psx_gpu: implement setup_sprite_untextured in asm
[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   if (&span_edge_data_element - psx_gpu->span_edge_data < psx_gpu->num_spans)  \
773   {                                                                            \
774     if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW)                 \
775       *(int *)0 = 1;                                                           \
776     if(span_edge_data_element.y > 2048)                                        \
777       *(int *)0 = 1;                                                           \
778   }                                                                            \
779
780 #else
781 #define setup_spans_debug_check(span_edge_data_element)                        \
782
783 #endif
784
785 #define setup_spans_prologue_alternate_yes()                                   \
786   vec_2x64s alternate_x;                                                       \
787   vec_2x64s alternate_dx_dy;                                                   \
788   vec_4x32s alternate_x_32;                                                    \
789   vec_2x32s alternate_x_16;                                                    \
790                                                                                \
791   vec_4x16u alternate_select;                                                  \
792   vec_4x16s y_mid_point;                                                       \
793                                                                                \
794   s32 y_b = v_b->y;                                                            \
795   s64 edge_alt;                                                                \
796   s32 edge_dx_dy_alt;                                                          \
797   u32 edge_shift_alt                                                           \
798
799 #define setup_spans_prologue_alternate_no()                                    \
800
801 #define setup_spans_prologue(alternate_active)                                 \
802   edge_data_struct *span_edge_data;                                            \
803   vec_4x32u *span_uvrg_offset;                                                 \
804   u32 *span_b_offset;                                                          \
805                                                                                \
806   s32 clip;                                                                    \
807                                                                                \
808   vec_2x64s edges_xy;                                                          \
809   vec_2x32s edges_dx_dy;                                                       \
810   vec_2x32u edge_shifts;                                                       \
811                                                                                \
812   vec_2x64s left_x, right_x;                                                   \
813   vec_2x64s left_dx_dy, right_dx_dy;                                           \
814   vec_4x32s left_x_32, right_x_32;                                             \
815   vec_8x16s left_right_x_16;                                                   \
816   vec_4x16s y_x4;                                                              \
817   vec_8x16s left_edge;                                                         \
818   vec_8x16s right_edge;                                                        \
819   vec_4x16u span_shift;                                                        \
820                                                                                \
821   vec_2x32u c_0x01;                                                            \
822   vec_4x16u c_0x04;                                                            \
823   vec_4x16u c_0xFFFE;                                                          \
824   vec_4x16u c_0x07;                                                            \
825                                                                                \
826   vec_2x32s x_starts;                                                          \
827   vec_2x32s x_ends;                                                            \
828                                                                                \
829   s32 x_a = v_a->x;                                                            \
830   s32 x_b = v_b->x;                                                            \
831   s32 x_c = v_c->x;                                                            \
832   s32 y_a = v_a->y;                                                            \
833   s32 y_c = v_c->y;                                                            \
834                                                                                \
835   vec_4x32u uvrg = psx_gpu->uvrg;                                              \
836   vec_4x32u uvrg_dy = psx_gpu->uvrg_dy;                                        \
837   u32 b = psx_gpu->b;                                                          \
838   u32 b_dy = psx_gpu->b_dy;                                                    \
839                                                                                \
840   dup_2x32b(c_0x01, 0x01);                                                     \
841   setup_spans_prologue_alternate_##alternate_active()                          \
842
843 #define setup_spans_prologue_b()                                               \
844   span_edge_data = psx_gpu->span_edge_data;                                    \
845   span_uvrg_offset = psx_gpu->span_uvrg_offset;                                \
846   span_b_offset = psx_gpu->span_b_offset;                                      \
847                                                                                \
848   vec_8x16u c_0x0001;                                                          \
849                                                                                \
850   dup_8x16b(c_0x0001, 0x0001);                                                 \
851   dup_8x16b(left_edge, psx_gpu->viewport_start_x);                             \
852   dup_8x16b(right_edge, psx_gpu->viewport_end_x);                              \
853   add_8x16b(right_edge, right_edge, c_0x0001);                                 \
854   dup_4x16b(c_0x04, 0x04);                                                     \
855   dup_4x16b(c_0x07, 0x07);                                                     \
856   dup_4x16b(c_0xFFFE, 0xFFFE);                                                 \
857
858
859 #define compute_edge_delta_x2()                                                \
860 {                                                                              \
861   vec_2x32s heights;                                                           \
862   vec_2x32s height_reciprocals;                                                \
863   vec_2x32s heights_b;                                                         \
864   vec_4x32u widths;                                                            \
865                                                                                \
866   u32 edge_shift = reciprocal_table[height];                                   \
867                                                                                \
868   dup_2x32b(heights, height);                                                  \
869   sub_2x32b(widths, x_ends, x_starts);                                         \
870                                                                                \
871   dup_2x32b(edge_shifts, edge_shift);                                          \
872   sub_2x32b(heights_b, heights, c_0x01);                                       \
873   shr_2x32b(height_reciprocals, edge_shifts, 10);                              \
874                                                                                \
875   mla_2x32b(heights_b, x_starts, heights);                                     \
876   bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
877   mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
878   mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
879 }                                                                              \
880
881 #define compute_edge_delta_x3(start_c, height_a, height_b)                     \
882 {                                                                              \
883   vec_2x32s heights;                                                           \
884   vec_2x32s height_reciprocals;                                                \
885   vec_2x32s heights_b;                                                         \
886   vec_2x32u widths;                                                            \
887                                                                                \
888   u32 width_alt;                                                               \
889   s32 height_b_alt;                                                            \
890   u32 height_reciprocal_alt;                                                   \
891                                                                                \
892   heights.e[0] = height_a;                                                     \
893   heights.e[1] = height_b;                                                     \
894                                                                                \
895   edge_shifts.e[0] = reciprocal_table[height_a];                               \
896   edge_shifts.e[1] = reciprocal_table[height_b];                               \
897   edge_shift_alt = reciprocal_table[height_minor_b];                           \
898                                                                                \
899   sub_2x32b(widths, x_ends, x_starts);                                         \
900   width_alt = x_c - start_c;                                                   \
901                                                                                \
902   shr_2x32b(height_reciprocals, edge_shifts, 10);                              \
903   height_reciprocal_alt = edge_shift_alt >> 10;                                \
904                                                                                \
905   bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0);              \
906   edge_shift_alt &= 0x1F;                                                      \
907                                                                                \
908   sub_2x32b(heights_b, heights, c_0x01);                                       \
909   height_b_alt = height_minor_b - 1;                                           \
910                                                                                \
911   mla_2x32b(heights_b, x_starts, heights);                                     \
912   height_b_alt += height_minor_b * start_c;                                    \
913                                                                                \
914   mul_long_2x32b(edges_xy, heights_b, height_reciprocals);                     \
915   edge_alt = (s64)height_b_alt * height_reciprocal_alt;                        \
916                                                                                \
917   mul_2x32b(edges_dx_dy, widths, height_reciprocals);                          \
918   edge_dx_dy_alt = width_alt * height_reciprocal_alt;                          \
919 }                                                                              \
920
921
922 #define setup_spans_adjust_y_up()                                              \
923   sub_4x32b(y_x4, y_x4, c_0x04)                                                \
924
925 #define setup_spans_adjust_y_down()                                            \
926   add_4x32b(y_x4, y_x4, c_0x04)                                                \
927
928 #define setup_spans_adjust_interpolants_up()                                   \
929   sub_4x32b(uvrg, uvrg, uvrg_dy);                                              \
930   b -= b_dy                                                                    \
931
932 #define setup_spans_adjust_interpolants_down()                                 \
933   add_4x32b(uvrg, uvrg, uvrg_dy);                                              \
934   b += b_dy                                                                    \
935
936
937 #define setup_spans_clip_interpolants_increment()                              \
938   mla_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
939   b += b_dy * clip                                                             \
940
941 #define setup_spans_clip_interpolants_decrement()                              \
942   mls_scalar_4x32b(uvrg, uvrg_dy, clip);                                       \
943   b -= b_dy * clip                                                             \
944
945 #define setup_spans_clip_alternate_yes()                                       \
946   edge_alt += edge_dx_dy_alt * (s64)(clip)                                     \
947
948 #define setup_spans_clip_alternate_no()                                        \
949
950 #define setup_spans_clip(direction, alternate_active)                          \
951 {                                                                              \
952   clipped_triangles++;                                                         \
953   mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip);                     \
954   setup_spans_clip_alternate_##alternate_active();                             \
955   setup_spans_clip_interpolants_##direction();                                 \
956 }                                                                              \
957
958
959 #define setup_spans_adjust_edges_alternate_no(left_index, right_index)         \
960 {                                                                              \
961   vec_2x64u edge_shifts_64;                                                    \
962   vec_2x64s edges_dx_dy_64;                                                    \
963                                                                                \
964   mov_wide_2x32b(edge_shifts_64, edge_shifts);                                 \
965   shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64);                      \
966                                                                                \
967   mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy);                                 \
968   shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64);          \
969                                                                                \
970   left_x.e[0] = edges_xy.e[left_index];                                        \
971   right_x.e[0] = edges_xy.e[right_index];                                      \
972                                                                                \
973   left_dx_dy.e[0] = edges_dx_dy_64.e[left_index];                              \
974   left_dx_dy.e[1] = edges_dx_dy_64.e[left_index];                              \
975   right_dx_dy.e[0] = edges_dx_dy_64.e[right_index];                            \
976   right_dx_dy.e[1] = edges_dx_dy_64.e[right_index];                            \
977                                                                                \
978   add_1x64b(left_x.high, left_x.low, left_dx_dy.low);                          \
979   add_1x64b(right_x.high, right_x.low, right_dx_dy.low);                       \
980                                                                                \
981   add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy);                               \
982   add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy);                            \
983 }                                                                              \
984
985 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index)        \
986 {                                                                              \
987   setup_spans_adjust_edges_alternate_no(left_index, right_index);              \
988   s64 edge_dx_dy_alt_64;                                                       \
989                                                                                \
990   dup_4x16b(y_mid_point, y_b);                                                 \
991                                                                                \
992   edge_alt <<= edge_shift_alt;                                                 \
993   edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt;                   \
994                                                                                \
995   alternate_x.e[0] = edge_alt;                                                 \
996   alternate_dx_dy.e[0] = edge_dx_dy_alt_64;                                    \
997   alternate_dx_dy.e[1] = edge_dx_dy_alt_64;                                    \
998                                                                                \
999   add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low);           \
1000   add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy);                \
1001 }                                                                              \
1002
1003
1004 #define setup_spans_y_select_up()                                              \
1005   cmplt_4x16b(alternate_select, y_x4, y_mid_point)                             \
1006
1007 #define setup_spans_y_select_down()                                            \
1008   cmpgt_4x16b(alternate_select, y_x4, y_mid_point)                             \
1009
1010 #define setup_spans_y_select_alternate_yes(direction)                          \
1011   setup_spans_y_select_##direction()                                           \
1012
1013 #define setup_spans_y_select_alternate_no(direction)                           \
1014
1015 #define setup_spans_alternate_select_left()                                    \
1016   bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select)             \
1017
1018 #define setup_spans_alternate_select_right()                                   \
1019   bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select)            \
1020
1021 #define setup_spans_alternate_select_none()                                    \
1022
1023 #define setup_spans_increment_alternate_yes()                                  \
1024   shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32);                       \
1025   add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
1026   shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32);                      \
1027   add_2x64b(alternate_x, alternate_x, alternate_dx_dy);                        \
1028   mov_narrow_4x32b(alternate_x_16, alternate_x_32)                             \
1029
1030 #define setup_spans_increment_alternate_no()                                   \
1031
1032 #define setup_spans_set_x4(alternate, direction, alternate_active)             \
1033 {                                                                              \
1034   span_uvrg_offset[0] = uvrg;                                                  \
1035   span_b_offset[0] = b;                                                        \
1036   setup_spans_adjust_interpolants_##direction();                               \
1037                                                                                \
1038   span_uvrg_offset[1] = uvrg;                                                  \
1039   span_b_offset[1] = b;                                                        \
1040   setup_spans_adjust_interpolants_##direction();                               \
1041                                                                                \
1042   span_uvrg_offset[2] = uvrg;                                                  \
1043   span_b_offset[2] = b;                                                        \
1044   setup_spans_adjust_interpolants_##direction();                               \
1045                                                                                \
1046   span_uvrg_offset[3] = uvrg;                                                  \
1047   span_b_offset[3] = b;                                                        \
1048   setup_spans_adjust_interpolants_##direction();                               \
1049                                                                                \
1050   span_uvrg_offset += 4;                                                       \
1051   span_b_offset += 4;                                                          \
1052                                                                                \
1053   shr_narrow_2x64b(left_x_32.low, left_x, 32);                                 \
1054   shr_narrow_2x64b(right_x_32.low, right_x, 32);                               \
1055                                                                                \
1056   add_2x64b(left_x, left_x, left_dx_dy);                                       \
1057   add_2x64b(right_x, right_x, right_dx_dy);                                    \
1058                                                                                \
1059   shr_narrow_2x64b(left_x_32.high, left_x, 32);                                \
1060   shr_narrow_2x64b(right_x_32.high, right_x, 32);                              \
1061                                                                                \
1062   add_2x64b(left_x, left_x, left_dx_dy);                                       \
1063   add_2x64b(right_x, right_x, right_dx_dy);                                    \
1064                                                                                \
1065   mov_narrow_4x32b(left_right_x_16.low, left_x_32);                            \
1066   mov_narrow_4x32b(left_right_x_16.high, right_x_32);                          \
1067                                                                                \
1068   setup_spans_increment_alternate_##alternate_active();                        \
1069   setup_spans_y_select_alternate_##alternate_active(direction);                \
1070   setup_spans_alternate_select_##alternate();                                  \
1071                                                                                \
1072   max_8x16b(left_right_x_16, left_right_x_16, left_edge);                      \
1073   min_8x16b(left_right_x_16, left_right_x_16, right_edge);                     \
1074                                                                                \
1075   sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low);  \
1076   add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07);               \
1077   and_4x16b(span_shift, left_right_x_16.high, c_0x07);                         \
1078   shl_variable_4x16b(span_shift, c_0xFFFE, span_shift);                        \
1079   shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3);                    \
1080                                                                                \
1081   u32 i;                                                                       \
1082   for(i = 0; i < 4; i++)                                                       \
1083   {                                                                            \
1084     span_edge_data[i].left_x = left_right_x_16.low.e[i];                       \
1085     span_edge_data[i].num_blocks = left_right_x_16.high.e[i];                  \
1086     span_edge_data[i].right_mask = span_shift.e[i];                            \
1087     span_edge_data[i].y = y_x4.e[i];                                           \
1088     setup_spans_debug_check(span_edge_data[i]);                                \
1089   }                                                                            \
1090                                                                                \
1091   span_edge_data += 4;                                                         \
1092                                                                                \
1093   setup_spans_adjust_y_##direction();                                          \
1094 }                                                                              \
1095
1096
1097 #define setup_spans_alternate_adjust_yes()                                     \
1098   edge_alt -= edge_dx_dy_alt * (s64)height_minor_a                             \
1099
1100 #define setup_spans_alternate_adjust_no()                                      \
1101
1102
1103 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1104   setup_spans_alternate_adjust_##alternate_active();                           \
1105   if(y_c > psx_gpu->viewport_end_y)                                            \
1106     height -= y_c - psx_gpu->viewport_end_y - 1;                               \
1107                                                                                \
1108   clip = psx_gpu->viewport_start_y - y_a;                                      \
1109   if(clip > 0)                                                                 \
1110   {                                                                            \
1111     height -= clip;                                                            \
1112     y_a += clip;                                                               \
1113     setup_spans_clip(increment, alternate_active);                             \
1114   }                                                                            \
1115                                                                                \
1116   setup_spans_prologue_b();                                                    \
1117                                                                                \
1118   if(height > 0)                                                               \
1119   {                                                                            \
1120     y_x4.e[0] = y_a;                                                           \
1121     y_x4.e[1] = y_a + 1;                                                       \
1122     y_x4.e[2] = y_a + 2;                                                       \
1123     y_x4.e[3] = y_a + 3;                                                       \
1124     setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
1125      right_index);                                                             \
1126                                                                                \
1127     psx_gpu->num_spans = height;                                               \
1128     do                                                                         \
1129     {                                                                          \
1130       setup_spans_set_x4(alternate, down, alternate_active);                   \
1131       height -= 4;                                                             \
1132     } while(height > 0);                                                       \
1133   }                                                                            \
1134
1135
1136 #define setup_spans_alternate_pre_increment_yes()                              \
1137   edge_alt += edge_dx_dy_alt                                                   \
1138
1139 #define setup_spans_alternate_pre_increment_no()                               \
1140
1141 #define setup_spans_up_decrement_height_yes()                                  \
1142   height--                                                                     \
1143
1144 #define setup_spans_up_decrement_height_no()                                   \
1145   {}                                                                           \
1146
1147 #define setup_spans_up(left_index, right_index, alternate, alternate_active)   \
1148   setup_spans_alternate_adjust_##alternate_active();                           \
1149   y_a--;                                                                       \
1150                                                                                \
1151   if(y_c < psx_gpu->viewport_start_y)                                          \
1152     height -= psx_gpu->viewport_start_y - y_c;                                 \
1153   else                                                                         \
1154     setup_spans_up_decrement_height_##alternate_active();                      \
1155                                                                                \
1156   clip = y_a - psx_gpu->viewport_end_y;                                        \
1157   if(clip > 0)                                                                 \
1158   {                                                                            \
1159     height -= clip;                                                            \
1160     y_a -= clip;                                                               \
1161     setup_spans_clip(decrement, alternate_active);                             \
1162   }                                                                            \
1163                                                                                \
1164   setup_spans_prologue_b();                                                    \
1165                                                                                \
1166   if(height > 0)                                                               \
1167   {                                                                            \
1168     y_x4.e[0] = y_a;                                                           \
1169     y_x4.e[1] = y_a - 1;                                                       \
1170     y_x4.e[2] = y_a - 2;                                                       \
1171     y_x4.e[3] = y_a - 3;                                                       \
1172     add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);                           \
1173     setup_spans_alternate_pre_increment_##alternate_active();                  \
1174     setup_spans_adjust_edges_alternate_##alternate_active(left_index,          \
1175      right_index);                                                             \
1176     setup_spans_adjust_interpolants_up();                                      \
1177                                                                                \
1178     psx_gpu->num_spans = height;                                               \
1179     while(height > 0)                                                          \
1180     {                                                                          \
1181       setup_spans_set_x4(alternate, up, alternate_active);                     \
1182       height -= 4;                                                             \
1183     }                                                                          \
1184   }                                                                            \
1185
1186 #define index_left  0
1187 #define index_right 1
1188
1189 #define setup_spans_up_up(minor, major)                                        \
1190   setup_spans_prologue(yes);                                                   \
1191   s32 height_minor_a = y_a - y_b;                                              \
1192   s32 height_minor_b = y_b - y_c;                                              \
1193   s32 height = y_a - y_c;                                                      \
1194                                                                                \
1195   dup_2x32b(x_starts, x_a);                                                    \
1196   x_ends.e[0] = x_c;                                                           \
1197   x_ends.e[1] = x_b;                                                           \
1198                                                                                \
1199   compute_edge_delta_x3(x_b, height, height_minor_a);                          \
1200   setup_spans_up(index_##major, index_##minor, minor, yes)                     \
1201
1202
1203 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1204  vertex_struct *v_b, vertex_struct *v_c);
1205 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1206  vertex_struct *v_b, vertex_struct *v_c);
1207 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1208  vertex_struct *v_b, vertex_struct *v_c);
1209 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1210  vertex_struct *v_b, vertex_struct *v_c);
1211 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1212  vertex_struct *v_b, vertex_struct *v_c);
1213 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1214  vertex_struct *v_b, vertex_struct *v_c);
1215 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1216  vertex_struct *v_b, vertex_struct *v_c);
1217 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1218  vertex_struct *v_b, vertex_struct *v_c);
1219 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1220  vertex_struct *v_b, vertex_struct *v_c);
1221
1222
1223 #ifndef NEON_BUILD
1224
1225 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1226  vertex_struct *v_b, vertex_struct *v_c)
1227 {
1228   setup_spans_up_up(left, right);
1229 }
1230
1231 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1232  vertex_struct *v_b, vertex_struct *v_c)
1233 {
1234   setup_spans_up_up(right, left);
1235 }
1236
1237 #define setup_spans_down_down(minor, major)                                    \
1238   setup_spans_prologue(yes);                                                   \
1239   s32 height_minor_a = y_b - y_a;                                              \
1240   s32 height_minor_b = y_c - y_b;                                              \
1241   s32 height = y_c - y_a;                                                      \
1242                                                                                \
1243   dup_2x32b(x_starts, x_a);                                                    \
1244   x_ends.e[0] = x_c;                                                           \
1245   x_ends.e[1] = x_b;                                                           \
1246                                                                                \
1247   compute_edge_delta_x3(x_b, height, height_minor_a);                          \
1248   setup_spans_down(index_##major, index_##minor, minor, yes)                   \
1249
1250 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1251  vertex_struct *v_b, vertex_struct *v_c)
1252 {
1253   setup_spans_down_down(left, right);
1254 }
1255
1256 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1257  vertex_struct *v_b, vertex_struct *v_c)
1258 {
1259   setup_spans_down_down(right, left);
1260 }
1261
1262 #define setup_spans_up_flat()                                                  \
1263   s32 height = y_a - y_c;                                                      \
1264                                                                                \
1265   flat_triangles++;                                                            \
1266   compute_edge_delta_x2();                                                     \
1267   setup_spans_up(index_left, index_right, none, no)                            \
1268
1269 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1270  vertex_struct *v_b, vertex_struct *v_c)
1271 {
1272   setup_spans_prologue(no);
1273   x_starts.e[0] = x_a;
1274   x_starts.e[1] = x_b;
1275   dup_2x32b(x_ends, x_c);
1276
1277   setup_spans_up_flat();
1278 }
1279
1280 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1281  vertex_struct *v_b, vertex_struct *v_c)
1282 {
1283   setup_spans_prologue(no);
1284   dup_2x32b(x_starts, x_a);
1285   x_ends.e[0] = x_b;
1286   x_ends.e[1] = x_c;
1287
1288   setup_spans_up_flat();
1289 }
1290
1291 #define setup_spans_down_flat()                                                \
1292   s32 height = y_c - y_a;                                                      \
1293                                                                                \
1294   flat_triangles++;                                                            \
1295   compute_edge_delta_x2();                                                     \
1296   setup_spans_down(index_left, index_right, none, no)                          \
1297
1298 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1299  vertex_struct *v_b, vertex_struct *v_c)
1300 {
1301   setup_spans_prologue(no);
1302   x_starts.e[0] = x_a;
1303   x_starts.e[1] = x_b;
1304   dup_2x32b(x_ends, x_c);
1305
1306   setup_spans_down_flat();
1307 }
1308
1309 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1310  vertex_struct *v_b, vertex_struct *v_c)
1311 {
1312   setup_spans_prologue(no);
1313   dup_2x32b(x_starts, x_a);
1314   x_ends.e[0] = x_b;
1315   x_ends.e[1] = x_c;
1316
1317   setup_spans_down_flat();
1318 }
1319
1320 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1321  vertex_struct *v_b, vertex_struct *v_c)
1322 {
1323   setup_spans_prologue(no);
1324
1325   s32 y_b = v_b->y;
1326   s64 edge_alt;
1327   s32 edge_dx_dy_alt;
1328   u32 edge_shift_alt;
1329
1330   s32 middle_y = y_a;
1331   s32 height_minor_a = y_a - y_b;
1332   s32 height_minor_b = y_c - y_a;
1333   s32 height_major = y_c - y_b;
1334
1335   vec_2x64s edges_xy_b;
1336   vec_2x32s edges_dx_dy_b;
1337   vec_2x32u edge_shifts_b;
1338
1339   vec_2x32s height_increment;
1340
1341   x_starts.e[0] = x_a;
1342   x_starts.e[1] = x_c;
1343   dup_2x32b(x_ends, x_b);
1344
1345   compute_edge_delta_x3(x_a, height_minor_a, height_major);
1346
1347   height_increment.e[0] = 0;
1348   height_increment.e[1] = height_minor_b;
1349
1350   mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1351
1352   edges_xy_b.e[0] = edge_alt;
1353   edges_xy_b.e[1] = edges_xy.e[1];
1354
1355   edge_shifts_b = edge_shifts;
1356   edge_shifts_b.e[0] = edge_shift_alt;
1357
1358   neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1359   edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1360   
1361   y_a--;
1362
1363   if(y_b < psx_gpu->viewport_start_y)
1364     height_minor_a -= psx_gpu->viewport_start_y - y_b;
1365
1366   clip = y_a - psx_gpu->viewport_end_y;
1367   if(clip > 0)
1368   {
1369     height_minor_a -= clip;
1370     y_a -= clip;
1371     setup_spans_clip(decrement, no);
1372   }
1373
1374   setup_spans_prologue_b();
1375
1376   if(height_minor_a > 0)
1377   {
1378     y_x4.e[0] = y_a;
1379     y_x4.e[1] = y_a - 1;
1380     y_x4.e[2] = y_a - 2;
1381     y_x4.e[3] = y_a - 3;
1382     add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1383     setup_spans_adjust_edges_alternate_no(index_left, index_right);
1384     setup_spans_adjust_interpolants_up();
1385
1386     psx_gpu->num_spans = height_minor_a;
1387     while(height_minor_a > 0)
1388     {
1389       setup_spans_set_x4(none, up, no);
1390       height_minor_a -= 4;
1391     }
1392
1393     span_edge_data += height_minor_a;
1394     span_uvrg_offset += height_minor_a;
1395     span_b_offset += height_minor_a;
1396   }
1397   
1398   edges_xy = edges_xy_b;
1399   edges_dx_dy = edges_dx_dy_b;
1400   edge_shifts = edge_shifts_b;
1401
1402   uvrg = psx_gpu->uvrg;
1403   b = psx_gpu->b;
1404
1405   y_a = middle_y;
1406
1407   if(y_c > psx_gpu->viewport_end_y)
1408     height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1409
1410   clip = psx_gpu->viewport_start_y - y_a;
1411   if(clip > 0)
1412   {
1413     height_minor_b -= clip;
1414     y_a += clip;
1415     setup_spans_clip(increment, no);
1416   }
1417
1418   if(height_minor_b > 0)
1419   {
1420     y_x4.e[0] = y_a;
1421     y_x4.e[1] = y_a + 1;
1422     y_x4.e[2] = y_a + 2;
1423     y_x4.e[3] = y_a + 3;
1424     setup_spans_adjust_edges_alternate_no(index_left, index_right);
1425
1426     psx_gpu->num_spans += height_minor_b;
1427     do
1428     {
1429       setup_spans_set_x4(none, down, no);
1430       height_minor_b -= 4;
1431     } while(height_minor_b > 0);
1432   }
1433
1434   left_split_triangles++;
1435 }
1436
1437 #endif
1438
1439
1440 #define dither_table_entry_normal(value)                                       \
1441   (value)                                                                      \
1442
1443
1444 #define setup_blocks_load_msb_mask_indirect()                                  \
1445
1446 #define setup_blocks_load_msb_mask_direct()                                    \
1447   vec_8x16u msb_mask;                                                          \
1448   dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
1449
1450
1451 #define setup_blocks_variables_shaded_textured(target)                         \
1452   vec_4x32u u_block;                                                           \
1453   vec_4x32u v_block;                                                           \
1454   vec_4x32u r_block;                                                           \
1455   vec_4x32u g_block;                                                           \
1456   vec_4x32u b_block;                                                           \
1457   vec_4x32u uvrg_dx = psx_gpu->uvrg_dx;                                        \
1458   vec_4x32u uvrg_dx4;                                                          \
1459   vec_4x32u uvrg_dx8;                                                          \
1460   vec_4x32u uvrg;                                                              \
1461   u32 b_dx = psx_gpu->b_block_span.e[1];                                       \
1462   u32 b_dx4 = b_dx << 2;                                                       \
1463   u32 b_dx8 = b_dx << 3;                                                       \
1464   u32 b;                                                                       \
1465                                                                                \
1466   vec_16x8u texture_mask;                                                      \
1467   shl_4x32b(uvrg_dx4, uvrg_dx, 2);                                             \
1468   shl_4x32b(uvrg_dx8, uvrg_dx, 3);                                             \
1469   dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
1470   dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
1471
1472 #define setup_blocks_variables_shaded_untextured(target)                       \
1473   vec_4x32u r_block;                                                           \
1474   vec_4x32u g_block;                                                           \
1475   vec_4x32u b_block;                                                           \
1476   vec_4x32u rgb_dx;                                                            \
1477   vec_4x32u rgb_dx4;                                                           \
1478   vec_4x32u rgb_dx8;                                                           \
1479   vec_4x32u rgb;                                                               \
1480                                                                                \
1481   vec_8x8u d64_0x07;                                                           \
1482   vec_8x8u d64_1;                                                              \
1483   vec_8x8u d64_4;                                                              \
1484   vec_8x8u d64_128;                                                            \
1485                                                                                \
1486   dup_8x8b(d64_0x07, 0x07);                                                    \
1487   dup_8x8b(d64_1, 1);                                                          \
1488   dup_8x8b(d64_4, 4);                                                          \
1489   dup_8x8b(d64_128, 128);                                                      \
1490                                                                                \
1491   rgb_dx.low = psx_gpu->uvrg_dx.high;                                          \
1492   rgb_dx.e[2] = psx_gpu->b_block_span.e[1];                                    \
1493   shl_4x32b(rgb_dx4, rgb_dx, 2);                                               \
1494   shl_4x32b(rgb_dx8, rgb_dx, 3)                                                \
1495
1496 #define setup_blocks_variables_unshaded_textured(target)                       \
1497   vec_4x32u u_block;                                                           \
1498   vec_4x32u v_block;                                                           \
1499   vec_2x32u uv_dx = psx_gpu->uvrg_dx.low;                                      \
1500   vec_2x32u uv_dx4;                                                            \
1501   vec_2x32u uv_dx8;                                                            \
1502   vec_2x32u uv = psx_gpu->uvrg.low;                                            \
1503                                                                                \
1504   vec_16x8u texture_mask;                                                      \
1505   shl_2x32b(uv_dx4, uv_dx, 2);                                                 \
1506   shl_2x32b(uv_dx8, uv_dx, 3);                                                 \
1507   dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width);                     \
1508   dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height)                    \
1509
1510
1511 #define setup_blocks_variables_unshaded_untextured_direct()                    \
1512   or_8x16b(colors, colors, msb_mask)                                           \
1513
1514 #define setup_blocks_variables_unshaded_untextured_indirect()                  \
1515
1516 #define setup_blocks_variables_unshaded_untextured(target)                     \
1517   u32 color = psx_gpu->triangle_color;                                         \
1518   vec_8x16u colors;                                                            \
1519                                                                                \
1520   u32 color_r = color & 0xFF;                                                  \
1521   u32 color_g = (color >> 8) & 0xFF;                                           \
1522   u32 color_b = (color >> 16) & 0xFF;                                          \
1523                                                                                \
1524   color = (color_r >> 3) | ((color_g >> 3) << 5) |                             \
1525    ((color_b >> 3) << 10);                                                     \
1526   dup_8x16b(colors, color);                                                    \
1527   setup_blocks_variables_unshaded_untextured_##target()                        \
1528
1529 #define setup_blocks_span_initialize_dithered_textured()                       \
1530   vec_8x16u dither_offsets;                                                    \
1531   shl_long_8x8b(dither_offsets, dither_offsets_short, 4)                       \
1532
1533 #define setup_blocks_span_initialize_dithered_untextured()                     \
1534   vec_8x8u dither_offsets;                                                     \
1535   add_8x8b(dither_offsets, dither_offsets_short, d64_4)                        \
1536
1537 #define setup_blocks_span_initialize_dithered(texturing)                       \
1538   u32 dither_row = psx_gpu->dither_table[y & 0x3];                             \
1539   u32 dither_shift = (span_edge_data->left_x & 0x3) * 8;                       \
1540   vec_8x8s dither_offsets_short;                                               \
1541                                                                                \
1542   dither_row =                                                                 \
1543    (dither_row >> dither_shift) | (dither_row << (32 - dither_shift));         \
1544   dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row);         \
1545   setup_blocks_span_initialize_dithered_##texturing()                          \
1546
1547 #define setup_blocks_span_initialize_undithered(texturing)                     \
1548
1549
1550 #define setup_blocks_span_initialize_shaded_textured()                         \
1551 {                                                                              \
1552   vec_4x32u block_span;                                                        \
1553   u32 offset = span_edge_data->left_x;                                         \
1554                                                                                \
1555   uvrg = *span_uvrg_offset;                                                    \
1556   mla_scalar_4x32b(uvrg, uvrg_dx, offset);                                     \
1557   b = *span_b_offset;                                                          \
1558   b += b_dx * offset;                                                          \
1559                                                                                \
1560   dup_4x32b(u_block, uvrg.e[0]);                                               \
1561   dup_4x32b(v_block, uvrg.e[1]);                                               \
1562   dup_4x32b(r_block, uvrg.e[2]);                                               \
1563   dup_4x32b(g_block, uvrg.e[3]);                                               \
1564   dup_4x32b(b_block, b);                                                       \
1565                                                                                \
1566   block_span = psx_gpu->u_block_span;                                          \
1567   add_4x32b(u_block, u_block, block_span);                                     \
1568   block_span = psx_gpu->v_block_span;                                          \
1569   add_4x32b(v_block, v_block, block_span);                                     \
1570   block_span = psx_gpu->r_block_span;                                          \
1571   add_4x32b(r_block, r_block, block_span);                                     \
1572   block_span = psx_gpu->g_block_span;                                          \
1573   add_4x32b(g_block, g_block, block_span);                                     \
1574   block_span = psx_gpu->b_block_span;                                          \
1575   add_4x32b(b_block, b_block, block_span);                                     \
1576 }
1577   
1578 #define setup_blocks_span_initialize_shaded_untextured()                       \
1579 {                                                                              \
1580   vec_4x32u block_span;                                                        \
1581   u32 offset = span_edge_data->left_x;                                         \
1582                                                                                \
1583   rgb.low = span_uvrg_offset->high;                                            \
1584   rgb.high.e[0] = *span_b_offset;                                              \
1585   mla_scalar_4x32b(rgb, rgb_dx, offset);                                       \
1586                                                                                \
1587   dup_4x32b(r_block, rgb.e[0]);                                                \
1588   dup_4x32b(g_block, rgb.e[1]);                                                \
1589   dup_4x32b(b_block, rgb.e[2]);                                                \
1590                                                                                \
1591   block_span = psx_gpu->r_block_span;                                          \
1592   add_4x32b(r_block, r_block, block_span);                                     \
1593   block_span = psx_gpu->g_block_span;                                          \
1594   add_4x32b(g_block, g_block, block_span);                                     \
1595   block_span = psx_gpu->b_block_span;                                          \
1596   add_4x32b(b_block, b_block, block_span);                                     \
1597 }                                                                              \
1598   
1599 #define setup_blocks_span_initialize_unshaded_textured()                       \
1600 {                                                                              \
1601   vec_4x32u block_span;                                                        \
1602   u32 offset = span_edge_data->left_x;                                         \
1603                                                                                \
1604   uv = span_uvrg_offset->low;                                                  \
1605   mla_scalar_2x32b(uv, uv_dx, offset);                                         \
1606                                                                                \
1607   dup_4x32b(u_block, uv.e[0]);                                                 \
1608   dup_4x32b(v_block, uv.e[1]);                                                 \
1609                                                                                \
1610   block_span = psx_gpu->u_block_span;                                          \
1611   add_4x32b(u_block, u_block, block_span);                                     \
1612   block_span = psx_gpu->v_block_span;                                          \
1613   add_4x32b(v_block, v_block, block_span);                                     \
1614 }                                                                              \
1615
1616 #define setup_blocks_span_initialize_unshaded_untextured()                     \
1617
1618
1619 #define setup_blocks_texture_swizzled()                                        \
1620 {                                                                              \
1621   vec_8x8u u_saved = u;                                                        \
1622   sli_8x8b(u, v, 4);                                                           \
1623   sri_8x8b(v, u_saved, 4);                                                     \
1624 }                                                                              \
1625
1626 #define setup_blocks_texture_unswizzled()                                      \
1627
1628 #define setup_blocks_store_shaded_textured(swizzling, dithering, target,       \
1629  edge_type)                                                                    \
1630 {                                                                              \
1631   vec_8x16u u_whole;                                                           \
1632   vec_8x16u v_whole;                                                           \
1633   vec_8x16u r_whole;                                                           \
1634   vec_8x16u g_whole;                                                           \
1635   vec_8x16u b_whole;                                                           \
1636                                                                                \
1637   vec_8x8u u;                                                                  \
1638   vec_8x8u v;                                                                  \
1639   vec_8x8u r;                                                                  \
1640   vec_8x8u g;                                                                  \
1641   vec_8x8u b;                                                                  \
1642   vec_8x16u uv;                                                                \
1643                                                                                \
1644   vec_4x32u dx4;                                                               \
1645   vec_4x32u dx8;                                                               \
1646                                                                                \
1647   shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
1648   shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
1649   shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
1650   shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
1651   shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
1652                                                                                \
1653   dup_4x32b(dx4, uvrg_dx4.e[0]);                                               \
1654   add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
1655   dup_4x32b(dx4, uvrg_dx4.e[1]);                                               \
1656   add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
1657   dup_4x32b(dx4, uvrg_dx4.e[2]);                                               \
1658   add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
1659   dup_4x32b(dx4, uvrg_dx4.e[3]);                                               \
1660   add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
1661   dup_4x32b(dx4, b_dx4);                                                       \
1662   add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
1663                                                                                \
1664   mov_narrow_8x16b(u, u_whole);                                                \
1665   mov_narrow_8x16b(v, v_whole);                                                \
1666   mov_narrow_8x16b(r, r_whole);                                                \
1667   mov_narrow_8x16b(g, g_whole);                                                \
1668   mov_narrow_8x16b(b, b_whole);                                                \
1669                                                                                \
1670   dup_4x32b(dx8, uvrg_dx8.e[0]);                                               \
1671   add_4x32b(u_block, u_block, dx8);                                            \
1672   dup_4x32b(dx8, uvrg_dx8.e[1]);                                               \
1673   add_4x32b(v_block, v_block, dx8);                                            \
1674   dup_4x32b(dx8, uvrg_dx8.e[2]);                                               \
1675   add_4x32b(r_block, r_block, dx8);                                            \
1676   dup_4x32b(dx8, uvrg_dx8.e[3]);                                               \
1677   add_4x32b(g_block, g_block, dx8);                                            \
1678   dup_4x32b(dx8, b_dx8);                                                       \
1679   add_4x32b(b_block, b_block, dx8);                                            \
1680                                                                                \
1681   and_8x8b(u, u, texture_mask.low);                                            \
1682   and_8x8b(v, v, texture_mask.high);                                           \
1683   setup_blocks_texture_##swizzling();                                          \
1684                                                                                \
1685   zip_8x16b(uv, u, v);                                                         \
1686   block->uv = uv;                                                              \
1687   block->r = r;                                                                \
1688   block->g = g;                                                                \
1689   block->b = b;                                                                \
1690   block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
1691   block->fb_ptr = fb_ptr;                                                      \
1692 }                                                                              \
1693
1694 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target,     \
1695  edge_type)                                                                    \
1696 {                                                                              \
1697   vec_8x16u u_whole;                                                           \
1698   vec_8x16u v_whole;                                                           \
1699                                                                                \
1700   vec_8x8u u;                                                                  \
1701   vec_8x8u v;                                                                  \
1702   vec_8x16u uv;                                                                \
1703                                                                                \
1704   vec_4x32u dx4;                                                               \
1705   vec_4x32u dx8;                                                               \
1706                                                                                \
1707   shr_narrow_4x32b(u_whole.low, u_block, 16);                                  \
1708   shr_narrow_4x32b(v_whole.low, v_block, 16);                                  \
1709                                                                                \
1710   dup_4x32b(dx4, uv_dx4.e[0]);                                                 \
1711   add_high_narrow_4x32b(u_whole.high, u_block, dx4);                           \
1712   dup_4x32b(dx4, uv_dx4.e[1]);                                                 \
1713   add_high_narrow_4x32b(v_whole.high, v_block, dx4);                           \
1714                                                                                \
1715   mov_narrow_8x16b(u, u_whole);                                                \
1716   mov_narrow_8x16b(v, v_whole);                                                \
1717                                                                                \
1718   dup_4x32b(dx8, uv_dx8.e[0]);                                                 \
1719   add_4x32b(u_block, u_block, dx8);                                            \
1720   dup_4x32b(dx8, uv_dx8.e[1]);                                                 \
1721   add_4x32b(v_block, v_block, dx8);                                            \
1722                                                                                \
1723   and_8x8b(u, u, texture_mask.low);                                            \
1724   and_8x8b(v, v, texture_mask.high);                                           \
1725   setup_blocks_texture_##swizzling();                                          \
1726                                                                                \
1727   zip_8x16b(uv, u, v);                                                         \
1728   block->uv = uv;                                                              \
1729   block->dither_offsets = vector_cast(vec_8x16u, dither_offsets);              \
1730   block->fb_ptr = fb_ptr;                                                      \
1731 }                                                                              \
1732
1733 #define setup_blocks_store_shaded_untextured_dithered()                        \
1734   addq_8x8b(r, r, dither_offsets);                                             \
1735   addq_8x8b(g, g, dither_offsets);                                             \
1736   addq_8x8b(b, b, dither_offsets);                                             \
1737                                                                                \
1738   subq_8x8b(r, r, d64_4);                                                      \
1739   subq_8x8b(g, g, d64_4);                                                      \
1740   subq_8x8b(b, b, d64_4)                                                       \
1741
1742 #define setup_blocks_store_shaded_untextured_undithered()                      \
1743   
1744
1745 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels)            \
1746   block->pixels = _pixels;                                                     \
1747   block->fb_ptr = fb_ptr                                                       \
1748
1749 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels)            \
1750   block->pixels = _pixels;                                                     \
1751   block->fb_ptr = fb_ptr                                                       \
1752
1753 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect()            \
1754   mul_long_8x8b(pixels, r, d64_1)                                              \
1755
1756
1757 #define setup_blocks_store_untextured_pixels_direct_full(_pixels)              \
1758   store_8x16b(_pixels, fb_ptr)                                                 \
1759
1760 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels)              \
1761 {                                                                              \
1762   vec_8x16u fb_pixels;                                                         \
1763   vec_8x16u draw_mask;                                                         \
1764   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
1765                                                                                \
1766   load_8x16b(fb_pixels, fb_ptr);                                               \
1767   dup_8x16b(draw_mask, span_edge_data->right_mask);                            \
1768   tst_8x16b(draw_mask, draw_mask, test_mask);                                  \
1769   bif_8x16b(fb_pixels, _pixels, draw_mask);                                    \
1770   store_8x16b(fb_pixels, fb_ptr);                                              \
1771 }                                                                              \
1772
1773 #define setup_blocks_store_shaded_untextured_seed_pixels_direct()              \
1774   pixels = msb_mask;                                                           \
1775   mla_long_8x8b(pixels, r, d64_1)                                              \
1776
1777
1778 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target,     \
1779  edge_type)                                                                    \
1780 {                                                                              \
1781   vec_8x16u r_whole;                                                           \
1782   vec_8x16u g_whole;                                                           \
1783   vec_8x16u b_whole;                                                           \
1784                                                                                \
1785   vec_8x8u r;                                                                  \
1786   vec_8x8u g;                                                                  \
1787   vec_8x8u b;                                                                  \
1788                                                                                \
1789   vec_4x32u dx4;                                                               \
1790   vec_4x32u dx8;                                                               \
1791                                                                                \
1792   vec_8x16u pixels;                                                            \
1793                                                                                \
1794   shr_narrow_4x32b(r_whole.low, r_block, 16);                                  \
1795   shr_narrow_4x32b(g_whole.low, g_block, 16);                                  \
1796   shr_narrow_4x32b(b_whole.low, b_block, 16);                                  \
1797                                                                                \
1798   dup_4x32b(dx4, rgb_dx4.e[0]);                                                \
1799   add_high_narrow_4x32b(r_whole.high, r_block, dx4);                           \
1800   dup_4x32b(dx4, rgb_dx4.e[1]);                                                \
1801   add_high_narrow_4x32b(g_whole.high, g_block, dx4);                           \
1802   dup_4x32b(dx4, rgb_dx4.e[2]);                                                \
1803   add_high_narrow_4x32b(b_whole.high, b_block, dx4);                           \
1804                                                                                \
1805   mov_narrow_8x16b(r, r_whole);                                                \
1806   mov_narrow_8x16b(g, g_whole);                                                \
1807   mov_narrow_8x16b(b, b_whole);                                                \
1808                                                                                \
1809   dup_4x32b(dx8, rgb_dx8.e[0]);                                                \
1810   add_4x32b(r_block, r_block, dx8);                                            \
1811   dup_4x32b(dx8, rgb_dx8.e[1]);                                                \
1812   add_4x32b(g_block, g_block, dx8);                                            \
1813   dup_4x32b(dx8, rgb_dx8.e[2]);                                                \
1814   add_4x32b(b_block, b_block, dx8);                                            \
1815                                                                                \
1816   setup_blocks_store_shaded_untextured_##dithering();                          \
1817                                                                                \
1818   shr_8x8b(r, r, 3);                                                           \
1819   bic_8x8b(g, g, d64_0x07);                                                    \
1820   bic_8x8b(b, b, d64_0x07);                                                    \
1821                                                                                \
1822   setup_blocks_store_shaded_untextured_seed_pixels_##target();                 \
1823   mla_long_8x8b(pixels, g, d64_4);                                             \
1824   mla_long_8x8b(pixels, b, d64_128)                                            \
1825                                                                                \
1826   setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels);         \
1827 }                                                                              \
1828
1829 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target,   \
1830  edge_type)                                                                    \
1831   setup_blocks_store_untextured_pixels_##target##_##edge_type(colors)          \
1832
1833
1834 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits)           \
1835   (_block)->draw_mask_bits = bits                                              \
1836
1837 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits)         \
1838 {                                                                              \
1839   vec_8x16u bits_mask;                                                         \
1840   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
1841   dup_8x16b(bits_mask, bits);                                                  \
1842   tst_8x16b(bits_mask, bits_mask, test_mask);                                  \
1843   (_block)->draw_mask = bits_mask;                                             \
1844 }                                                                              \
1845
1846 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits)           \
1847
1848
1849 #define setup_blocks_add_blocks_indirect()                                     \
1850   num_blocks += span_num_blocks;                                               \
1851                                                                                \
1852   if(num_blocks > MAX_BLOCKS)                                                  \
1853   {                                                                            \
1854     psx_gpu->num_blocks = num_blocks - span_num_blocks;                        \
1855     flush_render_block_buffer(psx_gpu);                                        \
1856     num_blocks = span_num_blocks;                                              \
1857     block = psx_gpu->blocks;                                                   \
1858   }                                                                            \
1859
1860 #define setup_blocks_add_blocks_direct()                                       \
1861   texel_blocks_untextured += span_num_blocks;                                  \
1862   span_pixel_blocks += span_num_blocks                                         \
1863
1864
1865 #define setup_blocks_builder(shading, texturing, dithering, sw, target)        \
1866 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target(     \
1867  psx_gpu_struct *psx_gpu)                                                      \
1868 {                                                                              \
1869   setup_blocks_load_msb_mask_##target();                                       \
1870   setup_blocks_variables_##shading##_##texturing(target);                      \
1871                                                                                \
1872   edge_data_struct *span_edge_data = psx_gpu->span_edge_data;                  \
1873   vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset;                     \
1874   u32 *span_b_offset = psx_gpu->span_b_offset;                                 \
1875                                                                                \
1876   block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks;                 \
1877                                                                                \
1878   u32 num_spans = psx_gpu->num_spans;                                          \
1879                                                                                \
1880   u16 *fb_ptr;                                                                 \
1881   u32 y;                                                                       \
1882                                                                                \
1883   u32 num_blocks = psx_gpu->num_blocks;                                        \
1884   u32 span_num_blocks;                                                         \
1885                                                                                \
1886   while(num_spans)                                                             \
1887   {                                                                            \
1888     span_num_blocks = span_edge_data->num_blocks;                              \
1889     if(span_num_blocks)                                                        \
1890     {                                                                          \
1891       y = span_edge_data->y;                                                   \
1892       fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024);    \
1893                                                                                \
1894       setup_blocks_span_initialize_##shading##_##texturing();                  \
1895       setup_blocks_span_initialize_##dithering(texturing);                     \
1896                                                                                \
1897       setup_blocks_add_blocks_##target();                                      \
1898                                                                                \
1899       s32 pixel_span = span_num_blocks * 8;                                    \
1900       pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF);     \
1901       span_pixels += pixel_span;                                               \
1902                                                                                \
1903       span_num_blocks--;                                                       \
1904       while(span_num_blocks)                                                   \
1905       {                                                                        \
1906         setup_blocks_store_##shading##_##texturing(sw, dithering, target,      \
1907          full);                                                                \
1908         setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00);      \
1909                                                                                \
1910         fb_ptr += 8;                                                           \
1911         block++;                                                               \
1912         span_num_blocks--;                                                     \
1913       }                                                                        \
1914                                                                                \
1915       setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1916       setup_blocks_store_draw_mask_##texturing##_##target(block,               \
1917        span_edge_data->right_mask);                                            \
1918                                                                                \
1919       block++;                                                                 \
1920     }                                                                          \
1921     else                                                                       \
1922     {                                                                          \
1923       zero_block_spans++;                                                      \
1924     }                                                                          \
1925                                                                                \
1926     num_spans--;                                                               \
1927     span_edge_data++;                                                          \
1928     span_uvrg_offset++;                                                        \
1929     span_b_offset++;                                                           \
1930   }                                                                            \
1931                                                                                \
1932   psx_gpu->num_blocks = num_blocks;                                            \
1933 }                                                                              \
1934
1935 void setup_blocks_shaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1936  *psx_gpu);
1937
1938 void setup_blocks_shaded_untextured_dithered_unswizzled_indirect(psx_gpu_struct
1939  *psx_gpu);
1940 void setup_blocks_shaded_untextured_undithered_unswizzled_indirect(
1941  psx_gpu_struct *psx_gpu);
1942 void setup_blocks_shaded_untextured_dithered_unswizzled_direct(psx_gpu_struct
1943  *psx_gpu);
1944 void setup_blocks_shaded_untextured_undithered_unswizzled_direct(
1945  psx_gpu_struct *psx_gpu);
1946
1947 void setup_blocks_unshaded_textured_dithered_unswizzled_indirect(psx_gpu_struct
1948  *psx_gpu);
1949 void setup_blocks_unshaded_untextured_undithered_unswizzled_indirect(
1950  psx_gpu_struct *psx_gpu);
1951 void setup_blocks_unshaded_untextured_undithered_unswizzled_direct(
1952  psx_gpu_struct *psx_gpu);
1953
1954 void setup_blocks_shaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1955  *psx_gpu);
1956 void setup_blocks_unshaded_textured_dithered_swizzled_indirect(psx_gpu_struct
1957  *psx_gpu);
1958
1959
1960 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1961
1962 #ifndef NEON_BUILD
1963
1964 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1965 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1966
1967 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1968 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1969
1970 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1971 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1972 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1973 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1974
1975 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1976 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1977
1978 #endif
1979
1980 void texture_blocks_untextured(psx_gpu_struct *psx_gpu);
1981 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu);
1982 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu);
1983 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu);
1984
1985 #ifndef NEON_BUILD
1986
1987 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1988 {
1989   if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1990     texel_blocks_untextured += psx_gpu->num_blocks;
1991 }
1992
1993 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1994 {
1995   block_struct *block = psx_gpu->blocks;
1996   u32 num_blocks = psx_gpu->num_blocks;
1997   texel_blocks_4bpp += num_blocks;
1998
1999   vec_8x8u texels_low;
2000   vec_8x8u texels_high;
2001   vec_8x8u texels;
2002   vec_8x16u pixels;
2003
2004   vec_8x16u clut_a;
2005   vec_8x16u clut_b;
2006   vec_16x8u clut_low;
2007   vec_16x8u clut_high;
2008
2009   u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2010   u16 *clut_ptr = psx_gpu->clut_ptr;
2011
2012   // Can be done with one deinterleaving load on NEON
2013   load_8x16b(clut_a, clut_ptr);
2014   load_8x16b(clut_b, clut_ptr + 8);
2015   unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
2016
2017   if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
2018     update_texture_4bpp_cache(psx_gpu);
2019
2020   while(num_blocks)
2021   {
2022     texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
2023     texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
2024     texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
2025     texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
2026     texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
2027     texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
2028     texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
2029     texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
2030
2031     tbl_16(texels_low, texels, clut_low);
2032     tbl_16(texels_high, texels, clut_high);
2033
2034     // Can be done with an interleaving store on NEON
2035     zip_8x16b(pixels, texels_low, texels_high);
2036
2037     block->texels = pixels;
2038
2039     num_blocks--;
2040     block++;
2041   }
2042 }
2043
2044 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2045 {
2046   block_struct *block = psx_gpu->blocks;
2047   u32 num_blocks = psx_gpu->num_blocks;
2048
2049   texel_blocks_8bpp += num_blocks;
2050
2051   if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2052     update_texture_8bpp_cache(psx_gpu);
2053
2054   vec_8x16u texels;
2055   u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2056
2057   u32 texel;
2058   u32 offset;
2059   u32 i;
2060
2061   while(num_blocks)
2062   {
2063     for(i = 0; i < 8; i++)
2064     {
2065       offset = block->uv.e[i];
2066
2067       texel = texture_ptr_8bpp[offset];
2068       texels.e[i] = psx_gpu->clut_ptr[texel];
2069     }
2070
2071     block->texels = texels;
2072
2073     num_blocks--;
2074     block++;
2075   }
2076 }
2077
2078 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2079 {
2080   block_struct *block = psx_gpu->blocks;
2081   u32 num_blocks = psx_gpu->num_blocks;
2082
2083   texel_blocks_16bpp += num_blocks;
2084
2085   vec_8x16u texels;
2086
2087   u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2088   u32 offset;
2089   u32 i;
2090
2091   while(num_blocks)
2092   {
2093     for(i = 0; i < 8; i++)
2094     {
2095       offset = block->uv.e[i];
2096       offset += ((offset & 0xFF00) * 3);
2097
2098       texels.e[i] = texture_ptr_16bpp[offset];
2099     }
2100
2101     block->texels = texels;
2102
2103     num_blocks--;
2104     block++;
2105   }
2106 }
2107
2108 #endif
2109
2110
2111 #define shade_blocks_load_msb_mask_indirect()                                  \
2112
2113 #define shade_blocks_load_msb_mask_direct()                                    \
2114   vec_8x16u msb_mask;                                                          \
2115   dup_8x16b(msb_mask, psx_gpu->mask_msb);                                      \
2116
2117 #define shade_blocks_store_indirect(_draw_mask, _pixels)                       \
2118   block->draw_mask = _draw_mask;                                               \
2119   block->pixels = _pixels                                                      \
2120
2121 #define shade_blocks_store_direct(_draw_mask, _pixels)                         \
2122 {                                                                              \
2123   vec_8x16u fb_pixels;                                                         \
2124   or_8x16b(_pixels, _pixels, msb_mask);                                        \
2125   load_8x16b(fb_pixels, block->fb_ptr);                                        \
2126   bif_8x16b(fb_pixels, _pixels, _draw_mask);                                   \
2127   store_8x16b(fb_pixels, block->fb_ptr);                                       \
2128 }                                                                              \
2129
2130
2131 #define shade_blocks_textured_false_modulated_check_dithered(target)           \
2132   if(psx_gpu->triangle_color == 0x808080)                                      \
2133   {                                                                            \
2134     false_modulated_blocks += num_blocks;                                      \
2135   }                                                                            \
2136
2137 #define shade_blocks_textured_false_modulated_check_undithered(target)         \
2138   if(psx_gpu->triangle_color == 0x808080)                                      \
2139   {                                                                            \
2140                                                                                \
2141     shade_blocks_textured_unmodulated_##target(psx_gpu);                       \
2142     false_modulated_blocks += num_blocks;                                      \
2143     return;                                                                    \
2144   }                                                                            \
2145
2146
2147 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering,       \
2148  target)                                                                       \
2149
2150 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering,     \
2151  target)                                                                       \
2152 {                                                                              \
2153   u32 color = psx_gpu->triangle_color;                                         \
2154   dup_8x8b(colors_r, color);                                                   \
2155   dup_8x8b(colors_g, color >> 8);                                              \
2156   dup_8x8b(colors_b, color >> 16);                                             \
2157   shade_blocks_textured_false_modulated_check_##dithering(target);             \
2158 }                                                                              \
2159
2160 #define shade_blocks_textured_modulated_shaded_block_load()                    \
2161   colors_r = block->r;                                                         \
2162   colors_g = block->g;                                                         \
2163   colors_b = block->b                                                          \
2164
2165 #define shade_blocks_textured_modulated_unshaded_block_load()                  \
2166
2167 #define shade_blocks_textured_modulate_dithered(component)                     \
2168   pixels_##component = block->dither_offsets;                                  \
2169   mla_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
2170
2171 #define shade_blocks_textured_modulate_undithered(component)                   \
2172   mul_long_8x8b(pixels_##component, texels_##component, colors_##component)    \
2173
2174 #define shade_blocks_textured_modulated_builder(shading, dithering, target)    \
2175 void shade_blocks_##shading##_textured_modulated_##dithering##_##target(       \
2176  psx_gpu_struct *psx_gpu)                                                      \
2177 {                                                                              \
2178   block_struct *block = psx_gpu->blocks;                                       \
2179   u32 num_blocks = psx_gpu->num_blocks;                                        \
2180   vec_8x16u texels;                                                            \
2181                                                                                \
2182   vec_8x8u texels_r;                                                           \
2183   vec_8x8u texels_g;                                                           \
2184   vec_8x8u texels_b;                                                           \
2185                                                                                \
2186   vec_8x8u colors_r;                                                           \
2187   vec_8x8u colors_g;                                                           \
2188   vec_8x8u colors_b;                                                           \
2189                                                                                \
2190   vec_8x8u pixels_r_low;                                                       \
2191   vec_8x8u pixels_g_low;                                                       \
2192   vec_8x8u pixels_b_low;                                                       \
2193   vec_8x16u pixels;                                                            \
2194                                                                                \
2195   vec_8x16u pixels_r;                                                          \
2196   vec_8x16u pixels_g;                                                          \
2197   vec_8x16u pixels_b;                                                          \
2198                                                                                \
2199   vec_8x16u draw_mask;                                                         \
2200   vec_8x16u zero_mask;                                                         \
2201                                                                                \
2202   vec_8x8u d64_0x07;                                                           \
2203   vec_8x8u d64_0x1F;                                                           \
2204   vec_8x8u d64_1;                                                              \
2205   vec_8x8u d64_4;                                                              \
2206   vec_8x8u d64_128;                                                            \
2207                                                                                \
2208   vec_8x16u d128_0x8000;                                                       \
2209                                                                                \
2210   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2211   u32 draw_mask_bits;                                                          \
2212   shade_blocks_load_msb_mask_##target();                                       \
2213                                                                                \
2214   dup_8x8b(d64_0x07, 0x07);                                                    \
2215   dup_8x8b(d64_0x1F, 0x1F);                                                    \
2216   dup_8x8b(d64_1, 1);                                                          \
2217   dup_8x8b(d64_4, 4);                                                          \
2218   dup_8x8b(d64_128, 128);                                                      \
2219                                                                                \
2220   dup_8x16b(d128_0x8000, 0x8000);                                              \
2221                                                                                \
2222   shade_blocks_textured_modulated_##shading##_primitive_load(dithering,        \
2223    target);                                                                    \
2224                                                                                \
2225   while(num_blocks)                                                            \
2226   {                                                                            \
2227     draw_mask_bits = block->draw_mask_bits;                                    \
2228     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2229     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2230                                                                                \
2231     shade_blocks_textured_modulated_##shading##_block_load();                  \
2232                                                                                \
2233     texels = block->texels;                                                    \
2234                                                                                \
2235     mov_narrow_8x16b(texels_r, texels);                                        \
2236     shr_narrow_8x16b(texels_g, texels, 5);                                     \
2237     shr_narrow_8x16b(texels_b, texels, 7);                                     \
2238                                                                                \
2239     and_8x8b(texels_r, texels_r, d64_0x1F);                                    \
2240     and_8x8b(texels_g, texels_g, d64_0x1F);                                    \
2241     shr_8x8b(texels_b, texels_b, 3);                                           \
2242                                                                                \
2243     shade_blocks_textured_modulate_##dithering(r);                             \
2244     shade_blocks_textured_modulate_##dithering(g);                             \
2245     shade_blocks_textured_modulate_##dithering(b);                             \
2246                                                                                \
2247     cmpeqz_8x16b(zero_mask, texels);                                           \
2248     and_8x16b(pixels, texels, d128_0x8000);                                    \
2249                                                                                \
2250     shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4);                       \
2251     shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4);                       \
2252     shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4);                       \
2253                                                                                \
2254     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2255                                                                                \
2256     shr_8x8b(pixels_r_low, pixels_r_low, 3);                                   \
2257     bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07);                            \
2258     bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07);                            \
2259                                                                                \
2260     mla_long_8x8b(pixels, pixels_r_low, d64_1);                                \
2261     mla_long_8x8b(pixels, pixels_g_low, d64_4);                                \
2262     mla_long_8x8b(pixels, pixels_b_low, d64_128);                              \
2263                                                                                \
2264     shade_blocks_store_##target(zero_mask, pixels);                            \
2265                                                                                \
2266     num_blocks--;                                                              \
2267     block++;                                                                   \
2268   }                                                                            \
2269 }                                                                              \
2270
2271 void shade_blocks_shaded_textured_modulated_dithered_direct(psx_gpu_struct
2272  *psx_gpu);
2273 void shade_blocks_shaded_textured_modulated_undithered_direct(psx_gpu_struct
2274  *psx_gpu);
2275 void shade_blocks_unshaded_textured_modulated_dithered_direct(psx_gpu_struct
2276  *psx_gpu);
2277 void shade_blocks_unshaded_textured_modulated_undithered_direct(psx_gpu_struct
2278  *psx_gpu);
2279
2280 void shade_blocks_shaded_textured_modulated_dithered_indirect(psx_gpu_struct
2281  *psx_gpu);
2282 void shade_blocks_shaded_textured_modulated_undithered_indirect(psx_gpu_struct
2283  *psx_gpu);
2284 void shade_blocks_unshaded_textured_modulated_dithered_indirect(psx_gpu_struct
2285  *psx_gpu);
2286 void shade_blocks_unshaded_textured_modulated_undithered_indirect(psx_gpu_struct
2287  *psx_gpu);
2288
2289 void shade_blocks_textured_unmodulated_indirect(psx_gpu_struct *psx_gpu);
2290 void shade_blocks_textured_unmodulated_direct(psx_gpu_struct *psx_gpu);
2291
2292 #ifndef NEON_BUILD
2293
2294 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2295 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2296 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2297 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2298
2299 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2300 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2301 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2302 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2303
2304 #endif
2305
2306
2307 #define shade_blocks_textured_unmodulated_builder(target)                      \
2308 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu)       \
2309 {                                                                              \
2310   block_struct *block = psx_gpu->blocks;                                       \
2311   u32 num_blocks = psx_gpu->num_blocks;                                        \
2312   vec_8x16u draw_mask;                                                         \
2313   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2314   u32 draw_mask_bits;                                                          \
2315                                                                                \
2316   vec_8x16u pixels;                                                            \
2317   shade_blocks_load_msb_mask_##target();                                       \
2318                                                                                \
2319   while(num_blocks)                                                            \
2320   {                                                                            \
2321     vec_8x16u zero_mask;                                                       \
2322                                                                                \
2323     draw_mask_bits = block->draw_mask_bits;                                    \
2324     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2325     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2326                                                                                \
2327     pixels = block->texels;                                                    \
2328                                                                                \
2329     cmpeqz_8x16b(zero_mask, pixels);                                           \
2330     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2331                                                                                \
2332     shade_blocks_store_##target(zero_mask, pixels);                            \
2333                                                                                \
2334     num_blocks--;                                                              \
2335     block++;                                                                   \
2336   }                                                                            \
2337 }                                                                              \
2338
2339 #define shade_blocks_textured_unmodulated_dithered_builder(target)             \
2340 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct        \
2341  *psx_gpu)                                                                     \
2342 {                                                                              \
2343   block_struct *block = psx_gpu->blocks;                                       \
2344   u32 num_blocks = psx_gpu->num_blocks;                                        \
2345   vec_8x16u draw_mask;                                                         \
2346   vec_8x16u test_mask = psx_gpu->test_mask;                                    \
2347   u32 draw_mask_bits;                                                          \
2348                                                                                \
2349   vec_8x16u pixels;                                                            \
2350   shade_blocks_load_msb_mask_##target();                                       \
2351                                                                                \
2352   while(num_blocks)                                                            \
2353   {                                                                            \
2354     vec_8x16u zero_mask;                                                       \
2355                                                                                \
2356     draw_mask_bits = block->draw_mask_bits;                                    \
2357     dup_8x16b(draw_mask, draw_mask_bits);                                      \
2358     tst_8x16b(draw_mask, draw_mask, test_mask);                                \
2359                                                                                \
2360     pixels = block->texels;                                                    \
2361                                                                                \
2362     cmpeqz_8x16b(zero_mask, pixels);                                           \
2363     or_8x16b(zero_mask, draw_mask, zero_mask);                                 \
2364                                                                                \
2365     shade_blocks_store_##target(zero_mask, pixels);                            \
2366                                                                                \
2367     num_blocks--;                                                              \
2368     block++;                                                                   \
2369   }                                                                            \
2370 }                                                                              \
2371
2372 #ifndef NEON_BUILD
2373
2374 shade_blocks_textured_unmodulated_builder(indirect)
2375 shade_blocks_textured_unmodulated_builder(direct)
2376
2377 #endif
2378
2379
2380 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu);
2381 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu);
2382
2383 #ifndef NEON_BUILD
2384                                                                                
2385 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2386 {
2387 }
2388
2389 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2390 {
2391   block_struct *block = psx_gpu->blocks;
2392   u32 num_blocks = psx_gpu->num_blocks;
2393
2394   vec_8x16u pixels = block->pixels;
2395   shade_blocks_load_msb_mask_direct();
2396
2397   while(num_blocks)
2398   {
2399     shade_blocks_store_direct(block->draw_mask, pixels);
2400
2401     num_blocks--;
2402     block++;
2403   }
2404 }
2405
2406 #endif
2407
2408 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2409 {
2410 }
2411
2412
2413 #define blend_blocks_mask_evaluate_on()                                        \
2414   vec_8x16u mask_pixels;                                                       \
2415   cmpltz_8x16b(mask_pixels, framebuffer_pixels);                               \
2416   or_8x16b(draw_mask, draw_mask, mask_pixels)                                  \
2417
2418 #define blend_blocks_mask_evaluate_off()                                       \
2419
2420 #define blend_blocks_average()                                                 \
2421 {                                                                              \
2422   vec_8x16u pixels_no_msb;                                                     \
2423   vec_8x16u fb_pixels_no_msb;                                                  \
2424                                                                                \
2425   vec_8x16u d128_0x0421;                                                       \
2426   vec_8x16u d128_0x8000;                                                       \
2427                                                                                \
2428   dup_8x16b(d128_0x0421, 0x0421);                                              \
2429   dup_8x16b(d128_0x8000, 0x8000);                                              \
2430                                                                                \
2431   eor_8x16b(blend_pixels, pixels, framebuffer_pixels);                         \
2432   bic_8x16b(pixels_no_msb, pixels, d128_0x8000);                               \
2433   and_8x16b(blend_pixels, blend_pixels, d128_0x0421);                          \
2434   sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels);                        \
2435   bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000);                \
2436   average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels);                 \
2437 }                                                                              \
2438
2439 #define blend_blocks_add()                                                     \
2440 {                                                                              \
2441   vec_8x16u pixels_rb, pixels_g;                                               \
2442   vec_8x16u fb_rb, fb_g;                                                       \
2443                                                                                \
2444   vec_8x16u d128_0x7C1F;                                                       \
2445   vec_8x16u d128_0x03E0;                                                       \
2446                                                                                \
2447   dup_8x16b(d128_0x7C1F, 0x7C1F);