psx_gpu: workaround overflow crash
[pcsx_rearmed.git] / plugins / gpu_neon / psx_gpu / psx_gpu.c
1 /*
2  * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
3  *
4  * This program is free software; you can redistribute it and/or
5  * modify it under the terms of the GNU General Public License as
6  * published by the Free Software Foundation; either version 2 of
7  * the License, or (at your option) any later version.
8  *
9  * This program is distributed in the hope that it will be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
12  * General Public License for more details.
13  */
14
15 #include <stdio.h>
16 #include <stdlib.h>
17 #include <string.h>
18
19 #include "common.h"
20
21 u32 span_pixels = 0;
22 u32 span_pixel_blocks = 0;
23 u32 spans = 0;
24 u32 triangles = 0;
25 u32 sprites = 0;
26 u32 sprites_4bpp = 0;
27 u32 sprites_8bpp = 0;
28 u32 sprites_16bpp = 0;
29 u32 sprite_blocks = 0;
30 u32 sprites_untextured = 0;
31 u32 lines = 0;
32 u32 trivial_rejects = 0;
33 u32 texels_4bpp = 0;
34 u32 texels_8bpp = 0;
35 u32 texels_16bpp = 0;
36 u32 texel_blocks_4bpp = 0;
37 u32 texel_blocks_8bpp = 0;
38 u32 texel_blocks_16bpp = 0;
39 u32 texel_blocks_untextured = 0;
40 u32 blend_blocks = 0;
41 u32 render_buffer_flushes = 0;
42 u32 state_changes = 0;
43 u32 left_split_triangles = 0;
44 u32 flat_triangles = 0;
45 u32 clipped_triangles = 0;
46 u32 zero_block_spans = 0;
47 u32 texture_cache_loads = 0;
48 u32 false_modulated_blocks = 0;
49
50 /* double size for enhancement */
51 u32 reciprocal_table[512 * 2];
52
53
54 typedef s32 fixed_type;
55
56 #define EDGE_STEP_BITS 32
57 #define FIXED_BITS     12
58
59 #define fixed_center(value)                                                    \
60   ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1)))            \
61
62 #define int_to_fixed(value)                                                    \
63   (((fixed_type)(value)) << FIXED_BITS)                                        \
64
65 #define fixed_to_int(value)                                                    \
66   ((value) >> FIXED_BITS)                                                      \
67
68 #define fixed_to_double(value)                                                 \
69   ((value) / (double)(1 << FIXED_BITS))                                        \
70
71 #define double_to_fixed(value)                                                 \
72   (fixed_type)(((value) * (double)(1 << FIXED_BITS)))                          \
73
74 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
75 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
76 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
77 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
78
79 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
80  s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
81
82 struct render_block_handler_struct
83 {
84   void *setup_blocks;
85   texture_blocks_function_type *texture_blocks;
86   shade_blocks_function_type *shade_blocks;
87   blend_blocks_function_type *blend_blocks; 
88 };
89
90 #ifndef NEON_BUILD
91
92 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
93 {
94   u32 shift = __builtin_clz(denominator);
95   u32 denominator_normalized = denominator << shift;
96
97   double numerator = (1ULL << 62) + denominator_normalized;
98   double numerator_b;
99
100   double denominator_normalized_dp_b;
101   u64 denominator_normalized_dp_u64;
102
103   u32 reciprocal;
104   double reciprocal_dp;
105
106   u64 numerator_u64 = (denominator_normalized >> 10) |
107    ((u64)(62 + 1023) << 52);
108   *((u64 *)(&numerator_b)) = numerator_u64;
109
110   denominator_normalized_dp_u64 =
111    (u64)(denominator_normalized << 21) |
112    ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
113   *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
114
115   // Implement with a DP divide
116   reciprocal_dp = numerator / denominator_normalized_dp_b;
117   reciprocal = reciprocal_dp;
118
119   if(reciprocal == 0x80000001)
120     reciprocal = 0x80000000;
121
122   *_shift = 62 - shift;
123   return reciprocal;
124 }
125
126 double reciprocal_estimate(double a)
127 {
128   int q, s;
129   double r;
130
131   q = (int)(a * 512.0);
132   /* a in units of 1/512 rounded down */
133   r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
134   s = (int)(256.0 * r + 0.5);
135
136   /* r in units of 1/256 rounded to nearest */
137   
138   return (double)s / 256.0;
139 }
140
141 u32 reciprocal_estimate_u32(u32 value)
142 {
143   u64 dp_value_u64;
144   volatile double dp_value;
145   volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
146
147   if((value >> 31) == 0)
148     return 0xFFFFFFFF;
149
150   dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
151
152   *dp_value_ptr = dp_value_u64;
153
154   dp_value = reciprocal_estimate(dp_value);
155   dp_value_u64 = *dp_value_ptr;
156
157   return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
158 }
159
160 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
161 {
162   u32 shift = __builtin_clz(value);
163   u32 value_normalized = value << shift;
164
165   *_shift = 62 - shift;
166
167   value_normalized -= 2;
168
169   u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
170
171   u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
172   reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
173   temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
174   reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
175   temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
176   reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
177
178   return reciprocal_normalized;
179 }
180
181 #endif
182
183
184 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
185 {
186         return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
187 }
188
189 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
190 {
191   s32 coverage_x, coverage_y;
192
193   u32 mask_up_left;
194   u32 mask_down_right;
195
196   coverage_x = x2 >> 6;
197   coverage_y = y2 >> 8;
198
199   if(coverage_x < 0)
200     coverage_x = 0;
201
202   if(coverage_x > 31)
203     coverage_x = 31;
204
205   mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
206
207   if(coverage_y >= 1)
208     mask_down_right |= mask_down_right << 16;
209
210   coverage_x = x1 >> 6;
211
212   mask_up_left = 0xFFFF0000 << coverage_x;
213   if(coverage_x < 0)
214     mask_up_left = 0xFFFF0000;
215
216   coverage_y = y1 >> 8;
217   if(coverage_y <= 0)
218     mask_up_left |= mask_up_left >> 16;
219
220   return mask_up_left & mask_down_right;
221 }
222
223 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
224  u32 x2, u32 y2)
225 {
226   u32 mask = texture_region_mask(x1, y1, x2, y2);
227
228   psx_gpu->dirty_textures_4bpp_mask |= mask;
229   psx_gpu->dirty_textures_8bpp_mask |= mask;
230   psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
231
232   return mask;
233 }
234
235 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
236  u32 y1, u32 x2, u32 y2)
237 {
238   u32 mask = texture_region_mask(x1, y1, x2, y2) &
239    psx_gpu->viewport_mask;
240
241   psx_gpu->dirty_textures_4bpp_mask |= mask;
242   psx_gpu->dirty_textures_8bpp_mask |= mask;
243   psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
244
245   return mask;
246 }
247
248 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
249  u32 x2, u32 y2)
250 {
251   u32 mask = texture_region_mask(x1, y1, x2, y2);
252   u32 texture_page;
253   u8 *texture_page_ptr;
254   u16 *vram_ptr;
255   u32 texel_block;
256   u32 sub_x, sub_y;
257
258   psx_gpu->dirty_textures_8bpp_mask |= mask;
259   psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
260
261   if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
262       (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
263   {
264     texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
265     texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
266     texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
267     vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
268     sub_x = 4;
269     sub_y = 16;
270
271     while(sub_y)
272     {
273       while(sub_x)
274       {
275         texel_block = *vram_ptr;
276
277         texture_page_ptr[0] = texel_block & 0xF;
278         texture_page_ptr[1] = (texel_block >> 4) & 0xF;
279         texture_page_ptr[2] = (texel_block >> 8) & 0xF;
280         texture_page_ptr[3] = texel_block >> 12;
281         
282         vram_ptr++;
283         texture_page_ptr += 4;
284
285         sub_x--;          
286       }
287
288       vram_ptr -= 4;
289       sub_x = 4;
290
291       sub_y--;
292       vram_ptr += 1024;
293     }
294   }
295   else
296   {
297     psx_gpu->dirty_textures_4bpp_mask |= mask;
298   }
299 }
300
301 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
302  u32 texture_page);
303
304 #ifndef NEON_BUILD
305
306 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
307 {
308   u32 current_texture_page = psx_gpu->current_texture_page;
309   u8 *texture_page_ptr = psx_gpu->texture_page_base;
310   u16 *vram_ptr = psx_gpu->vram_ptr;
311
312   u32 texel_block;
313   u32 tile_x, tile_y;
314   u32 sub_x, sub_y;
315
316   vram_ptr += (current_texture_page >> 4) * 256 * 1024;
317   vram_ptr += (current_texture_page & 0xF) * 64;
318
319   texture_cache_loads++;
320
321   tile_y = 16;
322   tile_x = 16;
323   sub_x = 4;
324   sub_y = 16;
325
326   psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
327
328   while(tile_y)
329   {
330     while(tile_x)
331     {
332       while(sub_y)
333       {
334         while(sub_x)
335         {
336           texel_block = *vram_ptr;
337
338           texture_page_ptr[0] = texel_block & 0xF;
339           texture_page_ptr[1] = (texel_block >> 4) & 0xF;
340           texture_page_ptr[2] = (texel_block >> 8) & 0xF;
341           texture_page_ptr[3] = texel_block >> 12;
342           
343           vram_ptr++;
344           texture_page_ptr += 4;
345
346           sub_x--;          
347         }
348
349         vram_ptr -= 4;
350         sub_x = 4;
351
352         sub_y--;
353         vram_ptr += 1024;
354       }
355
356       sub_y = 16;
357
358       vram_ptr -= (1024 * 16) - 4;
359       tile_x--;
360     }
361
362     tile_x = 16;
363
364     vram_ptr += (16 * 1024) - (4 * 16);
365     tile_y--;
366   }
367 }
368
369 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
370  u32 texture_page)
371 {
372   u16 *texture_page_ptr = psx_gpu->texture_page_base;
373   u16 *vram_ptr = psx_gpu->vram_ptr;
374
375   u32 tile_x, tile_y;
376   u32 sub_y;
377
378   vec_8x16u texels;
379
380   texture_cache_loads++;
381
382   vram_ptr += (texture_page >> 4) * 256 * 1024;
383   vram_ptr += (texture_page & 0xF) * 64;
384
385   if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
386     texture_page_ptr += (8 * 16) * 8;
387
388   tile_x = 8;
389   tile_y = 16;
390
391   sub_y = 16;
392
393   while(tile_y)
394   {
395     while(tile_x)
396     {
397       while(sub_y)
398       {
399         load_128b(texels, vram_ptr);
400         store_128b(texels, texture_page_ptr);
401
402         texture_page_ptr += 8;
403         vram_ptr += 1024;
404
405         sub_y--;
406       }
407
408       sub_y = 16;
409
410       vram_ptr -= (1024 * 16);
411       vram_ptr += 8;
412
413       tile_x--;
414     }
415
416     tile_x = 8;
417
418     vram_ptr -= (8 * 8);
419     vram_ptr += (16 * 1024);
420
421     texture_page_ptr += (8 * 16) * 8;
422     tile_y--;
423   }
424 }
425
426 #endif
427
428
429 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
430 {
431   u32 current_texture_page = psx_gpu->current_texture_page;
432   u32 update_textures =
433    psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
434
435   psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
436
437   if(update_textures & (1 << current_texture_page))
438   {
439     update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
440     update_textures &= ~(1 << current_texture_page);
441   }
442
443   if(update_textures)
444   {
445     u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
446      (current_texture_page & 0x10);
447
448     update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
449   }
450 }
451
452 void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(
453  psx_gpu_struct *psx_gpu);
454
455 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
456 {
457   if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
458    (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
459   {
460     u32 num_blocks_dest = 0;
461     block_struct *block_src = psx_gpu->blocks; 
462     block_struct *block_dest = psx_gpu->blocks;
463
464     u16 *vram_ptr = psx_gpu->vram_ptr;
465     u32 i;
466
467     if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
468     {
469       for(i = 0; i < psx_gpu->num_blocks; i++)
470       {
471         u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
472         if(fb_offset & (1 << 11))
473         {
474           *block_dest = *block_src;
475           num_blocks_dest++;
476           block_dest++;
477         }
478         block_src++;
479       }
480     }
481     else
482     {
483       for(i = 0; i < psx_gpu->num_blocks; i++)
484       {
485         u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
486         if((fb_offset & (1 << 11)) == 0)
487         {
488           *block_dest = *block_src;
489           num_blocks_dest++;
490           block_dest++;
491         }
492         block_src++;
493       }
494     }
495
496     psx_gpu->num_blocks = num_blocks_dest;
497   }
498
499   if(psx_gpu->num_blocks)
500   {
501     render_block_handler_struct *render_block_handler =
502      psx_gpu->render_block_handler;
503
504     render_block_handler->texture_blocks(psx_gpu);
505     render_block_handler->shade_blocks(psx_gpu);
506     render_block_handler->blend_blocks(psx_gpu);
507
508 #ifdef PROFILE
509     span_pixel_blocks += psx_gpu->num_blocks;
510     render_buffer_flushes++;
511 #endif
512
513     psx_gpu->num_blocks = 0;
514   }
515 }
516
517
518 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
519  vertex_struct *b, vertex_struct *c);
520
521 #ifndef NEON_BUILD
522
523 #define setup_gradient_calculation_input(set, vertex)                          \
524   /* First type is:  uvrg bxxx xxxx                                          */\
525   /* Second type is: yyyy ybyy uvrg                                          */\
526   /* Since x_a and y_c are the same the same variable is used for both.      */\
527   x##set##_a_y##set##_c.e[0] = vertex->u;                                      \
528   x##set##_a_y##set##_c.e[1] = vertex->v;                                      \
529   x##set##_a_y##set##_c.e[2] = vertex->r;                                      \
530   x##set##_a_y##set##_c.e[3] = vertex->g;                                      \
531   dup_4x16b(x##set##_b, vertex->x);                                            \
532   dup_4x16b(x##set##_c, vertex->x);                                            \
533   dup_4x16b(y##set##_a, vertex->y);                                            \
534   dup_4x16b(y##set##_b, vertex->y);                                            \
535   x##set##_b.e[0] = vertex->b;                                                 \
536   y##set##_b.e[1] = vertex->b                                                  \
537   
538
539 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
540  vertex_struct *b, vertex_struct *c)
541 {
542   u32 triangle_area = psx_gpu->triangle_area;
543   u32 winding_mask_scalar;
544
545   u32 triangle_area_shift;
546   u64 triangle_area_reciprocal =
547    fixed_reciprocal(triangle_area, &triangle_area_shift);
548   triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
549
550   // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
551   // ( d0       *  d1      ) - ( d2       *  d3      ) =
552   // ( m0                  ) - ( m1                  ) = gradient
553
554   // This is split to do 12 elements at a time over three sets: a, b, and c.
555   // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
556   // two of the slots are unused.
557
558   // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
559   // is g.
560
561   vec_4x16s x0_a_y0_c, x0_b, x0_c;
562   vec_4x16s y0_a, y0_b;
563   vec_4x16s x1_a_y1_c, x1_b, x1_c;
564   vec_4x16s y1_a, y1_b;
565   vec_4x16s x2_a_y2_c, x2_b, x2_c;
566   vec_4x16s y2_a, y2_b;
567
568   vec_4x32u uvrg_base;
569   vec_4x32u b_base;
570   vec_4x32u uvrgb_phase;
571
572   vec_4x16s d0_a_d3_c, d0_b, d0_c;
573   vec_4x16s d1_a, d1_b, d1_c_d2_a;
574   vec_4x16s d2_b, d2_c;
575   vec_4x16s d3_a, d3_b;
576
577   vec_4x32s m0_a, m0_b, m0_c;
578   vec_4x32s m1_a, m1_b, m1_c;
579
580   vec_4x32u gradient_area_a, gradient_area_c;
581   vec_2x32u gradient_area_b;  
582
583   vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
584   vec_2x32u gradient_area_sign_b;
585   vec_4x32u winding_mask;
586
587   vec_2x64u gradient_wide_a0, gradient_wide_a1;
588   vec_2x64u gradient_wide_c0, gradient_wide_c1;
589   vec_2x64u gradient_wide_b;
590
591   vec_4x32u gradient_a, gradient_c;
592   vec_2x32u gradient_b;
593   vec_16x8s gradient_shift;
594
595   setup_gradient_calculation_input(0, a);
596   setup_gradient_calculation_input(1, b);
597   setup_gradient_calculation_input(2, c);
598
599   dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
600   shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
601   shl_long_4x16b(b_base, x0_b, 16);
602
603   add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
604   add_4x32b(b_base, b_base, uvrgb_phase);
605
606   // Can probably pair these, but it'll require careful register allocation
607   sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
608   sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
609
610   sub_4x16b(d0_b, x1_b, x0_b);
611   sub_4x16b(d0_c, x1_c, x0_c);
612
613   sub_4x16b(d1_a, y2_a, y1_a);
614   sub_4x16b(d1_b, y2_b, y1_b);
615
616   sub_4x16b(d2_b, x2_b, x1_b);
617   sub_4x16b(d2_c, x2_c, x1_c);
618
619   sub_4x16b(d3_a, y1_a, y0_a);
620   sub_4x16b(d3_b, y1_b, y0_b);
621
622   mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
623   mul_long_4x16b(m0_b, d0_b, d1_b);
624   mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
625
626   mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
627   mul_long_4x16b(m1_b, d2_b, d3_b);
628   mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
629
630   sub_4x32b(gradient_area_a, m0_a, m1_a);
631   sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
632   sub_4x32b(gradient_area_c, m0_c, m1_c);
633
634   cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
635   cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
636   cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
637
638   abs_4x32b(gradient_area_a, gradient_area_a);
639   abs_2x32b(gradient_area_b, gradient_area_b);
640   abs_4x32b(gradient_area_c, gradient_area_c);
641
642   winding_mask_scalar = -psx_gpu->triangle_winding;
643
644   dup_4x32b(winding_mask, winding_mask_scalar);
645   eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
646   eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
647   eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
648
649   mul_scalar_long_2x32b(gradient_wide_a0, 
650    vector_cast(vec_2x32s, gradient_area_a.low), 
651    (s64)triangle_area_reciprocal);
652   mul_scalar_long_2x32b(gradient_wide_a1,
653    vector_cast(vec_2x32s, gradient_area_a.high),
654    (s64)triangle_area_reciprocal);
655   mul_scalar_long_2x32b(gradient_wide_b, 
656    vector_cast(vec_2x32s, gradient_area_b),
657    (s64)triangle_area_reciprocal);
658   mul_scalar_long_2x32b(gradient_wide_c0, 
659    vector_cast(vec_2x32s, gradient_area_c.low),
660    (s64)triangle_area_reciprocal);
661   mul_scalar_long_2x32b(gradient_wide_c1, 
662    vector_cast(vec_2x32s, gradient_area_c.high),
663    (s64)triangle_area_reciprocal);
664
665   dup_16x8b(gradient_shift, triangle_area_shift);
666   shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
667    vector_cast(vec_2x64u, gradient_shift));
668   shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
669    vector_cast(vec_2x64u, gradient_shift));
670   shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
671    vector_cast(vec_2x64u, gradient_shift));
672   shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
673    vector_cast(vec_2x64u, gradient_shift));
674   shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
675    vector_cast(vec_2x64u, gradient_shift));
676
677   mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
678   mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
679   mov_narrow_2x64b(gradient_b, gradient_wide_b);
680   mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
681   mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
682
683   shl_4x32b(gradient_a, gradient_a, 4);
684   shl_2x32b(gradient_b, gradient_b, 4);
685   shl_4x32b(gradient_c, gradient_c, 4);
686
687   eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
688   eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
689   eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
690
691   sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
692   sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
693   sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
694
695   u32 left_adjust = a->x;
696   mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
697   mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
698
699   vec_4x32u uvrg_dx2;
700   vec_2x32u b_dx2;
701
702   vec_4x32u uvrg_dx3;
703   vec_2x32u b_dx3;
704
705   vec_4x32u zero;
706
707   eor_4x32b(zero, zero, zero);
708   add_4x32b(uvrg_dx2, gradient_a, gradient_a);
709   add_2x32b(b_dx2, gradient_b, gradient_b);
710   add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
711   add_2x32b(b_dx3, gradient_b, b_dx2);
712
713   // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
714   // lined up properly
715   psx_gpu->u_block_span.e[0] = zero.e[0];
716   psx_gpu->u_block_span.e[1] = gradient_a.e[0];
717   psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
718   psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
719
720   psx_gpu->v_block_span.e[0] = zero.e[1];
721   psx_gpu->v_block_span.e[1] = gradient_a.e[1];
722   psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
723   psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
724
725   psx_gpu->r_block_span.e[0] = zero.e[2];
726   psx_gpu->r_block_span.e[1] = gradient_a.e[2];
727   psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
728   psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
729
730   psx_gpu->g_block_span.e[0] = zero.e[3];
731   psx_gpu->g_block_span.e[1] = gradient_a.e[3];
732   psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
733   psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
734
735   psx_gpu->b_block_span.e[0] = zero.e[0];
736   psx_gpu->b_block_span.e[1] = gradient_b.e[0];
737   psx_gpu->b_block_span.e[2] = b_dx2.e[0];
738   psx_gpu->b_block_span.e[3] = b_dx3.e[0];
739
740   psx_gpu->uvrg = uvrg_base;
741   psx_gpu->b = b_base.e[0];
742
743   psx_gpu->uvrg_dx = gradient_a;
744   psx_gpu->uvrg_dy = gradient_c;
745   psx_gpu->b_dy = gradient_b.e[1];
746 }
747 #endif
748
749 #define vector_check(_a, _b)                                                   \
750   if(memcmp(&_a, &_b, sizeof(_b)))                                             \
751   {                                                                            \
752     if(sizeof(_b) == 8)                                                        \
753     {                                                                          \
754       printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n",                     \
755        #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]);                          \
756     }                                                                          \
757     else                                                                       \
758     {                                                                          \
759       printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n",         \
760        #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1],         \
761        _b.e[2], _b.e[3]);                                                      \
762     }                                                                          \
763   }                                                                            \
764
765 #define scalar_check(_a, _b)                                                   \
766   if(_a != _b)                                                                 \
767     printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b)                  \
768
769
770 #ifndef NDEBUG
771 #define setup_spans_debug_check(span_edge_data_element)                        \
772 {                                                                              \
773   u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data;          \
774   if (_num_spans > MAX_SPANS)                                                  \
775     *(int *)0 = 1;                                                             \
776   if (_num_spans < psx_gpu->num_spans)                                         \
777   {                                                                            \
778     if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW)                 \
779       *(int *)0 = 1;                                                           \
780     if(span_edge_data_element.y > 2048)                                        \
781       *(int *)0 = 1;                                                           \
782   }                                                                            \
783 }                                                                              \
784
785 #else
786 #define setup_spans_debug_check(span_edge_data_element)                        \
787
788 #endif
789
790 #define setup_spans_prologue_alternate_yes()                                   \
791   vec_2x64s alternate_x;                                                       \
792   vec_2x64s alternate_dx_dy;                                                   \
793   vec_4x32s alternate_x_32;                                                    \
794   vec_2x32s alternate_x_16;                                                    \
795                                                                                \
796   vec_4x16u alternate_select;                                                  \
797   vec_4x16s y_mid_point;                                                       \
798                                                                                \
799   s32 y_b = v_b->y;                                                            \
800   s64 edge_alt;                                                                \
801   s32 edge_dx_dy_alt;                                                          \
802   u32 edge_shift_alt                                                           \
803
804 #define setup_spans_prologue_alternate_no()                                    \
805
806 #define setup_spans_prologue(alternate_active)                                 \
807   edge_data_struct *span_edge_data;                                            \
808   vec_4x32u *span_uvrg_offset;                                                 \
809   u32 *span_b_offset;                                                          \
810                                                                                \
811   s32 clip;                                                                    \
812                                                                                \
813   vec_2x64s edges_xy;                                                          \
814   vec_2x32s edges_dx_dy;                                                       \
815   vec_2x32u edge_shifts;                                                       \
816                                                                                \
817   vec_2x64s left_x, right_x;                                                   \
818   vec_2x64s left_dx_dy, right_dx_dy;                                           \
819   vec_4x32s left_x_32, right_x_32;                                             \
820   vec_8x16s left_right_x_16;                                                   \
821   vec_4x16s y_x4;                                                              \
822   vec_8x16s left_edge;                                                         \
823   vec_8x16s right_edge;                                                        \
824   vec_4x16u span_shift;                                                        \
825                                                                                \
826   vec_2x32u c_0x01;                                                            \
827   vec_4x16u c_0x04;                                                            \
828   vec_4x16u c_0xFFFE;                                                          \
829   vec_4x16u c_0x07;                                                            \
830                                                                                \
831   vec_2x32s x_starts;                                                          \
832   vec_2x32s x_ends;                                                            \
833                                                                                \
834   s32 x_a = v_a->x;                                                            \
835   s32 x_b = v_b->x;                                                            \
836   s32 x_c = v_c->x;                                                            \
837   s32 y_a = v_a->y;                                                            \
838   s32 y_c = v_c->y;                                                            \
839                                                                                \
840   vec_4x32u uvrg = psx_gpu->uvrg;                                              \
841   vec_4x32u uvrg_dy = psx_gpu->uvrg_dy;                                        \
842   u32 b = psx_gpu->b;                                                          \
843   u32 b_dy = psx_gpu->b_dy;                                                    \
844                                                                                \
845   dup_2x32b(c_0x01, 0x01);                                                     \
846   setup_spans_prologue_alternate_##alternate_active()                          \
847
848 #define setup_spans_prologue_b()                                               \
849   span_edge_data = psx_gpu->span_edge_data;                                    \
850   span_uvrg_offset = psx_gpu->span_uvrg_offset;                                \
851   span_b_offset = psx_gpu->span_b_offset;                                      \
852                                                                                \
853   vec_8x16u c_0x0001;                                                          \
854                                                                                \
855   dup_8x16b(c_0x0001, 0x0001);                                                 \
856   dup_8x16b(left_edge, psx_gpu->viewport_start_x);                             \
857   dup_8x16b(right_edge, psx_gpu->viewport_end_x);                              \
858   add_8x16b(right_edge, right_edge, c_0x0001);                                 \
859   dup_4x16b(c_0x04, 0x04);                                                     \
860   dup_4x16b(c_0x07, 0x07);                                                     \
861   dup_4x16b(c_0xFFFE, 0xFFFE);                                                 \
862
863
864 #define compute_edge_delta_x2()                                                \
865 {                                                                              \
866   vec_2x32s heights;                                                           \
867   vec_2x32s height_reciprocals;                                                \
868   vec_2x32s heights_b;                                                         \
869   vec_4x32u widths;                                                            \
870                                                                                \
871   u32 edge_shift = reciprocal_table[height];                                   \
872                                                                                \
873   dup_2x32b(heights, height);                                                  \
874   sub_2x32b(widths, x_ends, x_starts);                                         \
875                                                                                \
876   dup_2x32b(edge_shifts, edge_shift);                                          \
877   sub_2x32b(heights_b, heights, c_0x01);                                       \
878   shr_2x32b(height_reciprocals, edge_shifts, 10);                              \
879                                                                                \
880   mla_2x32b(heights_b, x_starts, heights);                                     \
881   bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
882   mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
883   mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
884 }                                                                              \
885
886 #define compute_edge_delta_x3(start_c, height_a, height_b)                     \
887 {                                                                              \
888   vec_2x32s heights;                                                           \
889   vec_2x32s height_reciprocals;                                                \
890   vec_2x32s heights_b;                                                         \
891   vec_2x32u widths;                                                            \
892                                                                                \
893   u32 width_alt;                                                               \
894   s32 height_b_alt;                                                            \
895   u32 height_reciprocal_alt;                                                   \
896                                                                                \
897   heights.e[0] = height_a;                                                     \
898   heights.e[1] = height_b;                                                     \
899                                                                                \
900   edge_shifts.e[0] = reciprocal_table[height_a];                               \
901   edge_shifts.e[1] = reciprocal_table[height_b];                               \
902   edge_shift_alt = reciprocal_table[height_minor_b];                           \
903                                                                                \
904   sub_2x32b(widths, x_ends, x_starts);                                         \
905   width_alt = x_c - start_c;                                                   \
906                                                                                \
907   shr_2x32b(height_reciprocals, edge_shifts, 10);                              \
908   height_reciprocal_alt = edge_shift_alt >> 10;                                \
909                                                                                \
910   bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
911   edge_shift_alt &= 0x1F;                                                      \
912                                                                                \
913   sub_2x32b(heights_b, heights, c_0x01);                                       \
914   height_b_alt = height_minor_b - 1;                                           \
915                                                                                \
916   mla_2x32b(heights_b, x_starts, heights);                                     \
917   height_b_alt += height_minor_b * start_c;                                    \
918                                                                                \
919   mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
920   edge_alt = (s64)height_b_alt * height_reciprocal_alt;                        \
921                                                                                \
922   mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
923   edge_dx_dy_alt = width_alt * height_reciprocal_alt;                          \
924 }                                                                              \
925
926
927 #define setup_spans_adjust_y_up()                                              \
928   sub_4x32b(y_x4, y_x4, c_0x04)                                                \
929
930 #define setup_spans_adjust_y_down()                                            \
931   add_4x32b(y_x4, y_x4, c_0x04)                                                \
932
933 #define setup_spans_adjust_interpolants_up()                                   \
934   sub_4x32b(uvrg, uvrg, uvrg_dy);                                              \
935   b -= b_dy                                                                    \
936
937 #define setup_spans_adjust_interpolants_down()                                 \
938   add_4x32b(uvrg, uvrg, uvrg_dy);                                              \
939   b += b_dy                                                                    \
940
941
942 #define setup_spans_clip_interpolants_increment()                              \
943   mla_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
944   b += b_dy * clip                                                             \
945
946 #define setup_spans_clip_interpolants_decrement()                              \
947   mls_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
948   b -= b_dy * clip                                                             \
949
950 #define setup_spans_clip_alternate_yes()                                       \
951   edge_alt += edge_dx_dy_alt * (s64)(clip)                                     \
952
953 #define setup_spans_clip_alternate_no()                                        \
954
955 #define setup_spans_clip(direction, alternate_active)                          \
956 {                                                                              \
957   clipped_triangles++;                                                         \
958   mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip);                     \
959   setup_spans_clip_alternate_##alternate_active();                             \
960   setup_spans_clip_interpolants_##direction();                                 \
961 }                                                                              \
962
963
964 #define setup_spans_adjust_edges_alternate_no(left_index, right_index)         \
965 {                                                                              \
966   vec_2x64u edge_shifts_64;                                                    \
967   vec_2x64s edges_dx_dy_64;                                                    \
968                                                                                \
969   mov_wide_2x32b(edge_shifts_64, edge_shifts);                                 \
970   shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64);                      \
971                                                                                \
972   mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy);                                 \
973   shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64);          \
974                                                                                \
975   left_x.e[0] = edges_xy.e[left_index];                                        \
976   right_x.e[0] = edges_xy.e[right_index];                                      \
977                                                                                \
978   left_dx_dy.e[0] = edges_dx_dy_64.e[left_index];                              \
979   left_dx_dy.e[1] = edges_dx_dy_64.e[left_index];                              \
980   right_dx_dy.e[0] = edges_dx_dy_64.e[right_index];                            \
981   right_dx_dy.e[1] = edges_dx_dy_64.e[right_index];                            \
982                                                                                \
983   add_1x64b(left_x.high, left_x.low, left_dx_dy.low);                          \
984   add_1x64b(right_x.high, right_x.low, right_dx_dy.low);                       \
985                                                                                \
986   add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy);                               \
987   add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy);                            \
988 }                                                                              \
989
990 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index)        \
991 {                                                                              \
992   setup_spans_adjust_edges_alternate_no(left_index, right_index);              \
993   s64 edge_dx_dy_alt_64;                                                       \
994                                                                                \
995   dup_4x16b(y_mid_point, y_b);                                                 \
996                                                                                \
997   edge_alt <<= edge_shift_alt;                                                 \
998   edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt;                   \
999                                                                                \
1000   alternate_x.e[0] = edge_alt;                                                 \
1001   alternate_dx_dy.e[0] = edge_dx_dy_alt_64;                                    \
1002   alternate_dx_dy.e[1] = edge_dx_dy_alt_64;                                    \
1003                                                                                \
1004   add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low);           \
1005   add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy);                \
1006 }                                                                              \
1007
1008
1009 #define setup_spans_y_select_up()                                              \
1010   cmplt_4x16b(alternate_select, y_x4, y_mid_point)                             \
1011
1012 #define setup_spans_y_select_down()                                            \
1013   cmpgt_4x16b(alternate_select, y_x4, y_mid_point)                             \
1014
1015 #define setup_spans_y_select_alternate_yes(direction)                          \
1016   setup_spans_y_select_##direction()                                           \
1017
1018 #define setup_spans_y_select_alternate_no(direction)                           \
1019
1020 #define setup_spans_alternate_select_left()                                    \
1021   bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select)             \
1022
1023 #define setup_spans_alternate_select_right()                                   \
1024   bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select)            \
1025
1026 #define setup_spans_alternate_select_none()                                    \
1027
1028 #define setup_spans_increment_alternate_yes()                                  \
1029   shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32);                       \
1030   add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
1031   shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32);                      \
1032   add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
1033   mov_narrow_4x32b(alternate_x_16, alternate_x_32)                             \
1034
1035 #define setup_spans_increment_alternate_no()                                   \
1036
1037 #define setup_spans_set_x4(alternate, direction, alternate_active)             \
1038 {                                                                              \
1039   span_uvrg_offset[0] = uvrg;                                                  \
1040   span_b_offset[0] = b;                                                        \
1041   setup_spans_adjust_interpolants_##direction();                               \
1042                                                                                \
1043   span_uvrg_offset[1] = uvrg;                                                  \
1044   span_b_offset[1] = b;                                                        \
1045   setup_spans_adjust_interpolants_##direction();                               \
1046                                                                                \
1047   span_uvrg_offset[2] = uvrg;                                                  \
1048   span_b_offset[2] = b;                                                        \
1049   setup_spans_adjust_interpolants_##direction();                               \
1050                                                                                \
1051   span_uvrg_offset[3] = uvrg;                                                  \
1052   span_b_offset[3] = b;                                                        \
1053   setup_spans_adjust_interpolants_##direction();                               \
1054                                                                                \
1055   span_uvrg_offset += 4;                                                       \
1056   span_b_offset += 4;                                                          \
1057                                                                                \
1058   shr_narrow_2x64b(left_x_32.low, left_x, 32);                                 \
1059   shr_narrow_2x64b(right_x_32.low, right_x, 32);                               \
1060                                                                                \
1061   add_2x64b(left_x, left_x, left_dx_dy);                                       \
1062   add_2x64b(right_x, right_x, right_dx_dy);                                    \
1063                                                                                \
1064   shr_narrow_2x64b(left_x_32.high, left_x, 32);                                \
1065   shr_narrow_2x64b(right_x_32.high, right_x, 32);                              \
1066                                                                                \
1067   add_2x64b(left_x, left_x, left_dx_dy);                                       \
1068   add_2x64b(right_x, right_x, right_dx_dy);                                    \
1069                                                                                \
1070   mov_narrow_4x32b(left_right_x_16.low, left_x_32);                            \
1071   mov_narrow_4x32b(left_right_x_16.high, right_x_32);                          \
1072                                                                                \
1073   setup_spans_increment_alternate_##alternate_active();                        \
1074   setup_spans_y_select_alternate_##alternate_active(direction);                \
1075   setup_spans_alternate_select_##alternate();                                  \
1076                                                                                \
1077   max_8x16b(left_right_x_16, left_right_x_16, left_edge);                      \
1078   min_8x16b(left_right_x_16, left_right_x_16, right_edge);                     \
1079                                                                                \
1080   sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low);  \
1081   add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07);               \
1082   and_4x16b(span_shift, left_right_x_16.high, c_0x07);                         \
1083   shl_variable_4x16b(span_shift, c_0xFFFE, span_shift);                        \
1084   shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3);                    \
1085                                                                                \
1086   u32 i;                                                                       \
1087   for(i = 0; i < 4; i++)                                                       \
1088   {                                                                            \
1089     span_edge_data[i].left_x = left_right_x_16.low.e[i];                       \
1090     span_edge_data[i].num_blocks = left_right_x_16.high.e[i];                  \
1091     span_edge_data[i].right_mask = span_shift.e[i];                            \
1092     span_edge_data[i].y = y_x4.e[i];                                           \
1093     setup_spans_debug_check(span_edge_data[i]);                                \
1094   }                                                                            \
1095                                                                                \
1096   span_edge_data += 4;                                                         \
1097                                                                                \
1098   setup_spans_adjust_y_##direction();                                          \
1099 }                                                                              \
1100
1101
1102 #define setup_spans_alternate_adjust_yes()                                     \
1103   edge_alt -= edge_dx_dy_alt * (s64)height_minor_a                             \
1104
1105 #define setup_spans_alternate_adjust_no()                                      \
1106
1107
1108 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1109   setup_spans_alternate_adjust_##alternate_active();                           \
1110   if(y_c > psx_gpu->viewport_end_y)                                            \
1111     height -= y_c - psx_gpu->viewport_end_y - 1;                               \
1112                                                                                \
1113   clip = psx_gpu->viewport_start_y - y_a;                                      \
1114   if(clip > 0)                                                                 \
1115   {                                                                            \
1116     height -= clip;                                                            \
1117     y_a += clip;                                                               \
1118     setup_spans_clip(increment, alternate_active);                             \
1119   }                                                                            \
1120                                                                                \
1121   setup_spans_prologue_b();                                                    \
1122                                                                                \
1123   if(height > 0)                                                               \
1124   {                                                                            \
1125     y_x4.e[0] = y_a;                                                           \
1126     y_x4.e[1] = y_a + 1;                                                       \
1127     y_x4.e[2] = y_a + 2;                                                       \
1128     y_x4.e[3] = y_a + 3;                                                       \
1129     setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
1130      right_index);                                                             \
1131                                                                                \
1132     psx_gpu->num_spans = height;                                               \
1133     do                                                                         \
1134     {                                                                          \
1135       setup_spans_set_x4(alternate, down, alternate_active);                   \
1136       height -= 4;                                                             \
1137     } while(height > 0);                                                       \
1138   }                                                                            \
1139
1140
1141 #define setup_spans_alternate_pre_increment_yes()                              \
1142   edge_alt += edge_dx_dy_alt                                                   \
1143
1144 #define setup_spans_alternate_pre_increment_no()                               \
1145
1146 #define setup_spans_up_decrement_height_yes()                                  \
1147   height--                                                                     \
1148
1149 #define setup_spans_up_decrement_height_no()                                   \
1150   {}                                                                           \
1151
1152 #define setup_spans_up(left_index, right_index, alternate, alternate_active)   \
1153   setup_spans_alternate_adjust_##alternate_active();                           \
1154   y_a--;                                                                       \
1155                                                                                \
1156   if(y_c < psx_gpu->viewport_start_y)                                          \
1157     height -= psx_gpu->viewport_start_y - y_c;                                 \
1158   else                                                                         \
1159     setup_spans_up_decrement_height_##alternate_active();                      \
1160                                                                                \
1161   clip = y_a - psx_gpu->viewport_end_y;                                        \
1162   if(clip > 0)                                                                 \
1163   {                                                                            \
1164     height -= clip;                                                            \
1165     y_a -= clip;                                                               \
1166     setup_spans_clip(decrement, alternate_active);                             \
1167   }                                                                            \
1168                                                                                \
1169   setup_spans_prologue_b();                                                    \
1170                                                                                \
1171   if(height > 0)                                                               \
1172   {                                                                            \
1173     y_x4.e[0] = y_a;                                                           \
1174     y_x4.e[1] = y_a - 1;                                                       \
1175     y_x4.e[2] = y_a - 2;                                                       \
1176     y_x4.e[3] = y_a - 3;                                                       \
1177     add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);                           \
1178     setup_spans_alternate_pre_increment_##alternate_active();                  \
1179     setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
1180      right_index);                                                             \
1181     setup_spans_adjust_interpolants_up();                                      \
1182                                                                                \
1183     psx_gpu->num_spans = height;                                               \
1184     while(height > 0)                                                          \
1185     {                                                                          \
1186       setup_spans_set_x4(alternate, up, alternate_active);                     \
1187       height -= 4;                                                             \
1188     }                                                                          \
1189   }                                                                            \
1190
1191 #define index_left  0
1192 #define index_right 1
1193
1194 #define setup_spans_up_up(minor, major)                                        \
1195   setup_spans_prologue(yes);                                                   \
1196   s32 height_minor_a = y_a - y_b;                                              \
1197   s32 height_minor_b = y_b - y_c;                                              \
1198   s32 height = y_a - y_c;                                                      \
1199                                                                                \
1200   dup_2x32b(x_starts, x_a);                                                    \
1201   x_ends.e[0] = x_c;                                                           \
1202   x_ends.e[1] = x_b;                                                           \
1203                                                                                \
1204   compute_edge_delta_x3(x_b, height, height_minor_a);                          \
1205   setup_spans_up(index_##major, index_##minor, minor, yes)                     \
1206
1207
1208 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1209  vertex_struct *v_b, vertex_struct *v_c);
1210 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1211  vertex_struct *v_b, vertex_struct *v_c);
1212 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1213  vertex_struct *v_b, vertex_struct *v_c);
1214 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1215  vertex_struct *v_b, vertex_struct *v_c);
1216 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1217  vertex_struct *v_b, vertex_struct *v_c);
1218 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1219  vertex_struct *v_b, vertex_struct *v_c);
1220 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1221  vertex_struct *v_b, vertex_struct *v_c);
1222 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1223  vertex_struct *v_b, vertex_struct *v_c);
1224 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1225  vertex_struct *v_b, vertex_struct *v_c);
1226
1227
1228 #ifndef NEON_BUILD
1229
1230 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1231  vertex_struct *v_b, vertex_struct *v_c)
1232 {
1233   setup_spans_up_up(left, right);
1234 }
1235
1236 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1237  vertex_struct *v_b, vertex_struct *v_c)
1238 {
1239   setup_spans_up_up(right, left);
1240 }
1241
1242 #define setup_spans_down_down(minor, major)                                    \
1243   setup_spans_prologue(yes);                                                   \
1244   s32 height_minor_a = y_b - y_a;                                              \
1245   s32 height_minor_b = y_c - y_b;                                              \
1246   s32 height = y_c - y_a;                                                      \
1247                                                                                \
1248   dup_2x32b(x_starts, x_a);                                                    \
1249   x_ends.e[0] = x_c;                                                           \
1250   x_ends.e[1] = x_b;                                                           \
1251                                                                                \
1252   compute_edge_delta_x3(x_b, height, height_minor_a);                          \
1253   setup_spans_down(index_##major, index_##minor, minor, yes)                   \
1254
1255 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1256  vertex_struct *v_b, vertex_struct *v_c)
1257 {
1258   setup_spans_down_down(left, right);
1259 }
1260
1261 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1262  vertex_struct *v_b, vertex_struct *v_c)
1263 {
1264   setup_spans_down_down(right, left);
1265 }
1266
1267 #define setup_spans_up_flat()                                                  \
1268   s32 height = y_a - y_c;                                                      \
1269                                                                                \
1270   flat_triangles++;                                                            \
1271   compute_edge_delta_x2();                                                     \
1272   setup_spans_up(index_left, index_right, none, no)                            \
1273
1274 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1275  vertex_struct *v_b, vertex_struct *v_c)
1276 {
1277   setup_spans_prologue(no);
1278   x_starts.e[0] = x_a;
1279   x_starts.e[1] = x_b;
1280   dup_2x32b(x_ends, x_c);
1281
1282   setup_spans_up_flat();
1283 }
1284
1285 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1286  vertex_struct *v_b, vertex_struct *v_c)
1287 {
1288   setup_spans_prologue(no);
1289   dup_2x32b(x_starts, x_a);
1290   x_ends.e[0] = x_b;
1291   x_ends.e[1] = x_c;
1292
1293   setup_spans_up_flat();
1294 }
1295
1296 #define setup_spans_down_flat()                                                \
1297   s32 height = y_c - y_a;                                                      \
1298                                                                                \
1299   flat_triangles++;                                                            \
1300   compute_edge_delta_x2();                                                     \
1301   setup_spans_down(index_left, index_right, none, no)                          \
1302
1303 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1304  vertex_struct *v_b, vertex_struct *v_c)
1305 {
1306   setup_spans_prologue(no);
1307   x_starts.e[0] = x_a;
1308   x_starts.e[1] = x_b;
1309   dup_2x32b(x_ends, x_c);
1310
1311   setup_spans_down_flat();
1312 }
1313
1314 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1315  vertex_struct *v_b, vertex_struct *v_c)
1316 {
1317   setup_spans_prologue(no);
1318   dup_2x32b(x_starts, x_a);
1319   x_ends.e[0] = x_b;
1320   x_ends.e[1] = x_c;
1321
1322   setup_spans_down_flat();
1323 }
1324
1325 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1326  vertex_struct *v_b, vertex_struct *v_c)
1327 {
1328   setup_spans_prologue(no);
1329
1330   s32 y_b = v_b->y;
1331   s64 edge_alt;
1332   s32 edge_dx_dy_alt;
1333   u32 edge_shift_alt;
1334
1335   s32 middle_y = y_a;
1336   s32 height_minor_a = y_a - y_b;
1337   s32 height_minor_b = y_c - y_a;
1338   s32 height_major = y_c - y_b;
1339
1340   vec_2x64s edges_xy_b;
1341   vec_2x32s edges_dx_dy_b;
1342   vec_2x32u edge_shifts_b;
1343
1344   vec_2x32s height_increment;
1345
1346   x_starts.e[0] = x_a;
1347   x_starts.e[1] = x_c;
1348   dup_2x32b(x_ends, x_b);
1349
1350   compute_edge_delta_x3(x_a, height_minor_a, height_major);
1351
1352   height_increment.e[0] = 0;
1353   height_increment.e[1] = height_minor_b;
1354
1355   mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1356
1357   edges_xy_b.e[0] = edge_alt;
1358   edges_xy_b.e[1] = edges_xy.e[1];
1359
1360   edge_shifts_b = edge_shifts;
1361   edge_shifts_b.e[0] = edge_shift_alt;
1362
1363   neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1364   edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1365   
1366   y_a--;
1367
1368   if(y_b < psx_gpu->viewport_start_y)
1369     height_minor_a -= psx_gpu->viewport_start_y - y_b;
1370
1371   clip = y_a - psx_gpu->viewport_end_y;
1372   if(clip > 0)
1373   {
1374     height_minor_a -= clip;
1375     y_a -= clip;
1376     setup_spans_clip(decrement, no);
1377   }
1378
1379   setup_spans_prologue_b();
1380
1381   if(height_minor_a > 0)
1382   {
1383     y_x4.e[0] = y_a;
1384     y_x4.e[1] = y_a - 1;
1385     y_x4.e[2] = y_a - 2;
1386     y_x4.e[3] = y_a - 3;
1387     add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1388     setup_spans_adjust_edges_alternate_no(index_left, index_right);
1389     setup_spans_adjust_interpolants_up();
1390
1391     psx_gpu->num_spans = height_minor_a;
1392     while(height_minor_a > 0)
1393     {
1394       setup_spans_set_x4(none, up, no);
1395       height_minor_a -= 4;
1396     }
1397
1398     span_edge_data += height_minor_a;
1399     span_uvrg_offset += height_minor_a;
1400     span_b_offset += height_minor_a;
1401   }
1402   
1403   edges_xy = edges_xy_b;
1404   edges_dx_dy = edges_dx_dy_b;
1405   edge_shifts = edge_shifts_b;
1406
1407   uvrg = psx_gpu->uvrg;
1408   b = psx_gpu->b;
1409
1410   y_a = middle_y;
1411
1412   if(y_c > psx_gpu->viewport_end_y)
1413     height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1414
1415   clip = psx_gpu->viewport_start_y - y_a;
1416   if(clip > 0)
1417   {
1418     height_minor_b -= clip;
1419     y_a += clip;
1420     setup_spans_clip(increment, no);
1421   }
1422
1423   if(height_minor_b > 0)
1424   {
1425     y_x4.e[0] = y_a;
1426     y_x4.e[1] = y_a + 1;
1427     y_x4.e[2] = y_a + 2;
1428     y_x4.e[3] = y_a + 3;
1429     setup_spans_adjust_edges_alternate_no(index_left, index_right);
1430
1431     // FIXME: overflow corner case
1432     if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1433       height_minor_b &= ~3;
1434
1435     psx_gpu->num_spans += height_minor_b;
1436     while(height_minor_b > 0)
1437     {
1438       setup_spans_set_x4(none, down, no);
1439       height_minor_b -= 4;
1440     }
1441   }
1442
1443   left_split_triangles++;
1444 }
1445
1446 #endif
1447
1448
1449 #define dither_table_entry_normal(value)                                       \
1450   (value)                                                                      \
1451
1452
1453 #define setup_blocks_load_msb_mask_indirect()                                  \
1454
1455 #define setup_blocks_load_msb_mask_direct()                                    \
1456   vec_8x16u msb_mask;                                                          \
1457   dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
1458
1459
1460 #define setup_blocks_variables_shaded_textured(target)                         \
1461   vec_4x32u u_block;                                                           \
1462   vec_4x32u v_block;                                                           \
1463   vec_4x32u r_block;                                                           \
1464   vec_4x32u g_block;                                                           \
1465   vec_4x32u b_block;                                                           \
1466   vec_4x32u uvrg_dx = psx_gpu->uvrg_dx;                                        \
1467   vec_4x32u uvrg_dx4;                                                          \
1468   vec_4x32u uvrg_dx8;                                                          \
1469   vec_4x32u uvrg;                                                              \
1470   u32 b_dx = psx_gpu->b_block_span.e[1];                                       \
1471   u32 b_dx4 = b_dx << 2;                                                       \
1472   u32 b_dx8 = b_dx << 3;                                                       \
1473   u32 b;                                                                       \
1474                                                                                \
1475   vec_16x8u texture_mask;                                                      \
1476   shl_4x32b(uvrg_dx4, uvrg_dx, 2);                                             \
1477   shl_4x32b(uvrg_dx8, uvrg_dx, 3);                                             \
1478   dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
1479   dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
1480
1481 #define setup_blocks_variables_shaded_untextured(target)                       \
1482   vec_4x32u r_block;                                                           \
1483   vec_4x32u g_block;                                                           \
1484   vec_4x32u b_block;                                                           \
1485   vec_4x32u rgb_dx;                                                            \
1486   vec_4x32u rgb_dx4;                                                           \
1487   vec_4x32u rgb_dx8;                                                           \
1488   vec_4x32u rgb;                                                               \
1489                                                                                \
1490   vec_8x8u d64_0x07;                                                           \
1491   vec_8x8u d64_1;                                                              \
1492   vec_8x8u d64_4;                                                              \
1493   vec_8x8u d64_128;                                                            \
1494                                                                                \
1495   dup_8x8b(d64_0x07, 0x07);                                                    \
1496   dup_8x8b(d64_1, 1);                                                          \
1497   dup_8x8b(d64_4, 4);                                                          \
1498   dup_8x8b(d64_128, 128);                                                      \
1499                                                                                \
1500   rgb_dx.low = psx_gpu->uvrg_dx.high;                                          \
1501   rgb_dx.e[2] = psx_gpu->b_block_span.e[1];                                    \
1502   shl_4x32b(rgb_dx4, rgb_dx, 2);                                               \
1503   shl_4x32b(rgb_dx8, rgb_dx, 3)                                                \
1504
1505 #define setup_blocks_variables_unshaded_textured(target)                       \
1506   vec_4x32u u_block;                                                           \
1507   vec_4x32u v_block;                                                           \
1508   vec_2x32u uv_dx = psx_gpu->uvrg_dx.low;                                      \
1509   vec_2x32u uv_dx4;                                                            \
1510   vec_2x32u uv_dx8;                                                            \
1511   vec_2x32u uv = psx_gpu->uvrg.low;                                            \
1512                                                                                \
1513   vec_16x8u texture_mask;                                                      \
1514   shl_2x32b(uv_dx4, uv_dx, 2);                                                 \
1515   shl_2x32b(uv_dx8, uv_dx, 3);                                                 \
1516   dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
1517   dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
1518
1519
1520 #define setup_blocks_variables_unshaded_untextured_direct()                    \
1521   or_8x16b(colors, colors, msb_mask)                                           \
1522
1523 #define setup_blocks_variables_unshaded_untextured_indirect()                  \
1524
1525 #define setup_blocks_variables_unshaded_untextured(target)                     \
1526   u32 color = psx_gpu->triangle_color;                                         \
1527   vec_8x16u colors;                                                            \
1528                                                                                \
1529   u32 color_r = color & 0xFF;                                                  \
1530   u32 color_g = (color >> 8) & 0xFF;                                           \
1531   u32 color_b = (color >> 16) & 0xFF;                                          \
1532                                                                                \
1533   color = (color_r >> 3) | ((color_g >> 3) << 5) |                             \
1534    ((color_b >> 3) << 10);                                                     \
1535   dup_8x16b(colors, color);                                                    \
1536   setup_blocks_variables_unshaded_untextured_##target()                        \
1537
1538 #define setup_blocks_span_initialize_dithered_textured()                       \
1539   vec_8x16u dither_offsets;                                                    \
1540   shl_long_8x8b(dither_offsets, dither_offsets_short, 4)                       \
1541
1542 #define setup_blocks_span_initialize_dithered_untextured()                     \
1543   vec_8x8u dither_offsets;                                                     \
1544   add_8x8b(dither_offsets, dither_offsets_short, d64_4)                        \
1545
1546 #define setup_blocks_span_initialize_dithered(texturing)                       \
1547   u32 dither_row = psx_gpu->dither_table[y & 0x3];                             \
1548   u32 dither_shift = (span_edge_data->left_x & 0x3) * 8;                       \
1549   vec_8x8s dither_offsets_short;                                               \
1550                                                                                \
1551   dither_row =                                                                 \
1552    (dither_row >> dither_shift) | (dither_row << (32 - dither_shift));         \
1553   dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row);         \
1554   setup_blocks_span_initialize_dithered_##texturing()                          \
1555
1556 #define setup_blocks_span_initialize_undithered(texturing)                     \
1557
1558
1559 #define setup_blocks_span_initialize_shaded_textured()                         \
1560 {                                                                              \
1561   vec_4x32u block_span;                                                        \
1562   u32 offset = span_edge_data->left_x;                                         \
1563                                                                                \
1564   uvrg = *span_uvrg_offset;                                                    \
1565   mla_scalar_4x32b(uvrg, uvrg_dx, offset);                                     \
1566   b = *span_b_offset;                                                          \
1567   b += b_dx * offset;                                                          \
1568                                                                                \
1569   dup_4x32b(u_block, uvrg.e[0]);                                               \
1570   dup_4x32b(v_block, uvrg.e[1]);                                               \
1571   dup_4x32b(r_block, uvrg.e[2]);                                               \
1572   dup_4x32b(g_block, uvrg.e[3]);                                               \
1573   dup_4x32b(b_block, b);                                                       \
1574                                                                                \
1575   block_span = psx_gpu->u_block_span;                                          \
1576   add_4x32b(u_block, u_block, block_span);                                     \
1577   block_span = psx_gpu->v_block_span;                                          \
1578   add_4x32b(v_block, v_block, block_span);                                     \
1579   block_span = psx_gpu->r_block_span;                                          \
1580   add_4x32b(r_block, r_block, block_span);                                     \
1581   block_span = psx_gpu->g_block_span;                                          \
1582   add_4x32b(g_block, g_block, block_span);                                     \
1583   block_span = psx_gpu->b_block_span;                                          \
1584   add_4x32b(b_block, b_block, block_span);                                     \
1585 }
1586   
1587 #define setup_blocks_span_initialize_shaded_untextured()                       \
1588 {                                                                              \
1589   vec_4x32u block_span;                                                        \
1590   u32 offset = span_edge_data->left_x;                                         \
1591                                                                                \
1592   rgb.low = span_uvrg_offset->high;                                            \
1593   rgb.high.e[0] = *span_b_offset;                                              \
1594   mla_scalar_4x32b(rgb, rgb_dx, offset);                                       \
1595                                                                                \
1596   dup_4x32b(r_block, rgb.e[0]);                                                \
1597   dup_4x32b(g_block, rgb.e[1]);                                                \
1598   dup_4x32b(b_block, rgb.e[2]);                                                \
1599                                                                                \
1600   block_span = psx_gpu->r_block_span;                                          \
1601   add_4x32b(r_block, r_block, block_span);                                     \
1602   block_span = psx_gpu->g_block_span;                                          \
1603   add_4x32b(g_block, g_block, block_span);                                     \
1604   block_span = psx_gpu->b_block_span;                                          \
1605   add_4x32b(b_block, b_block, block_span);                                     \
1606 }                                                                              \
1607   
1608 #define setup_blocks_span_initialize_unshaded_textured()                       \
1609 {                                                                              \
1610   vec_4x32u block_span;                                                        \
1611   u32 offset = span_edge_data->left_x;                                         \
1612                                                                                \
1613   uv = span_uvrg_offset->low;                                                  \
1614   mla_scalar_2x32b(uv, uv_dx, offset);                                         \
1615                                                                                \
1616   dup_4x32b(u_block, uv.e[0]);                                                 \
1617   dup_4x32b(v_block, uv.e[1]);                                                 \
1618                                                                                \
1619   block_span = psx_gpu->u_block_span;                                          \
1620   add_4x32b(u_block, u_block, block_span);                                     \
1621   block_span = psx_gpu->v_block_span;                                          \
1622   add_4x32b(v_block, v_block, block_span);                                     \
1623 }                                                                              \
1624
1625 #define setup_blocks_span_initialize_unshaded_untextured()                     \
1626
1627
1628 #define setup_blocks_texture_swizzled()                                        \
1629 {                                                                              \
1630   vec_8x8u u_saved = u;                                                        \
1631   sli_8x8b(u, v, 4);                                                           \
1632   sri_8x8b(v, u_saved, 4);                                                     \
1633 }                                                                              \
1634
1635 #define setup_blocks_texture_unswizzled()                                      \
1636
1637 #define setup_blocks_store_shaded_textured(swizzling, dithering, target,       \
1638  edge_type)                                                                    \
1639 {                                                                              \
1640   vec_8x16u u_whole;                                                           \
1641   vec_8x16u v_whole;                                                           \
1642   vec_8x16u r_whole;                                                           \
1643   vec_8x16u g_whole;                                                           \
1644   vec_8x16u b_whole;                                                           \
1645                                                                                \
1646   vec_8x8u u;                                                                  \
1647   vec_8x8u v;                                                                  \
1648   vec_8x8u r;                                                                  \
1649   vec_8x8u g;                                                                  \
1650   vec_8x8u b;                                                                  \
1651   vec_8x16u uv;                                                                \
1652                                                                                \
1653   vec_4x32u dx4;                                                               \
1654   vec_4x32u dx8;                                                               \
1655                                                                                \
1656   shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
1657   shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
1658   shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
1659   shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
1660   shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
1661                                                                                \
1662   dup_4x32b(dx4, uvrg_dx4.e[0]);                                               \
1663   add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
1664   dup_4x32b(dx4, uvrg_dx4.e[1]);                                               \
1665   add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
1666   dup_4x32b(dx4, uvrg_dx4.e[2]);                                               \
1667   add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
1668   dup_4x32b(dx4, uvrg_dx4.e[3]);                                               \
1669   add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
1670   dup_4x32b(dx4, b_dx4);                                                       \
1671   add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
1672                                                                                \
1673   mov_narrow_8x16b(u, u_whole);                                                \
1674   mov_narrow_8x16b(v, v_whole);                                                \
1675   mov_narrow_8x16b(r, r_whole);                                                \
1676   mov_narrow_8x16b(g, g_whole);                                                \
1677   mov_narrow_8x16b(b, b_whole);                                                \
1678                                                                                \
1679   dup_4x32b(dx8, uvrg_dx8.e[0]);                                               \
1680   add_4x32b(u_block, u_block, dx8);                                            \
1681   dup_4x32b(dx8, uvrg_dx8.e[1]);                                               \
1682   add_4x32b(v_block, v_block, dx8);                                            \
1683   dup_4x32b(dx8, uvrg_dx8.e[2]);                                               \
1684   add_4x32b(r_block, r_block, dx8);                                            \
1685   dup_4x32b(dx8, uvrg_dx8.e[3]);                                               \
1686   add_4x32b(g_block, g_block, dx8);                                            \
1687   dup_4x32b(dx8, b_dx8);                                                       \
1688   add_4x32b(b_block, b_block, dx8);                                            \
1689                                                                                \
1690   and_8x8b(u, u, texture_mask.low);                                            \
1691   and_8x8b(v, v, texture_mask.high);                                           \
1692   setup_blocks_texture_##swizzling();                                          \
1693                                                                                \
1694   zip_8x16b(uv, u, v);                                                         \
1695   block->uv = uv;                                                              \
1696   block->r = r;                                                                \
1697   block->g = g;                                                                \
1698   block->b = b;                                                                \
1699   block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
1700   block->fb_ptr = fb_ptr;                                                      \
1701 }                                                                              \
1702
1703 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target,     \
1704  edge_type)                                                                    \
1705 {                                                                              \
1706   vec_8x16u u_whole;                                                           \
1707   vec_8x16u v_whole;                                                           \
1708                                                                                \
1709   vec_8x8u u;                                                                  \
1710   vec_8x8u v;                                                                  \
1711   vec_8x16u uv;                                                                \
1712                                                                                \
1713   vec_4x32u dx4;                                                               \
1714   vec_4x32u dx8;                                                               \
1715                                                                                \
1716   shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
1717   shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
1718                                                                                \
1719   dup_4x32b(dx4, uv_dx4.e[0]);                                                 \
1720   add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
1721   dup_4x32b(dx4, uv_dx4.e[1]);                                                 \
1722   add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
1723                                                                                \
1724   mov_narrow_8x16b(u, u_whole);                                                \
1725   mov_narrow_8x16b(v, v_whole);                                                \
1726                                                                                \
1727   dup_4x32b(dx8, uv_dx8.e[0]);                                                 \
1728   add_4x32b(u_block, u_block, dx8);                                            \
1729   dup_4x32b(dx8, uv_dx8.e[1]);                                                 \
1730   add_4x32b(v_block, v_block, dx8);                                            \
1731                                                                                \
1732   and_8x8b(u, u, texture_mask.low);                                            \
1733   and_8x8b(v, v, texture_mask.high);                                           \
1734   setup_blocks_texture_##swizzling();                                          \
1735                                                                                \
1736   zip_8x16b(uv, u, v);                                                         \
1737   block->uv = uv;                                                              \
1738   block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
1739   block->fb_ptr = fb_ptr;                                                      \
1740 }                                                                              \
1741
1742 #define setup_blocks_store_shaded_untextured_dithered()                        \
1743   addq_8x8b(r, r, dither_offsets);                                             \
1744   addq_8x8b(g, g, dither_offsets);                                             \
1745   addq_8x8b(b, b, dither_offsets);                                             \
1746                                                                                \
1747   subq_8x8b(r, r, d64_4);                                                      \
1748   subq_8x8b(g, g, d64_4);                                                      \
1749   subq_8x8b(b, b, d64_4)                                                       \
1750
1751 #define setup_blocks_store_shaded_untextured_undithered()                      \
1752   
1753
1754 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels)            \
1755   block->pixels = _pixels;                                                     \
1756   block->fb_ptr = fb_ptr                                                       \
1757
1758 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels)            \
1759   block->pixels = _pixels;                                                     \
1760   block->fb_ptr = fb_ptr                                                       \
1761
1762 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect()            \
1763   mul_long_8x8b(pixels, r, d64_1)                                              \
1764
1765
1766 #define setup_blocks_store_untextured_pixels_direct_full(_pixels)              \
1767   store_8x16b(_pixels, fb_ptr)                                                 \
1768
1769 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels)              \
1770 {                                                                              \
1771   vec_8x16u fb_pixels;                                                         \
1772   vec_8x16u draw_mask;                                                         \
1773   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
1774                                                                                \
1775   load_8x16b(fb_pixels, fb_ptr);                                               \
1776   dup_8x16b(draw_mask, span_edge_data->right_mask);                            \
1777   tst_8x16b(draw_mask, draw_mask, test_mask);                                  \
1778   bif_8x16b(fb_pixels, _pixels, draw_mask);                                    \
1779   store_8x16b(fb_pixels, fb_ptr);                                              \
1780 }                                                                              \
1781
1782 #define setup_blocks_store_shaded_untextured_seed_pixels_direct()              \
1783   pixels = msb_mask;                                                           \
1784   mla_long_8x8b(pixels, r, d64_1)                                              \
1785
1786
1787 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target,     \
1788  edge_type)                                                                    \
1789 {                                                                              \
1790   vec_8x16u r_whole;                                                           \
1791   vec_8x16u g_whole;                                                           \
1792   vec_8x16u b_whole;                                                           \
1793                                                                                \
1794   vec_8x8u r;                                                                  \
1795   vec_8x8u g;                                                                  \
1796   vec_8x8u b;                                                                  \
1797                                                                                \
1798   vec_4x32u dx4;                                                               \
1799   vec_4x32u dx8;                                                               \
1800                                                                                \
1801   vec_8x16u pixels;                                                            \
1802                                                                                \
1803   shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
1804   shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
1805   shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
1806                                                                                \
1807   dup_4x32b(dx4, rgb_dx4.e[0]);                                                \
1808   add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
1809   dup_4x32b(dx4, rgb_dx4.e[1]);                                                \
1810   add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
1811   dup_4x32b(dx4, rgb_dx4.e[2]);                                                \
1812   add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
1813                                                                                \
1814   mov_narrow_8x16b(r, r_whole);                                                \
1815   mov_narrow_8x16b(g, g_whole);                                                \
1816   mov_narrow_8x16b(b, b_whole);                                                \
1817                                                                                \
1818   dup_4x32b(dx8, rgb_dx8.e[0]);                                                \
1819   add_4x32b(r_block, r_block, dx8);                                            \
1820   dup_4x32b(dx8, rgb_dx8.e[1]);                                                \
1821   add_4x32b(g_block, g_block, dx8);                                            \
1822   dup_4x32b(dx8, rgb_dx8.e[2]);                                                \
1823   add_4x32b(b_block, b_block, dx8);                                            \
1824                                                                                \
1825   setup_blocks_store_shaded_untextured_##dithering();                          \
1826                                                                                \
1827   shr_8x8b(r, r, 3);                                                           \
1828   bic_8x8b(g, g, d64_0x07);                                                    \
1829   bic_8x8b(b, b, d64_0x07);                                                    \
1830                                                                                \
1831   setup_blocks_store_shaded_untextured_seed_pixels_##target();                 \
1832   mla_long_8x8b(pixels, g, d64_4);                                             \
1833   mla_long_8x8b(pixels, b, d64_128)                                            \
1834                                                                                \
1835   setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels);         \
1836 }                                                                              \
1837
1838 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target,   \
1839  edge_type)                                                                    \
1840   setup_blocks_store_untextured_pixels_##target##_##edge_type(colors)          \
1841
1842
1843 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits)           \
1844   (_block)->draw_mask_bits = bits                                              \
1845
1846 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits)         \
1847 {                                                                              \
1848   vec_8x16u bits_mask;                                                         \
1849   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
1850   dup_8x16b(bits_mask, bits);                                                  \
1851   tst_8x16b(bits_mask, bits_mask, test_mask);                                  \
1852   (_block)->draw_mask = bits_mask;                                             \
1853 }                                                                              \
1854
1855 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits)           \
1856
1857
1858 #define setup_blocks_add_blocks_indirect()                                     \
1859   num_blocks += span_num_blocks;                                               \
1860                                                                                \
1861   if(num_blocks > MAX_BLOCKS)                                                  \
1862   {                                                                            \
1863     psx_gpu->num_blocks = num_blocks - span_num_blocks;                        \
1864     flush_render_block_buffer(psx_gpu);                                        \
1865     num_blocks = span_num_blocks;                                              \
1866     block = psx_gpu->blocks;                                                   \
1867   }                                                                            \
1868
1869 #define setup_blocks_add_blocks_direct()                                       \
1870   texel_blocks_untextured += span_num_blocks;                                  \
1871   span_pixel_blocks += span_num_blocks                                         \
1872
1873
1874 #define setup_blocks_builder(shading, texturing, dithering, sw, target)        \
1875 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target(     \
1876  psx_gpu_struct *psx_gpu)                                                      \
1877 {                                                                              \
1878   setup_blocks_load_msb_mask_##target();                                       \
1879   setup_blocks_variables_##shading##_##texturing(target);                      \
1880                                                                                \
1881   edge_data_struct *span_edge_data = psx_gpu->span_edge_data;                  \
1882   vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset;                     \
1883   u32 *span_b_offset = psx_gpu->span_b_offset;                                 \
1884                                                                                \
1885   block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks;                 \
1886                                                                                \
1887   u32 num_spans = psx_gpu->num_spans;                                          \
1888                                                                                \
1889   u16 *fb_ptr;                                                                 \
1890   u32 y;                                                                       \
1891                                                                                \
1892   u32 num_blocks = psx_gpu->num_blocks;                                        \
1893   u32 span_num_blocks;                                                         \
1894                                                                                \
1895   while(num_spans)                                                             \
1896   {                                                                            \
1897     span_num_blocks = span_edge_data->num_blocks;                              \
1898     if(span_num_blocks)                                                        \
1899     {                                                                          \
1900       y = span_edge_data->y;                                                   \
1901       fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024);    \
1902                                                                                \
1903       setup_blocks_span_initialize_##shading##_##texturing();                  \
1904       setup_blocks_span_initialize_##dithering(texturing);                     \
1905                                                                                \
1906       setup_blocks_add_blocks_##target();                                      \
1907                                                                                \
1908       s32 pixel_span = span_num_blocks * 8;                                    \
1909       pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF);     \
1910       span_pixels += pixel_span;                                               \
1911                                                                                \
1912       span_num_blocks--;                                                       \
1913       while(span_num_blocks)                                                   \
1914       {                                                                        \
1915         setup_blocks_store_##shading##_##texturing(sw, dithering, target,      \
1916          full);                                                                \
1917         setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00);      \
1918                                                                                \
1919         fb_ptr += 8;                                                           \
1920         block++;                                                               \
1921         span_num_blocks--;                                                     \
1922       }                                                                        \
1923                                                                                \
1924       setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1925       setup_blocks_store_draw_mask_##texturing##_##target(block,               \
1926        span_edge_data->right_mask);                                            \
1927                                                                                \
1928       block++;                                                                 \
1929     }                                                                          \
1930     else                                                                       \
1931     {                                                                          \
1932       zero_block_spans++;                                                      \
1933     }                                                                          \
1934                                                                                \
1935     num_spans--;                                                               \
1936     span_edge_data++;                                                          \
1937     span_uvrg_offset++;                                                        \
1938     span_b_offset++;                                                           \
1939   }                                                                            \
1940                                                                                \
1941   psx_gpu->num_blocks = num_blocks;                                            \
1942 }                                                                              \
1943
1944 void setup_blocks_shaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1945  *psx_gpu);
1946
1947 void setup_blocks_shaded_untextured_dithered_unswizzled_indirect(psx_gpu_struct
1948  *psx_gpu);
1949 void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(
1950  psx_gpu_struct *psx_gpu);
1951 void setup_blocks_shaded_untextured_dithered_unswizzled_direct(psx_gpu_struct
1952  *psx_gpu);
1953 void setup_blocks_shaded_untextured_undithered_unswizzled_direct(
1954  psx_gpu_struct *psx_gpu);
1955
1956 void setup_blocks_unshaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1957  *psx_gpu);
1958 void setup_blocks_unshaded_untextured_undithered_unswizzled_indirect(
1959  psx_gpu_struct *psx_gpu);
1960 void setup_blocks_unshaded_untextured_undithered_unswizzled_direct(
1961  psx_gpu_struct *psx_gpu);
1962
1963 void setup_blocks_shaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1964  *psx_gpu);
1965 void setup_blocks_unshaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1966  *psx_gpu);
1967
1968
1969 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1970
1971 #ifndef NEON_BUILD
1972
1973 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1974 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1975
1976 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1977 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1978
1979 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1980 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1981 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1982 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1983
1984 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1985 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1986
1987 #endif
1988
1989 void texture_blocks_untextured(psx_gpu_struct *psx_gpu);
1990 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu);
1991 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu);
1992 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu);
1993
1994 #ifndef NEON_BUILD
1995
1996 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1997 {
1998   if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1999     texel_blocks_untextured += psx_gpu->num_blocks;
2000 }
2001
2002 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
2003 {
2004   block_struct *block = psx_gpu->blocks;
2005   u32 num_blocks = psx_gpu->num_blocks;
2006   texel_blocks_4bpp += num_blocks;
2007
2008   vec_8x8u texels_low;
2009   vec_8x8u texels_high;
2010   vec_8x8u texels;
2011   vec_8x16u pixels;
2012
2013   vec_8x16u clut_a;
2014   vec_8x16u clut_b;
2015   vec_16x8u clut_low;
2016   vec_16x8u clut_high;
2017
2018   u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2019   u16 *clut_ptr = psx_gpu->clut_ptr;
2020
2021   // Can be done with one deinterleaving load on NEON
2022   load_8x16b(clut_a, clut_ptr);
2023   load_8x16b(clut_b, clut_ptr + 8);
2024   unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
2025
2026   if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
2027     update_texture_4bpp_cache(psx_gpu);
2028
2029   while(num_blocks)
2030   {
2031     texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
2032     texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
2033     texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
2034     texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
2035     texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
2036     texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
2037     texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
2038     texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
2039
2040     tbl_16(texels_low, texels, clut_low);
2041     tbl_16(texels_high, texels, clut_high);
2042
2043     // Can be done with an interleaving store on NEON
2044     zip_8x16b(pixels, texels_low, texels_high);
2045
2046     block->texels = pixels;
2047
2048     num_blocks--;
2049     block++;
2050   }
2051 }
2052
2053 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2054 {
2055   block_struct *block = psx_gpu->blocks;
2056   u32 num_blocks = psx_gpu->num_blocks;
2057
2058   texel_blocks_8bpp += num_blocks;
2059
2060   if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2061     update_texture_8bpp_cache(psx_gpu);
2062
2063   vec_8x16u texels;
2064   u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2065
2066   u32 texel;
2067   u32 offset;
2068   u32 i;
2069
2070   while(num_blocks)
2071   {
2072     for(i = 0; i < 8; i++)
2073     {
2074       offset = block->uv.e[i];
2075
2076       texel = texture_ptr_8bpp[offset];
2077       texels.e[i] = psx_gpu->clut_ptr[texel];
2078     }
2079
2080     block->texels = texels;
2081
2082     num_blocks--;
2083     block++;
2084   }
2085 }
2086
2087 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2088 {
2089   block_struct *block = psx_gpu->blocks;
2090   u32 num_blocks = psx_gpu->num_blocks;
2091
2092   texel_blocks_16bpp += num_blocks;
2093
2094   vec_8x16u texels;
2095
2096   u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2097   u32 offset;
2098   u32 i;
2099
2100   while(num_blocks)
2101   {
2102     for(i = 0; i < 8; i++)
2103     {
2104       offset = block->uv.e[i];
2105       offset += ((offset & 0xFF00) * 3);
2106
2107       texels.e[i] = texture_ptr_16bpp[offset];
2108     }
2109
2110     block->texels = texels;
2111
2112     num_blocks--;
2113     block++;
2114   }
2115 }
2116
2117 #endif
2118
2119
2120 #define shade_blocks_load_msb_mask_indirect()                                  \
2121
2122 #define shade_blocks_load_msb_mask_direct()                                    \
2123   vec_8x16u msb_mask;                                                          \
2124   dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
2125
2126 #define shade_blocks_store_indirect(_draw_mask, _pixels)                       \
2127   block->draw_mask = _draw_mask;                                               \
2128   block->pixels = _pixels                                                      \
2129
2130 #define shade_blocks_store_direct(_draw_mask, _pixels)                         \
2131 {                                                                              \
2132   vec_8x16u fb_pixels;                                                         \
2133   or_8x16b(_pixels, _pixels, msb_mask);                                        \
2134   load_8x16b(fb_pixels, block->fb_ptr);                                        \
2135   bif_8x16b(fb_pixels, _pixels, _draw_mask);                                   \
2136   store_8x16b(fb_pixels, block->fb_ptr);                                       \
2137 }                                                                              \
2138
2139
2140 #define shade_blocks_textured_false_modulated_check_dithered(target)           \
2141   if(psx_gpu->triangle_color == 0x808080)                                      \
2142   {                                                                            \
2143     false_modulated_blocks += num_blocks;                                      \
2144   }                                                                            \
2145
2146 #define shade_blocks_textured_false_modulated_check_undithered(target)         \
2147   if(psx_gpu->triangle_color == 0x808080)                                      \
2148   {                                                                            \
2149                                                                                \
2150     shade_blocks_textured_unmodulated_##target(psx_gpu);                       \
2151     false_modulated_blocks += num_blocks;                                      \
2152     return;                                                                    \
2153   }                                                                            \
2154
2155
2156 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering,       \
2157  target)                                                                       \
2158
2159 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering,     \
2160  target)                                                                       \
2161 {                                                                              \
2162   u32 color = psx_gpu->triangle_color;                                         \
2163   dup_8x8b(colors_r, color);                                                   \
2164   dup_8x8b(colors_g, color >> 8);                                              \
2165   dup_8x8b(colors_b, color >> 16);                                             \
2166   shade_blocks_textured_false_modulated_check_##dithering(target);             \
2167 }                                                                              \
2168
2169 #define shade_blocks_textured_modulated_shaded_block_load()                    \
2170   colors_r = block->r;                                                         \
2171   colors_g = block->g;                                                         \
2172   colors_b = block->b                                                          \
2173
2174 #define shade_blocks_textured_modulated_unshaded_block_load()                  \
2175
2176 #define shade_blocks_textured_modulate_dithered(component)                     \
2177   pixels_##component = block->dither_offsets;                                  \
2178   mla_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
2179
2180 #define shade_blocks_textured_modulate_undithered(component)                   \
2181   mul_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
2182
2183 #define shade_blocks_textured_modulated_builder(shading, dithering, target)    \
2184 void shade_blocks_##shading##_textured_modulated_##dithering##_##target(       \
2185  psx_gpu_struct *psx_gpu)                                                      \
2186 {                                                                              \
2187   block_struct *block = psx_gpu->blocks;                                       \
2188   u32 num_blocks = psx_gpu->num_blocks;                                        \
2189   vec_8x16u texels;                                                            \
2190                                                                                \
2191   vec_8x8u texels_r;                                                           \
2192   vec_8x8u texels_g;                                                           \
2193   vec_8x8u texels_b;                                                           \
2194                                                                                \
2195   vec_8x8u colors_r;                                                           \
2196   vec_8x8u colors_g;                                                           \
2197   vec_8x8u colors_b;                                                           \
2198                                                                                \
2199   vec_8x8u pixels_r_low;                                                       \
2200   vec_8x8u pixels_g_low;                                                       \
2201   vec_8x8u pixels_b_low;                                                       \
2202   vec_8x16u pixels;                                                            \
2203                                                                                \
2204   vec_8x16u pixels_r;                                                          \
2205   vec_8x16u pixels_g;                                                          \
2206   vec_8x16u pixels_b;                                                          \
2207                                                                                \
2208   vec_8x16u draw_mask;                                                         \
2209   vec_8x16u zero_mask;                                                         \
2210                                                                                \
2211   vec_8x8u d64_0x07;                                                           \
2212   vec_8x8u d64_0x1F;                                                           \
2213   vec_8x8u d64_1;                                                              \
2214   vec_8x8u d64_4;                                                              \
2215   vec_8x8u d64_128;                                                            \
2216                                                                                \
2217   vec_8x16u d128_0x8000;                                                       \
2218                                                                                \
2219   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2220   u32 draw_mask_bits;                                                          \
2221   shade_blocks_load_msb_mask_##target();                                       \
2222                                                                                \
2223   dup_8x8b(d64_0x07, 0x07);                                                    \
2224   dup_8x8b(d64_0x1F, 0x1F);                                                    \
2225   dup_8x8b(d64_1, 1);                                                          \
2226   dup_8x8b(d64_4, 4);                                                          \
2227   dup_8x8b(d64_128, 128);                                                      \
2228                                                                                \
2229   dup_8x16b(d128_0x8000, 0x8000);                                              \
2230                                                                                \
2231   shade_blocks_textured_modulated_##shading##_primitive_load(dithering,        \
2232    target);                                                                    \
2233                                                                                \
2234   while(num_blocks)                                                            \
2235   {                                                                            \
2236     draw_mask_bits = block->draw_mask_bits;                                    \
2237     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2238     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2239                                                                                \
2240     shade_blocks_textured_modulated_##shading##_block_load();                  \
2241                                                                                \
2242     texels = block->texels;                                                    \
2243                                                                                \
2244     mov_narrow_8x16b(texels_r, texels);                                        \
2245     shr_narrow_8x16b(texels_g, texels, 5);                                     \
2246     shr_narrow_8x16b(texels_b, texels, 7);                                     \
2247                                                                                \
2248     and_8x8b(texels_r, texels_r, d64_0x1F);                                    \
2249     and_8x8b(texels_g, texels_g, d64_0x1F);                                    \
2250     shr_8x8b(texels_b, texels_b, 3);                                           \
2251                                                                                \
2252     shade_blocks_textured_modulate_##dithering(r);                             \
2253     shade_blocks_textured_modulate_##dithering(g);                             \
2254     shade_blocks_textured_modulate_##dithering(b);                             \
2255                                                                                \
2256     cmpeqz_8x16b(zero_mask, texels);                                           \
2257     and_8x16b(pixels, texels, d128_0x8000);                                    \
2258                                                                                \
2259     shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4);                       \
2260     shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4);                       \
2261     shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4);                       \
2262                                                                                \
2263     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2264                                                                                \
2265     shr_8x8b(pixels_r_low, pixels_r_low, 3);                                   \
2266     bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07);                            \
2267     bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07);                            \
2268                                                                                \
2269     mla_long_8x8b(pixels, pixels_r_low, d64_1);                                \
2270     mla_long_8x8b(pixels, pixels_g_low, d64_4);                                \
2271     mla_long_8x8b(pixels, pixels_b_low, d64_128);                              \
2272                                                                                \
2273     shade_blocks_store_##target(zero_mask, pixels);                            \
2274                                                                                \
2275     num_blocks--;                                                              \
2276     block++;                                                                   \
2277   }                                                                            \
2278 }                                                                              \
2279
2280 void shade_blocks_shaded_textured_modulated_dithered_direct(psx_gpu_struct
2281  *psx_gpu);
2282 void shade_blocks_shaded_textured_modulated_undithered_direct(psx_gpu_struct
2283  *psx_gpu);
2284 void shade_blocks_unshaded_textured_modulated_dithered_direct(psx_gpu_struct
2285  *psx_gpu);
2286 void shade_blocks_unshaded_textured_modulated_undithered_direct(psx_gpu_struct
2287  *psx_gpu);
2288
2289 void shade_blocks_shaded_textured_modulated_dithered_indirect(psx_gpu_struct
2290  *psx_gpu);
2291 void shade_blocks_shaded_textured_modulated_undithered_indirect(psx_gpu_struct
2292  *psx_gpu);
2293 void shade_blocks_unshaded_textured_modulated_dithered_indirect(psx_gpu_struct
2294  *psx_gpu);
2295 void shade_blocks_unshaded_textured_modulated_undithered_indirect(psx_gpu_struct
2296  *psx_gpu);
2297
2298 void shade_blocks_textured_unmodulated_indirect(psx_gpu_struct *psx_gpu);
2299 void shade_blocks_textured_unmodulated_direct(psx_gpu_struct *psx_gpu);
2300
2301 #ifndef NEON_BUILD
2302
2303 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2304 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2305 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2306 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2307
2308 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2309 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2310 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2311 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2312
2313 #endif
2314
2315
2316 #define shade_blocks_textured_unmodulated_builder(target)                      \
2317 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu)       \
2318 {                                                                              \
2319   block_struct *block = psx_gpu->blocks;                                       \
2320   u32 num_blocks = psx_gpu->num_blocks;                                        \
2321   vec_8x16u draw_mask;                                                         \
2322   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2323   u32 draw_mask_bits;                                                          \
2324                                                                                \
2325   vec_8x16u pixels;                                                            \
2326   shade_blocks_load_msb_mask_##target();                                       \
2327                                                                                \
2328   while(num_blocks)                                                            \
2329   {                                                                            \
2330     vec_8x16u zero_mask;                                                       \
2331                                                                                \
2332     draw_mask_bits = block->draw_mask_bits;                                    \
2333     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2334     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2335                                                                                \
2336     pixels = block->texels;                                                    \
2337                                                                                \
2338     cmpeqz_8x16b(zero_mask, pixels);                                           \
2339     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2340                                                                                \
2341     shade_blocks_store_##target(zero_mask, pixels);                            \
2342                                                                                \
2343     num_blocks--;                                                              \
2344     block++;                                                                   \
2345   }                                                                            \
2346 }                                                                              \
2347
2348 #define shade_blocks_textured_unmodulated_dithered_builder(target)             \
2349 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct        \
2350  *psx_gpu)                                                                     \
2351 {                                                                              \
2352   block_struct *block = psx_gpu->blocks;                                       \
2353   u32 num_blocks = psx_gpu->num_blocks;                                        \
2354   vec_8x16u draw_mask;                                                         \
2355   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2356   u32 draw_mask_bits;                                                          \
2357                                                                                \
2358   vec_8x16u pixels;                                                            \
2359   shade_blocks_load_msb_mask_##target();                                       \
2360                                                                                \
2361   while(num_blocks)                                                            \
2362   {                                                                            \
2363     vec_8x16u zero_mask;                                                       \
2364                                                                                \
2365     draw_mask_bits = block->draw_mask_bits;                                    \
2366     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2367     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2368                                                                                \
2369     pixels = block->texels;                                                    \
2370                                                                                \
2371     cmpeqz_8x16b(zero_mask, pixels);                                           \
2372     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2373                                                                                \
2374     shade_blocks_store_##target(zero_mask, pixels);                            \
2375                                                                                \
2376     num_blocks--;                                                              \
2377     block++;                                                                   \
2378   }                                                                            \
2379 }                                                                              \
2380
2381 #ifndef NEON_BUILD
2382
2383 shade_blocks_textured_unmodulated_builder(indirect)
2384 shade_blocks_textured_unmodulated_builder(direct)
2385
2386 #endif
2387
2388
2389 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu);
2390 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu);
2391
2392 #ifndef NEON_BUILD
2393                                                                                
2394 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2395 {
2396 }
2397
2398 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2399 {
2400   block_struct *block = psx_gpu->blocks;
2401   u32 num_blocks = psx_gpu->num_blocks;
2402
2403   vec_8x16u pixels = block->pixels;
2404   shade_blocks_load_msb_mask_direct();
2405
2406   while(num_blocks)
2407   {
2408     shade_blocks_store_direct(block->draw_mask, pixels);
2409
2410     num_blocks--;
2411     block++;
2412   }
2413 }
2414
2415 #endif
2416
2417 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2418 {
2419 }
2420
2421
2422 #define blend_blocks_mask_evaluate_on()                                        \
2423   vec_8x16u mask_pixels;                                                       \
2424   cmpltz_8x16b(mask_pixels, framebuffer_pixels);                               \
2425   or_8x16b(draw_mask, draw_mask, mask_pixels)                                  \
2426
2427 #define blend_blocks_mask_evaluate_off()                                       \
2428
2429 #define blend_blocks_average()                                                 \
2430 {                                                                              \
2431   vec_8x16u pixels_no_msb;                                                     \
2432   vec_8x16u fb_pixels_no_msb;                                                  \
2433                                                                                \
2434   vec_8x16u d128_0x0421;                                                       \
2435   vec_8x16u d128_0x8000;                                                       \
2436                                                                                \
2437   dup_8x16b(d128_0x0421, 0x0421);                                              \
2438   dup_8x16b(d128_0x8000, 0x8000);                                              \
2439                                                                                \
2440   eor_8x16b(blend_pixels, pixels, framebuffer_pixels);                         \
2441   bic_8x16b(pixels_no_msb, pixels, d128_0x8000);                               \
2442   and_8x16b(blend_pixels, blend_pixels, d128_0x0421);                          \
2443   sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels);                        \
2444   bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000);                \
2445   average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels);                 \
2446 }                                                                              \
2447
2448 #define blend_blocks_add()                                                     \
2449 {                          &