls=%d-%d\n", __FUNCTION__,
			(void *)rmesa->radeon.glCtx, (void *)t->base.tObj,
			t->base.totalSize, t->base.firstLevel,
			t->base.lastLevel);
	}

	if (!t || t->base.totalSize == 0)
		return 0;

	if (RADEON_DEBUG & DEBUG_SYNC) {
		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
		radeonFinish(rmesa->radeon.glCtx);
	}

	LOCK_HARDWARE(&rmesa->radeon);

	if (t->base.memBlock == NULL) {
		int heap;

		heap = driAllocateTexture(rmesa->texture_heaps, rmesa->nr_heaps,
					  (driTextureObject *) t);
		if (heap == -1) {
			UNLOCK_HARDWARE(&rmesa->radeon);
			return -1;
		}

		/* Set the base offset of the texture image */
		t->bufAddr = rmesa->radeon.radeonScreen->texOffset[heap]
		    + t->base.memBlock->ofs;
		t->offset = t->bufAddr;

		/* Mark this texobj as dirty on all units:
		 */
		t->dirty_state = TEX_ALL;
	}

	/* Let the world know we've used this memory recently.
	 */
	driUpdateTextureLRU((driTextureObject *) t);
	UNLOCK_HARDWARE(&rmesa->radeon);

	/* Upload any images that are new */
	if (t->base.dirty_images[face]) {
		int i;
		for (i = 0; i < numLevels; i++) {
			if ((t->base.
			     dirty_images[face] & (1 <<
						   (i + t->base.firstLevel))) !=
			    0) {
				uploadSubImage(rmesa, t, i, 0, 0,
					       t->image[face][i].width,
					       t->image[face][i].height, face);
			}
		}
		t->base.dirty_images[face] = 0;
	}

	if (RADEON_DEBUG & DEBUG_SYNC) {
		fprintf(stderr, "%s: Syncing\n", __FUNCTION__);
		radeonFinish(rmesa->radeon.glCtx);
	}

	return 0;
}