summaryrefslogtreecommitdiffstats
path: root/src/gallium/drivers/cell/spu/spu_texture.c
blob: 69784c89788c7cd72bf2d656192dd55e4b7e51eb (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
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
/**************************************************************************
 * 
 * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
 * All Rights Reserved.
 *
 * 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 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 TUNGSTEN GRAPHICS AND/OR ITS 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.
 * 
 **************************************************************************/


#include <math.h>

#include "pipe/p_compiler.h"
#include "spu_main.h"
#include "spu_texture.h"
#include "spu_tile.h"
#include "spu_colorpack.h"
#include "spu_dcache.h"


/**
 * Mark all tex cache entries as invalid.
 */
void
invalidate_tex_cache(void)
{
   uint lvl;
   for (lvl = 0; lvl < CELL_MAX_TEXTURE_LEVELS; lvl++) {
      uint unit = 0;
      uint bytes = 4 * spu.texture[unit].level[lvl].width
         * spu.texture[unit].level[lvl].height;

      if (spu.texture[unit].target == PIPE_TEXTURE_CUBE)
         bytes *= 6;
      else if (spu.texture[unit].target == PIPE_TEXTURE_3D)
         bytes *= spu.texture[unit].level[lvl].depth;

      spu_dcache_mark_dirty((unsigned) spu.texture[unit].level[lvl].start, bytes);
   }
}


/**
 * Get four texels from locations (x[0], y[0]), (x[1], y[1]) ...
 *
 * NOTE: in the typical case of bilinear filtering, the four texels
 * are in a 2x2 group so we could get by with just two dcache fetches
 * (two side-by-side texels per fetch).  But when bilinear filtering
 * wraps around a texture edge, we'll probably need code like we have
 * now.
 * FURTHERMORE: since we're rasterizing a quad of 2x2 pixels at a time,
 * it's quite likely that the four pixels in a quad will need some of the
 * same texels.  So look into doing texture fetches for four pixels at
 * a time.
 */
static void
get_four_texels(const struct spu_texture_level *tlevel, uint face,
                vec_int4 x, vec_int4 y,
                vec_uint4 *texels)
{
   unsigned texture_ea = (uintptr_t) tlevel->start;
   const vec_int4 tile_x = spu_rlmask(x, -5);  /* tile_x = x / 32 */
   const vec_int4 tile_y = spu_rlmask(y, -5);  /* tile_y = y / 32 */
   const qword offset_x = si_andi((qword) x, 0x1f); /* offset_x = x & 0x1f */
   const qword offset_y = si_andi((qword) y, 0x1f); /* offset_y = y & 0x1f */

   const qword tiles_per_row = (qword) spu_splats(tlevel->tiles_per_row);
   const qword tile_size = (qword) spu_splats((unsigned) sizeof(tile_t));

   qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x);
   tile_offset = si_mpy((qword) tile_offset, tile_size);

   qword texel_offset = si_a(si_mpyui(offset_y, 32), offset_x);
   texel_offset = si_mpyui(texel_offset, 4);
   
   vec_uint4 offset = (vec_uint4) si_a(tile_offset, texel_offset);
   
   texture_ea = texture_ea + face * tlevel->bytes_per_image;

   spu_dcache_fetch_unaligned((qword *) & texels[0],
                              texture_ea + spu_extract(offset, 0), 4);
   spu_dcache_fetch_unaligned((qword *) & texels[1],
                              texture_ea + spu_extract(offset, 1), 4);
   spu_dcache_fetch_unaligned((qword *) & texels[2],
                              texture_ea + spu_extract(offset, 2), 4);
   spu_dcache_fetch_unaligned((qword *) & texels[3],
                              texture_ea + spu_extract(offset, 3), 4);
}


/** clamp vec to [0, max] */
static INLINE vector signed int
spu_clamp(vector signed int vec, vector signed int max)
{
   static const vector signed int zero = {0,0,0,0};
   vector unsigned int c;
   c = spu_cmpgt(vec, zero);    /* c = vec > zero ? ~0 : 0 */
   vec = spu_sel(zero, vec, c);
   c = spu_cmpgt(vec, max);    /* c = vec > max ? ~0 : 0 */
   vec = spu_sel(vec, max, c);
   return vec;
}



/**
 * Do nearest texture sampling for four pixels.
 * \param colors  returned colors in SOA format (rrrr, gggg, bbbb, aaaa).
 */
void
sample_texture_2d_nearest(vector float s, vector float t,
                          uint unit, uint level, uint face,
                          vector float colors[4])
{
   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
   vector float ss = spu_mul(s, tlevel->scale_s);
   vector float tt = spu_mul(t, tlevel->scale_t);
   vector signed int is = spu_convts(ss, 0);
   vector signed int it = spu_convts(tt, 0);
   vec_uint4 texels[4];

   /* PIPE_TEX_WRAP_REPEAT */
   is = spu_and(is, tlevel->mask_s);
   it = spu_and(it, tlevel->mask_t);

   /* PIPE_TEX_WRAP_CLAMP */
   is = spu_clamp(is, tlevel->max_s);
   it = spu_clamp(it, tlevel->max_t);

   get_four_texels(tlevel, face, is, it, texels);

   /* convert four packed ARGBA pixels to float RRRR,GGGG,BBBB,AAAA */
   spu_unpack_A8R8G8B8_transpose4(texels, colors);
}


/**
 * Do bilinear texture sampling for four pixels.
 * \param colors  returned colors in SOA format (rrrr, gggg, bbbb, aaaa).
 */
void
sample_texture_2d_bilinear(vector float s, vector float t,
                           uint unit, uint level, uint face,
                           vector float colors[4])
{
   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
   static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f};

   vector float ss = spu_madd(s, tlevel->scale_s, half);
   vector float tt = spu_madd(t, tlevel->scale_t, half);

   vector signed int is0 = spu_convts(ss, 0);
   vector signed int it0 = spu_convts(tt, 0);

   /* is + 1, it + 1 */
   vector signed int is1 = spu_add(is0, 1);
   vector signed int it1 = spu_add(it0, 1);

   /* PIPE_TEX_WRAP_REPEAT */
   is0 = spu_and(is0, tlevel->mask_s);
   it0 = spu_and(it0, tlevel->mask_t);
   is1 = spu_and(is1, tlevel->mask_s);
   it1 = spu_and(it1, tlevel->mask_t);

   /* PIPE_TEX_WRAP_CLAMP */
   is0 = spu_clamp(is0, tlevel->max_s);
   it0 = spu_clamp(it0, tlevel->max_t);
   is1 = spu_clamp(is1, tlevel->max_s);
   it1 = spu_clamp(it1, tlevel->max_t);

   /* get packed int texels */
   vector unsigned int texels[16];
   get_four_texels(tlevel, face, is0, it0, texels + 0);  /* upper-left */
   get_four_texels(tlevel, face, is1, it0, texels + 4);  /* upper-right */
   get_four_texels(tlevel, face, is0, it1, texels + 8);  /* lower-left */
   get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */

   /* convert packed int texels to float colors */
   vector float ftexels[16];
   spu_unpack_A8R8G8B8_transpose4(texels + 0, ftexels + 0);
   spu_unpack_A8R8G8B8_transpose4(texels + 4, ftexels + 4);
   spu_unpack_A8R8G8B8_transpose4(texels + 8, ftexels + 8);
   spu_unpack_A8R8G8B8_transpose4(texels + 12, ftexels + 12);

   /* Compute weighting factors in [0,1]
    * Multiply texcoord by 1024, AND with 1023, convert back to float.
    */
   vector float ss1024 = spu_mul(ss, spu_splats(1024.0f));
   vector signed int iss1024 = spu_convts(ss1024, 0);
   iss1024 = spu_and(iss1024, 1023);
   vector float sWeights0 = spu_convtf(iss1024, 10);

   vector float tt1024 = spu_mul(tt, spu_splats(1024.0f));
   vector signed int itt1024 = spu_convts(tt1024, 0);
   itt1024 = spu_and(itt1024, 1023);
   vector float tWeights0 = spu_convtf(itt1024, 10);

   /* 1 - sWeight and 1 - tWeight */
   vector float sWeights1 = spu_sub(spu_splats(1.0f), sWeights0);
   vector float tWeights1 = spu_sub(spu_splats(1.0f), tWeights0);

   /* reds, for four pixels */
   ftexels[ 0] = spu_mul(ftexels[ 0], spu_mul(sWeights1, tWeights1)); /*ul*/
   ftexels[ 4] = spu_mul(ftexels[ 4], spu_mul(sWeights0, tWeights1)); /*ur*/
   ftexels[ 8] = spu_mul(ftexels[ 8], spu_mul(sWeights1, tWeights0)); /*ll*/
   ftexels[12] = spu_mul(ftexels[12], spu_mul(sWeights0, tWeights0)); /*lr*/
   colors[0] = spu_add(spu_add(ftexels[0], ftexels[4]),
                       spu_add(ftexels[8], ftexels[12]));

   /* greens, for four pixels */
   ftexels[ 1] = spu_mul(ftexels[ 1], spu_mul(sWeights1, tWeights1)); /*ul*/
   ftexels[ 5] = spu_mul(ftexels[ 5], spu_mul(sWeights0, tWeights1)); /*ur*/
   ftexels[ 9] = spu_mul(ftexels[ 9], spu_mul(sWeights1, tWeights0)); /*ll*/
   ftexels[13] = spu_mul(ftexels[13], spu_mul(sWeights0, tWeights0)); /*lr*/
   colors[1] = spu_add(spu_add(ftexels[1], ftexels[5]),
                       spu_add(ftexels[9], ftexels[13]));

   /* blues, for four pixels */
   ftexels[ 2] = spu_mul(ftexels[ 2], spu_mul(sWeights1, tWeights1)); /*ul*/
   ftexels[ 6] = spu_mul(ftexels[ 6], spu_mul(sWeights0, tWeights1)); /*ur*/
   ftexels[10] = spu_mul(ftexels[10], spu_mul(sWeights1, tWeights0)); /*ll*/
   ftexels[14] = spu_mul(ftexels[14], spu_mul(sWeights0, tWeights0)); /*lr*/
   colors[2] = spu_add(spu_add(ftexels[2], ftexels[6]),
                       spu_add(ftexels[10], ftexels[14]));

   /* alphas, for four pixels */
   ftexels[ 3] = spu_mul(ftexels[ 3], spu_mul(sWeights1, tWeights1)); /*ul*/
   ftexels[ 7] = spu_mul(ftexels[ 7], spu_mul(sWeights0, tWeights1)); /*ur*/
   ftexels[11] = spu_mul(ftexels[11], spu_mul(sWeights1, tWeights0)); /*ll*/
   ftexels[15] = spu_mul(ftexels[15], spu_mul(sWeights0, tWeights0)); /*lr*/
   colors[3] = spu_add(spu_add(ftexels[3], ftexels[7]),
                       spu_add(ftexels[11], ftexels[15]));
}



/**
 * Adapted from /opt/cell/sdk/usr/spu/include/transpose_matrix4x4.h
 */
static INLINE void
transpose(vector unsigned int *mOut0,
          vector unsigned int *mOut1,
          vector unsigned int *mOut2,
          vector unsigned int *mOut3,
          vector unsigned int *mIn)
{
  vector unsigned int abcd, efgh, ijkl, mnop;	/* input vectors */
  vector unsigned int aeim, bfjn, cgko, dhlp;	/* output vectors */
  vector unsigned int aibj, ckdl, emfn, gohp;	/* intermediate vectors */

  vector unsigned char shufflehi = ((vector unsigned char) {
					       0x00, 0x01, 0x02, 0x03,
					       0x10, 0x11, 0x12, 0x13,
					       0x04, 0x05, 0x06, 0x07,
					       0x14, 0x15, 0x16, 0x17});
  vector unsigned char shufflelo = ((vector unsigned char) {
					       0x08, 0x09, 0x0A, 0x0B,
					       0x18, 0x19, 0x1A, 0x1B,
					       0x0C, 0x0D, 0x0E, 0x0F,
					       0x1C, 0x1D, 0x1E, 0x1F});
  abcd = *(mIn+0);
  efgh = *(mIn+1);
  ijkl = *(mIn+2);
  mnop = *(mIn+3);

  aibj = spu_shuffle(abcd, ijkl, shufflehi);
  ckdl = spu_shuffle(abcd, ijkl, shufflelo);
  emfn = spu_shuffle(efgh, mnop, shufflehi);
  gohp = spu_shuffle(efgh, mnop, shufflelo);

  aeim = spu_shuffle(aibj, emfn, shufflehi);
  bfjn = spu_shuffle(aibj, emfn, shufflelo);
  cgko = spu_shuffle(ckdl, gohp, shufflehi);
  dhlp = spu_shuffle(ckdl, gohp, shufflelo);

  *mOut0 = aeim;
  *mOut1 = bfjn;
  *mOut2 = cgko;
  *mOut3 = dhlp;
}


/**
 * Bilinear filtering, using int instead of float arithmetic for computing
 * sample weights.
 */
void
sample_texture_2d_bilinear_int(vector float s, vector float t,
                               uint unit, uint level, uint face,
                               vector float colors[4])
{
   const struct spu_texture_level *tlevel = &spu.texture[unit].level[level];
   static const vector float half = {-0.5f, -0.5f, -0.5f, -0.5f};

   /* Scale texcoords by size of texture, and add half pixel bias */
   vector float ss = spu_madd(s, tlevel->scale_s, half);
   vector float tt = spu_madd(t, tlevel->scale_t, half);

   /* convert float coords to fixed-pt coords with 7 fraction bits */
   vector signed int is = spu_convts(ss, 7);  /* XXX really need floor() here */
   vector signed int it = spu_convts(tt, 7);  /* XXX really need floor() here */

   /* compute integer texel weights in [0, 127] */
   vector signed int sWeights0 = spu_and(is, 127);
   vector signed int tWeights0 = spu_and(it, 127);
   vector signed int sWeights1 = spu_sub(127, sWeights0);
   vector signed int tWeights1 = spu_sub(127, tWeights0);

   /* texel coords: is0 = is / 128, it0 = is / 128 */
   vector signed int is0 = spu_rlmask(is, -7);
   vector signed int it0 = spu_rlmask(it, -7);

   /* texel coords: i1 = is0 + 1, it1 = it0 + 1 */
   vector signed int is1 = spu_add(is0, 1);
   vector signed int it1 = spu_add(it0, 1);

   /* PIPE_TEX_WRAP_REPEAT */
   is0 = spu_and(is0, tlevel->mask_s);
   it0 = spu_and(it0, tlevel->mask_t);
   is1 = spu_and(is1, tlevel->mask_s);
   it1 = spu_and(it1, tlevel->mask_t);

   /* PIPE_TEX_WRAP_CLAMP */
   is0 = spu_clamp(is0, tlevel->max_s);
   it0 = spu_clamp(it0, tlevel->max_t);
   is1 = spu_clamp(is1, tlevel->max_s);
   it1 = spu_clamp(it1, tlevel->max_t);

   /* get packed int texels */
   vector unsigned int texels[16];
   get_four_texels(tlevel, face, is0, it0, texels + 0);  /* upper-left */
   get_four_texels(tlevel, face, is1, it0, texels + 4);  /* upper-right */
   get_four_texels(tlevel, face, is0, it1, texels + 8);  /* lower-left */
   get_four_texels(tlevel, face, is1, it1, texels + 12); /* lower-right */

   /* twiddle packed 32-bit BGRA pixels into RGBA as four unsigned ints */
   {
      static const unsigned char ZERO = 0x80;
      int i;
      for (i = 0; i < 16; i++) {
         texels[i] = spu_shuffle(texels[i], texels[i],
                                 ((vector unsigned char) {
                                    ZERO, ZERO, ZERO, 1,
                                    ZERO, ZERO, ZERO, 2,
                                    ZERO, ZERO, ZERO, 3,
                                    ZERO, ZERO, ZERO, 0}));
      }
   }

   /* convert RGBA,RGBA,RGBA,RGBA to RRRR,GGGG,BBBB,AAAA */
   vector unsigned int texel0, texel1, texel2, texel3, texel4, texel5, texel6, texel7,
      texel8, texel9, texel10, texel11, texel12, texel13, texel14, texel15;
   transpose(&texel0, &texel1, &texel2, &texel3, texels + 0);
   transpose(&texel4, &texel5, &texel6, &texel7, texels + 4);
   transpose(&texel8, &texel9, &texel10, &texel11, texels + 8);
   transpose(&texel12, &texel13, &texel14, &texel15, texels + 12);

   /* computed weighted colors */
   vector unsigned int c0, c1, c2, c3, cSum;

   /* red */
   c0 = (vector unsigned int) si_mpy((qword) texel0, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
   c1 = (vector unsigned int) si_mpy((qword) texel4, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
   c2 = (vector unsigned int) si_mpy((qword) texel8, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
   c3 = (vector unsigned int) si_mpy((qword) texel12, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
   colors[0] = spu_convtf(cSum, 22);

   /* green */
   c0 = (vector unsigned int) si_mpy((qword) texel1, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
   c1 = (vector unsigned int) si_mpy((qword) texel5, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
   c2 = (vector unsigned int) si_mpy((qword) texel9, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
   c3 = (vector unsigned int) si_mpy((qword) texel13, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
   colors[1] = spu_convtf(cSum, 22);

   /* blue */
   c0 = (vector unsigned int) si_mpy((qword) texel2, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
   c1 = (vector unsigned int) si_mpy((qword) texel6, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
   c2 = (vector unsigned int) si_mpy((qword) texel10, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
   c3 = (vector unsigned int) si_mpy((qword) texel14, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
   colors[2] = spu_convtf(cSum, 22);

   /* alpha */
   c0 = (vector unsigned int) si_mpy((qword) texel3, si_mpy((qword) sWeights1, (qword) tWeights1)); /*ul*/
   c1 = (vector unsigned int) si_mpy((qword) texel7, si_mpy((qword) sWeights0, (qword) tWeights1)); /*ur*/
   c2 = (vector unsigned int) si_mpy((qword) texel11, si_mpy((qword) sWeights1, (qword) tWeights0)); /*ll*/
   c3 = (vector unsigned int) si_mpy((qword) texel15, si_mpy((qword) sWeights0, (qword) tWeights0)); /*lr*/
   cSum = spu_add(spu_add(c0, c1), spu_add(c2, c3));
   colors[3] = spu_convtf(cSum, 22);
}



/**
 * Compute level of detail factor from texcoords.
 */
static INLINE float
compute_lambda_2d(uint unit, vector float s, vector float t)
{
   uint baseLevel = 0;
   float width = spu.texture[unit].level[baseLevel].width;
   float height = spu.texture[unit].level[baseLevel].width;
   float dsdx = width * (spu_extract(s, 1) - spu_extract(s, 0));
   float dsdy = width * (spu_extract(s, 2) - spu_extract(s, 0));
   float dtdx = height * (spu_extract(t, 1) - spu_extract(t, 0));
   float dtdy = height * (spu_extract(t, 2) - spu_extract(t, 0));
#if 0
   /* ideal value */
   float x = dsdx * dsdx + dtdx * dtdx;
   float y = dsdy * dsdy + dtdy * dtdy;
   float rho = x > y ? x : y;
   rho = sqrtf(rho);
#else
   /* approximation */
   dsdx = fabsf(dsdx);
   dsdy = fabsf(dsdy);
   dtdx = fabsf(dtdx);
   dtdy = fabsf(dtdy);
   float rho = (dsdx + dsdy + dtdx + dtdy) * 0.5;
#endif
   float lambda = logf(rho) * 1.442695f; /* compute logbase2(rho) */
   return lambda;
}


/**
 * Blend two sets of colors according to weight.
 */
static void
blend_colors(vector float c0[4], const vector float c1[4], float weight)
{
   vector float t = spu_splats(weight);
   vector float dc0 = spu_sub(c1[0], c0[0]);
   vector float dc1 = spu_sub(c1[1], c0[1]);
   vector float dc2 = spu_sub(c1[2], c0[2]);
   vector float dc3 = spu_sub(c1[3], c0[3]);
   c0[0] = spu_madd(dc0, t, c0[0]);
   c0[1] = spu_madd(dc1, t, c0[1]);
   c0[2] = spu_madd(dc2, t, c0[2]);
   c0[3] = spu_madd(dc3, t, c0[3]);
}


/**
 * Texture sampling with level of detail selection and possibly mipmap
 * interpolation.
 */
void
sample_texture_2d_lod(vector float s, vector float t,
                      uint unit, uint level_ignored, uint face,
                      vector float colors[4])
{
   /*
    * Note that we're computing a lambda/lod here that's used for all
    * four pixels in the quad.
    */
   float lambda = compute_lambda_2d(unit, s, t);

   (void) face;
   (void) level_ignored;

   /* apply lod bias */
   lambda += spu.sampler[unit].lod_bias;

   /* clamp */
   if (lambda < spu.sampler[unit].min_lod)
      lambda = spu.sampler[unit].min_lod;
   else if (lambda > spu.sampler[unit].max_lod)
      lambda = spu.sampler[unit].max_lod;

   if (lambda <= 0.0f) {
      /* magnify */
      spu.mag_sample_texture_2d[unit](s, t, unit, 0, face, colors);
   }
   else {
      /* minify */
      if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) {
         /* sample two mipmap levels and interpolate */
         int level = (int) lambda;
         if (level > (int) spu.texture[unit].max_level)
            level = spu.texture[unit].max_level;
         spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors);
         if (spu.sampler[unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) {
            /* sample second mipmap level */
            float weight = lambda - (float) level;
            level++;
            if (level <= (int) spu.texture[unit].max_level) {
               vector float colors2[4];
               spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors2);
               blend_colors(colors, colors2, weight);
            }
         }
      }
      else {
         /* sample one mipmap level */
         int level = (int) (lambda + 0.5f);
         if (level > (int) spu.texture[unit].max_level)
            level = spu.texture[unit].max_level;
         spu.min_sample_texture_2d[unit](s, t, unit, level, face, colors);
      }
   }
}


/** XXX need a SIMD version of this */
static unsigned
choose_cube_face(float rx, float ry, float rz, float *newS, float *newT)
{
   /*
      major axis
      direction     target                             sc     tc    ma
      ----------    -------------------------------    ---    ---   ---
       +rx          TEXTURE_CUBE_MAP_POSITIVE_X_EXT    -rz    -ry   rx
       -rx          TEXTURE_CUBE_MAP_NEGATIVE_X_EXT    +rz    -ry   rx
       +ry          TEXTURE_CUBE_MAP_POSITIVE_Y_EXT    +rx    +rz   ry
       -ry          TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT    +rx    -rz   ry
       +rz          TEXTURE_CUBE_MAP_POSITIVE_Z_EXT    +rx    -ry   rz
       -rz          TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT    -rx    -ry   rz
   */
   const float arx = fabsf(rx);
   const float ary = fabsf(ry);
   const float arz = fabsf(rz);
   unsigned face;
   float sc, tc, ma;

   if (arx > ary && arx > arz) {
      if (rx >= 0.0F) {
         face = PIPE_TEX_FACE_POS_X;
         sc = -rz;
         tc = -ry;
         ma = arx;
      }
      else {
         face = PIPE_TEX_FACE_NEG_X;
         sc = rz;
         tc = -ry;
         ma = arx;
      }
   }
   else if (ary > arx && ary > arz) {
      if (ry >= 0.0F) {
         face = PIPE_TEX_FACE_POS_Y;
         sc = rx;
         tc = rz;
         ma = ary;
      }
      else {
         face = PIPE_TEX_FACE_NEG_Y;
         sc = rx;
         tc = -rz;
         ma = ary;
      }
   }
   else {
      if (rz > 0.0F) {
         face = PIPE_TEX_FACE_POS_Z;
         sc = rx;
         tc = -ry;
         ma = arz;
      }
      else {
         face = PIPE_TEX_FACE_NEG_Z;
         sc = -rx;
         tc = -ry;
         ma = arz;
      }
   }

   *newS = (sc / ma + 1.0F) * 0.5F;
   *newT = (tc / ma + 1.0F) * 0.5F;

   return face;
}



void
sample_texture_cube(vector float s, vector float t, vector float r,
                    uint unit, vector float colors[4])
{
   uint p, faces[4], level = 0;
   float newS[4], newT[4];

   /* Compute cube faces referenced by the four sets of texcoords.
    * XXX we should SIMD-ize this.
    */
   for (p = 0; p < 4; p++) {      
      float rx = spu_extract(s, p);
      float ry = spu_extract(t, p);
      float rz = spu_extract(r, p);
      faces[p] = choose_cube_face(rx, ry, rz, &newS[p], &newT[p]);
   }

   if (faces[0] == faces[1] &&
       faces[0] == faces[2] &&
       faces[0] == faces[3]) {
      /* GOOD!  All four texcoords refer to the same cube face */
      s = (vector float) {newS[0], newS[1], newS[2], newS[3]};
      t = (vector float) {newT[0], newT[1], newT[2], newT[3]};
      spu.sample_texture_2d[unit](s, t, unit, level, faces[0], colors);
   }
   else {
      /* BAD!  The four texcoords refer to different faces */
      for (p = 0; p < 4; p++) {      
         vector float c[4];

         spu.sample_texture_2d[unit](spu_splats(newS[p]), spu_splats(newT[p]),
                                     unit, level, faces[p], c);

         float red = spu_extract(c[0], p);
         float green = spu_extract(c[1], p);
         float blue = spu_extract(c[2], p);
         float alpha = spu_extract(c[3], p);

         colors[0] = spu_insert(red,   colors[0], p);
         colors[1] = spu_insert(green, colors[1], p);
         colors[2] = spu_insert(blue,  colors[2], p);
         colors[3] = spu_insert(alpha, colors[3], p);
      }
   }
}