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
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
|
/**************************************************************************
*
* Copyright 2007 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.
*
**************************************************************************/
/**
* Triangle rendering within a tile.
*/
#include <transpose_matrix4x4.h>
#include "pipe/p_compiler.h"
#include "pipe/p_format.h"
#include "util/u_math.h"
#include "spu_colorpack.h"
#include "spu_main.h"
#include "spu_shuffle.h"
#include "spu_texture.h"
#include "spu_tile.h"
#include "spu_tri.h"
/** Masks are uint[4] vectors with each element being 0 or 0xffffffff */
typedef vector unsigned int mask_t;
/**
* Simplified types taken from other parts of Gallium
*/
struct vertex_header {
vector float data[1];
};
/* XXX fix this */
#undef CEILF
#define CEILF(X) ((float) (int) ((X) + 0.99999f))
#define QUAD_TOP_LEFT 0
#define QUAD_TOP_RIGHT 1
#define QUAD_BOTTOM_LEFT 2
#define QUAD_BOTTOM_RIGHT 3
#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT)
#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT)
#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT)
#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT)
#define MASK_ALL 0xf
#define DEBUG_VERTS 0
/**
* Triangle edge info
*/
struct edge {
union {
struct {
float dx; /**< X(v1) - X(v0), used only during setup */
float dy; /**< Y(v1) - Y(v0), used only during setup */
};
vec_float4 ds; /**< vector accessor for dx and dy */
};
float dxdy; /**< dx/dy */
float sx, sy; /**< first sample point coord */
int lines; /**< number of lines on this edge */
};
struct interp_coef
{
vector float a0;
vector float dadx;
vector float dady;
};
/**
* Triangle setup info (derived from draw_stage).
* Also used for line drawing (taking some liberties).
*/
struct setup_stage {
/* Vertices are just an array of floats making up each attribute in
* turn. Currently fixed at 4 floats, but should change in time.
* Codegen will help cope with this.
*/
union {
struct {
const struct vertex_header *vmin;
const struct vertex_header *vmid;
const struct vertex_header *vmax;
const struct vertex_header *vprovoke;
};
qword vertex_headers;
};
struct edge ebot;
struct edge etop;
struct edge emaj;
float oneOverArea; /* XXX maybe make into vector? */
uint facing;
uint tx, ty; /**< position of current tile (x, y) */
int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy;
struct interp_coef coef[PIPE_MAX_SHADER_INPUTS];
struct {
vec_int4 quad; /**< [0] = row0, [1] = row1; {left[0],left[1],right[0],right[1]} */
int y;
unsigned y_flags;
unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */
} span;
};
static struct setup_stage setup;
/**
* Evaluate attribute coefficients (plane equations) to compute
* attribute values for the four fragments in a quad.
* Eg: four colors will be computed (in AoS format).
*/
static INLINE void
eval_coeff(uint slot, float x, float y, vector float w, vector float result[4])
{
switch (spu.vertex_info.attrib[slot].interp_mode) {
case INTERP_CONSTANT:
result[QUAD_TOP_LEFT] =
result[QUAD_TOP_RIGHT] =
result[QUAD_BOTTOM_LEFT] =
result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0;
break;
case INTERP_LINEAR:
{
vector float dadx = setup.coef[slot].dadx;
vector float dady = setup.coef[slot].dady;
vector float topLeft =
spu_add(setup.coef[slot].a0,
spu_add(spu_mul(spu_splats(x), dadx),
spu_mul(spu_splats(y), dady)));
result[QUAD_TOP_LEFT] = topLeft;
result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx);
result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady);
result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady);
}
break;
case INTERP_PERSPECTIVE:
{
vector float dadx = setup.coef[slot].dadx;
vector float dady = setup.coef[slot].dady;
vector float topLeft =
spu_add(setup.coef[slot].a0,
spu_add(spu_mul(spu_splats(x), dadx),
spu_mul(spu_splats(y), dady)));
vector float wInv = spu_re(w); /* 1.0 / w */
result[QUAD_TOP_LEFT] = spu_mul(topLeft, wInv);
result[QUAD_TOP_RIGHT] = spu_mul(spu_add(topLeft, dadx), wInv);
result[QUAD_BOTTOM_LEFT] = spu_mul(spu_add(topLeft, dady), wInv);
result[QUAD_BOTTOM_RIGHT] = spu_mul(spu_add(spu_add(topLeft, dadx), dady), wInv);
}
break;
case INTERP_POS:
case INTERP_NONE:
break;
default:
ASSERT(0);
}
}
/**
* As above, but return 4 vectors in SOA format.
* XXX this will all be re-written someday.
*/
static INLINE void
eval_coeff_soa(uint slot, float x, float y, vector float w, vector float result[4])
{
eval_coeff(slot, x, y, w, result);
_transpose_matrix4x4(result, result);
}
static INLINE vector float
splatz(vector float v)
{
return spu_splats(spu_extract(v, 2));
}
static INLINE vector float
splatw(vector float v)
{
return spu_splats(spu_extract(v, 3));
}
/**
* Compute quad's Z and W vectors for the quad at (x,y).
*/
static INLINE void
eval_zw(float x, float y, vector float *zOut, vector float *wOut)
{
static const vector float fragX = (const vector float) { 0.0, 1.0, 0.0, 1.0 };
static const vector float fragY = (const vector float) { 0.0, 0.0, 1.0, 1.0 };
const uint slot = 0; /* vertex position attribute */
const vector float pos = setup.coef[slot].a0;
const vector float dposdx = setup.coef[slot].dadx;
const vector float dposdy = setup.coef[slot].dady;
const vector float xVec = spu_splats(x) + fragX;
const vector float yVec = spu_splats(y) + fragY;
*zOut = splatz(pos) + xVec * splatz(dposdx) + yVec * splatz(dposdy);
*wOut = splatw(pos) + xVec * splatw(dposdx) + yVec * splatw(dposdy);
}
/**
* Emit a quad (pass to next stage). No clipping is done.
* Note: about 1/5 to 1/7 of the time, mask is zero and this function
* should be skipped. But adding the test for that slows things down
* overall.
*/
static INLINE void
emit_quad( int x, int y, mask_t mask)
{
/* If any bits in mask are set... */
if (spu_extract(spu_orx(mask), 0)) {
const int ix = x - setup.cliprect_minx;
const int iy = y - setup.cliprect_miny;
spu.cur_ctile_status = TILE_STATUS_DIRTY;
spu.cur_ztile_status = TILE_STATUS_DIRTY;
{
/*
* Run fragment shader, execute per-fragment ops, update fb/tile.
*/
vector float inputs[4*4], outputs[2*4];
vector float fragZ, fragW;
vector unsigned int kill_mask;
eval_zw((float) x, (float) y, &fragZ, &fragW);
/* setup inputs */
#if 0
eval_coeff_soa(1, (float) x, (float) y, fragW, inputs);
#else
uint i;
for (i = 0; i < spu.vertex_info.num_attribs; i++) {
eval_coeff_soa(i+1, (float) x, (float) y, fragW, inputs + i * 4);
}
#endif
ASSERT(spu.fragment_program);
ASSERT(spu.fragment_ops);
/* Execute the current fragment program */
kill_mask = spu.fragment_program(inputs, outputs, spu.constants);
mask = spu_andc(mask, kill_mask);
/* Execute per-fragment/quad operations, including:
* alpha test, z test, stencil test, blend and framebuffer writing.
* Note that there are two different fragment operations functions
* that can be called, one for front-facing fragments, and one
* for back-facing fragments. (Often the two are the same;
* but in some cases, like two-sided stenciling, they can be
* very different.) So choose the correct function depending
* on the calculated facing.
*/
spu.fragment_ops[setup.facing](ix, iy, &spu.ctile, &spu.ztile,
fragZ,
outputs[0*4+0],
outputs[0*4+1],
outputs[0*4+2],
outputs[0*4+3],
mask);
}
}
}
/**
* Given an X or Y coordinate, return the block/quad coordinate that it
* belongs to.
*/
static INLINE int
block(int x)
{
return x & ~1;
}
/**
* Render a horizontal span of quads
*/
static void
flush_spans(void)
{
int minleft, maxright;
const int l0 = spu_extract(setup.span.quad, 0);
const int l1 = spu_extract(setup.span.quad, 1);
const int r0 = spu_extract(setup.span.quad, 2);
const int r1 = spu_extract(setup.span.quad, 3);
switch (setup.span.y_flags) {
case 0x3:
/* both odd and even lines written (both quad rows) */
minleft = MIN2(l0, l1);
maxright = MAX2(r0, r1);
break;
case 0x1:
/* only even line written (quad top row) */
minleft = l0;
maxright = r0;
break;
case 0x2:
/* only odd line written (quad bottom row) */
minleft = l1;
maxright = r1;
break;
default:
return;
}
/* OK, we're very likely to need the tile data now.
* clear or finish waiting if needed.
*/
if (spu.cur_ctile_status == TILE_STATUS_GETTING) {
/* wait for mfc_get() to complete */
//printf("SPU: %u: waiting for ctile\n", spu.init.id);
wait_on_mask(1 << TAG_READ_TILE_COLOR);
spu.cur_ctile_status = TILE_STATUS_CLEAN;
}
else if (spu.cur_ctile_status == TILE_STATUS_CLEAR) {
//printf("SPU %u: clearing C tile %u, %u\n", spu.init.id, setup.tx, setup.ty);
clear_c_tile(&spu.ctile);
spu.cur_ctile_status = TILE_STATUS_DIRTY;
}
ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED);
if (spu.read_depth_stencil) {
if (spu.cur_ztile_status == TILE_STATUS_GETTING) {
/* wait for mfc_get() to complete */
//printf("SPU: %u: waiting for ztile\n", spu.init.id);
wait_on_mask(1 << TAG_READ_TILE_Z);
spu.cur_ztile_status = TILE_STATUS_CLEAN;
}
else if (spu.cur_ztile_status == TILE_STATUS_CLEAR) {
//printf("SPU %u: clearing Z tile %u, %u\n", spu.init.id, setup.tx, setup.ty);
clear_z_tile(&spu.ztile);
spu.cur_ztile_status = TILE_STATUS_DIRTY;
}
ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED);
}
/* XXX this loop could be moved into the above switch cases... */
/* Setup for mask calculation */
const vec_int4 quad_LlRr = setup.span.quad;
const vec_int4 quad_RrLl = spu_rlqwbyte(quad_LlRr, 8);
const vec_int4 quad_LLll = spu_shuffle(quad_LlRr, quad_LlRr, SHUFFLE4(A,A,B,B));
const vec_int4 quad_RRrr = spu_shuffle(quad_RrLl, quad_RrLl, SHUFFLE4(A,A,B,B));
const vec_int4 twos = spu_splats(2);
const int x = block(minleft);
vec_int4 xs = {x, x+1, x, x+1};
for (; spu_extract(xs, 0) <= block(maxright); xs += twos) {
/**
* Computes mask to indicate which pixels in the 2x2 quad are actually
* inside the triangle's bounds.
*/
/* Calculate ({x,x+1,x,x+1} >= {l[0],l[0],l[1],l[1]}) */
const mask_t gt_LLll_xs = spu_cmpgt(quad_LLll, xs);
const mask_t gte_xs_LLll = spu_nand(gt_LLll_xs, gt_LLll_xs);
/* Calculate ({r[0],r[0],r[1],r[1]} > {x,x+1,x,x+1}) */
const mask_t gt_RRrr_xs = spu_cmpgt(quad_RRrr, xs);
/* Combine results to create mask */
const mask_t mask = spu_and(gte_xs_LLll, gt_RRrr_xs);
emit_quad(spu_extract(xs, 0), setup.span.y, mask);
}
setup.span.y = 0;
setup.span.y_flags = 0;
/* Zero right elements */
setup.span.quad = spu_shuffle(setup.span.quad, setup.span.quad, SHUFFLE4(A,B,0,0));
}
#if DEBUG_VERTS
static void
print_vertex(const struct vertex_header *v)
{
uint i;
fprintf(stderr, " Vertex: (%p)\n", v);
for (i = 0; i < spu.vertex_info.num_attribs; i++) {
fprintf(stderr, " %d: %f %f %f %f\n", i,
spu_extract(v->data[i], 0),
spu_extract(v->data[i], 1),
spu_extract(v->data[i], 2),
spu_extract(v->data[i], 3));
}
}
#endif
/**
* Sort vertices from top to bottom.
* Compute area and determine front vs. back facing.
* Do coarse clip test against tile bounds
* \return FALSE if tri is totally outside tile, TRUE otherwise
*/
static boolean
setup_sort_vertices(const struct vertex_header *v0,
const struct vertex_header *v1,
const struct vertex_header *v2)
{
float area, sign;
#if DEBUG_VERTS
if (spu.init.id==0) {
fprintf(stderr, "SPU %u: Triangle:\n", spu.init.id);
print_vertex(v0);
print_vertex(v1);
print_vertex(v2);
}
#endif
/* determine bottom to top order of vertices */
{
/* A table of shuffle patterns for putting vertex_header pointers into
correct order. Quite magical. */
const vec_uchar16 sort_order_patterns[] = {
SHUFFLE4(A,B,C,C),
SHUFFLE4(C,A,B,C),
SHUFFLE4(A,C,B,C),
SHUFFLE4(B,C,A,C),
SHUFFLE4(B,A,C,C),
SHUFFLE4(C,B,A,C) };
/* The vertex_header pointers, packed for easy shuffling later */
const vec_uint4 vs = {(unsigned)v0, (unsigned)v1, (unsigned)v2};
/* Collate y values into two vectors for comparison.
Using only one shuffle constant! ;) */
const vec_float4 y_02_ = spu_shuffle(v0->data[0], v2->data[0], SHUFFLE4(0,B,b,C));
const vec_float4 y_10_ = spu_shuffle(v1->data[0], v0->data[0], SHUFFLE4(0,B,b,C));
const vec_float4 y_012 = spu_shuffle(y_02_, v1->data[0], SHUFFLE4(0,B,b,C));
const vec_float4 y_120 = spu_shuffle(y_10_, v2->data[0], SHUFFLE4(0,B,b,C));
/* Perform comparison: {y0,y1,y2} > {y1,y2,y0} */
const vec_uint4 compare = spu_cmpgt(y_012, y_120);
/* Compress the result of the comparison into 4 bits */
const vec_uint4 gather = spu_gather(compare);
/* Subtract one to attain the index into the LUT. Magical. */
const unsigned int index = spu_extract(gather, 0) - 1;
/* Load the appropriate pattern and construct the desired vector. */
setup.vertex_headers = (qword)spu_shuffle(vs, vs, sort_order_patterns[index]);
/* Using the result of the comparison, set sign.
Very magical. */
sign = ((si_to_uint(si_cntb((qword)gather)) == 2) ? 1.0f : -1.0f);
}
/* Check if triangle is completely outside the tile bounds */
if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy)
return FALSE;
if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny)
return FALSE;
if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx &&
spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx &&
spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx)
return FALSE;
if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx &&
spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx &&
spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx)
return FALSE;
setup.ebot.ds = spu_sub(setup.vmid->data[0], setup.vmin->data[0]);
setup.emaj.ds = spu_sub(setup.vmax->data[0], setup.vmin->data[0]);
setup.etop.ds = spu_sub(setup.vmax->data[0], setup.vmid->data[0]);
/*
* Compute triangle's area. Use 1/area to compute partial
* derivatives of attributes later.
*/
area = setup.emaj.dx * setup.ebot.dy - setup.ebot.dx * setup.emaj.dy;
setup.oneOverArea = 1.0f / area;
/* The product of area * sign indicates front/back orientation (0/1).
* Just in case someone gets the bright idea of switching the front
* and back constants without noticing that we're assuming their
* values in this operation, also assert that the values are
* what we think they are.
*/
ASSERT(CELL_FACING_FRONT == 0);
ASSERT(CELL_FACING_BACK == 1);
setup.facing = (area * sign > 0.0f)
^ (spu.rasterizer.front_winding == PIPE_WINDING_CW);
return TRUE;
}
/**
* Compute a0 for a constant-valued coefficient (GL_FLAT shading).
* The value value comes from vertex->data[slot].
* The result will be put into setup.coef[slot].a0.
* \param slot which attribute slot
*/
static INLINE void
const_coeff4(uint slot)
{
setup.coef[slot].dadx = (vector float) {0.0, 0.0, 0.0, 0.0};
setup.coef[slot].dady = (vector float) {0.0, 0.0, 0.0, 0.0};
setup.coef[slot].a0 = setup.vprovoke->data[slot];
}
/**
* As above, but interp setup all four vector components.
*/
static INLINE void
tri_linear_coeff4(uint slot)
{
const vector float vmin_d = setup.vmin->data[slot];
const vector float vmid_d = setup.vmid->data[slot];
const vector float vmax_d = setup.vmax->data[slot];
const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f);
const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f);
vector float botda = vmid_d - vmin_d;
vector float majda = vmax_d - vmin_d;
vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda),
spu_mul(botda, spu_splats(setup.emaj.dy)));
vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda),
spu_mul(majda, spu_splats(setup.ebot.dx)));
setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea));
setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea));
vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx);
vector float tempy = spu_mul(setup.coef[slot].dady, yyyy);
setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy));
}
/**
* Compute a0, dadx and dady for a perspective-corrected interpolant,
* for a triangle.
* We basically multiply the vertex value by 1/w before computing
* the plane coefficients (a0, dadx, dady).
* Later, when we compute the value at a particular fragment position we'll
* divide the interpolated value by the interpolated W at that fragment.
*/
static void
tri_persp_coeff4(uint slot)
{
const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f);
const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f);
const vector float vmin_w = spu_splats(spu_extract(setup.vmin->data[0], 3));
const vector float vmid_w = spu_splats(spu_extract(setup.vmid->data[0], 3));
const vector float vmax_w = spu_splats(spu_extract(setup.vmax->data[0], 3));
vector float vmin_d = setup.vmin->data[slot];
vector float vmid_d = setup.vmid->data[slot];
vector float vmax_d = setup.vmax->data[slot];
vmin_d = spu_mul(vmin_d, vmin_w);
vmid_d = spu_mul(vmid_d, vmid_w);
vmax_d = spu_mul(vmax_d, vmax_w);
vector float botda = vmid_d - vmin_d;
vector float majda = vmax_d - vmin_d;
vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda),
spu_mul(botda, spu_splats(setup.emaj.dy)));
vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda),
spu_mul(majda, spu_splats(setup.ebot.dx)));
setup.coef[slot].dadx = spu_mul(a, spu_splats(setup.oneOverArea));
setup.coef[slot].dady = spu_mul(b, spu_splats(setup.oneOverArea));
vector float tempx = spu_mul(setup.coef[slot].dadx, xxxx);
vector float tempy = spu_mul(setup.coef[slot].dady, yyyy);
setup.coef[slot].a0 = spu_sub(vmin_d, spu_add(tempx, tempy));
}
/**
* Compute the setup.coef[] array dadx, dady, a0 values.
* Must be called after setup.vmin,vmid,vmax,vprovoke are initialized.
*/
static void
setup_tri_coefficients(void)
{
uint i;
for (i = 0; i < spu.vertex_info.num_attribs; i++) {
switch (spu.vertex_info.attrib[i].interp_mode) {
case INTERP_NONE:
break;
case INTERP_CONSTANT:
const_coeff4(i);
break;
case INTERP_POS:
/* fall-through */
case INTERP_LINEAR:
tri_linear_coeff4(i);
break;
case INTERP_PERSPECTIVE:
tri_persp_coeff4(i);
break;
default:
ASSERT(0);
}
}
}
static void
setup_tri_edges(void)
{
float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f;
float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f;
float vmin_y = spu_extract(setup.vmin->data[0], 1) - 0.5f;
float vmid_y = spu_extract(setup.vmid->data[0], 1) - 0.5f;
float vmax_y = spu_extract(setup.vmax->data[0], 1) - 0.5f;
setup.emaj.sy = CEILF(vmin_y);
setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy);
setup.emaj.dxdy = setup.emaj.dx / setup.emaj.dy;
setup.emaj.sx = vmin_x + (setup.emaj.sy - vmin_y) * setup.emaj.dxdy;
setup.etop.sy = CEILF(vmid_y);
setup.etop.lines = (int) CEILF(vmax_y - setup.etop.sy);
setup.etop.dxdy = setup.etop.dx / setup.etop.dy;
setup.etop.sx = vmid_x + (setup.etop.sy - vmid_y) * setup.etop.dxdy;
setup.ebot.sy = CEILF(vmin_y);
setup.ebot.lines = (int) CEILF(vmid_y - setup.ebot.sy);
setup.ebot.dxdy = setup.ebot.dx / setup.ebot.dy;
setup.ebot.sx = vmin_x + (setup.ebot.sy - vmin_y) * setup.ebot.dxdy;
}
/**
* Render the upper or lower half of a triangle.
* Scissoring/cliprect is applied here too.
*/
static void
subtriangle(struct edge *eleft, struct edge *eright, unsigned lines)
{
const int minx = setup.cliprect_minx;
const int maxx = setup.cliprect_maxx;
const int miny = setup.cliprect_miny;
const int maxy = setup.cliprect_maxy;
int y, start_y, finish_y;
int sy = (int)eleft->sy;
ASSERT((int)eleft->sy == (int) eright->sy);
/* clip top/bottom */
start_y = sy;
finish_y = sy + lines;
if (start_y < miny)
start_y = miny;
if (finish_y > maxy)
finish_y = maxy;
start_y -= sy;
finish_y -= sy;
/*
_mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y);
*/
for (y = start_y; y < finish_y; y++) {
/* avoid accumulating adds as floats don't have the precision to
* accurately iterate large triangle edges that way. luckily we
* can just multiply these days.
*
* this is all drowned out by the attribute interpolation anyway.
*/
int left = (int)(eleft->sx + y * eleft->dxdy);
int right = (int)(eright->sx + y * eright->dxdy);
/* clip left/right */
if (left < minx)
left = minx;
if (right > maxx)
right = maxx;
if (left < right) {
int _y = sy + y;
if (block(_y) != setup.span.y) {
flush_spans();
setup.span.y = block(_y);
}
int offset = _y&1;
vec_int4 quad_LlRr = {left, left, right, right};
/* Store left and right in 0 or 1 row of quad based on offset */
setup.span.quad = spu_sel(quad_LlRr, setup.span.quad, spu_maskw(5<<offset));
setup.span.y_flags |= 1<<offset;
}
}
/* save the values so that emaj can be restarted:
*/
eleft->sx += lines * eleft->dxdy;
eright->sx += lines * eright->dxdy;
eleft->sy += lines;
eright->sy += lines;
}
/**
* Draw triangle into tile at (tx, ty) (tile coords)
* The tile data should have already been fetched.
*/
boolean
tri_draw(const float *v0, const float *v1, const float *v2,
uint tx, uint ty)
{
setup.tx = tx;
setup.ty = ty;
/* set clipping bounds to tile bounds */
setup.cliprect_minx = tx * TILE_SIZE;
setup.cliprect_miny = ty * TILE_SIZE;
setup.cliprect_maxx = (tx + 1) * TILE_SIZE;
setup.cliprect_maxy = (ty + 1) * TILE_SIZE;
if (!setup_sort_vertices((struct vertex_header *) v0,
(struct vertex_header *) v1,
(struct vertex_header *) v2)) {
return FALSE; /* totally clipped */
}
setup_tri_coefficients();
setup_tri_edges();
setup.span.y = 0;
setup.span.y_flags = 0;
/* Zero right elements */
setup.span.quad = spu_shuffle(setup.span.quad, setup.span.quad, SHUFFLE4(A,B,0,0));
if (setup.oneOverArea < 0.0) {
/* emaj on left */
subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines );
subtriangle( &setup.emaj, &setup.etop, setup.etop.lines );
}
else {
/* emaj on right */
subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines );
subtriangle( &setup.etop, &setup.emaj, setup.etop.lines );
}
flush_spans();
return TRUE;
}
|