2 * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
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.
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.
23 #include "vector_ops.h"
25 #include "psx_gpu_simd.h"
28 void dump_r_d(const char *name, void *dump);
29 void dump_r_q(const char *name, void *dump);
30 #define dumprd(n) dump_r_d(#n, n.e)
31 #define dumprq(n) dump_r_q(#n, n.e)
35 u32 span_pixel_blocks = 0;
41 u32 sprites_16bpp = 0;
42 u32 sprite_blocks = 0;
43 u32 sprites_untextured = 0;
45 u32 trivial_rejects = 0;
49 u32 texel_blocks_4bpp = 0;
50 u32 texel_blocks_8bpp = 0;
51 u32 texel_blocks_16bpp = 0;
52 u32 texel_blocks_untextured = 0;
54 u32 render_buffer_flushes = 0;
55 u32 state_changes = 0;
56 u32 left_split_triangles = 0;
57 u32 flat_triangles = 0;
58 u32 clipped_triangles = 0;
59 u32 zero_block_spans = 0;
60 u32 texture_cache_loads = 0;
61 u32 false_modulated_blocks = 0;
63 #define stats_add(stat, count) // stat += count
65 /* double size for enhancement */
66 u32 reciprocal_table[512 * 2];
69 typedef s32 fixed_type;
71 #define EDGE_STEP_BITS 32
74 #define fixed_center(value) \
75 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
77 #define int_to_fixed(value) \
78 (((fixed_type)(value)) << FIXED_BITS) \
80 #define fixed_to_int(value) \
81 ((value) >> FIXED_BITS) \
83 #define fixed_to_double(value) \
84 ((value) / (double)(1 << FIXED_BITS)) \
86 #define double_to_fixed(value) \
87 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
89 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
90 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
91 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
92 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
94 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
95 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
97 struct render_block_handler_struct
100 texture_blocks_function_type *texture_blocks;
101 shade_blocks_function_type *shade_blocks;
102 blend_blocks_function_type *blend_blocks;
107 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
109 u32 shift = __builtin_clz(denominator);
110 u32 denominator_normalized = denominator << shift;
112 double numerator = (1ULL << 62) + denominator_normalized;
115 double denominator_normalized_dp_b;
116 u64 denominator_normalized_dp_u64;
119 double reciprocal_dp;
121 u64 numerator_u64 = (denominator_normalized >> 10) |
122 ((u64)(62 + 1023) << 52);
123 *((u64 *)(&numerator_b)) = numerator_u64;
125 denominator_normalized_dp_u64 =
126 (u64)(denominator_normalized << 21) |
127 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
128 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
130 // Implement with a DP divide
131 reciprocal_dp = numerator / denominator_normalized_dp_b;
132 reciprocal = reciprocal_dp;
134 if(reciprocal == 0x80000001)
135 reciprocal = 0x80000000;
137 *_shift = 62 - shift;
141 double reciprocal_estimate(double a)
146 q = (int)(a * 512.0);
147 /* a in units of 1/512 rounded down */
148 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
149 s = (int)(256.0 * r + 0.5);
151 /* r in units of 1/256 rounded to nearest */
153 return (double)s / 256.0;
156 u32 reciprocal_estimate_u32(u32 value)
159 volatile double dp_value;
160 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
162 if((value >> 31) == 0)
165 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
167 *dp_value_ptr = dp_value_u64;
169 dp_value = reciprocal_estimate(dp_value);
170 dp_value_u64 = *dp_value_ptr;
172 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
175 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
177 u32 shift = __builtin_clz(value);
178 u32 value_normalized = value << shift;
180 *_shift = 62 - shift;
182 value_normalized -= 2;
184 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
186 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
187 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
188 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
189 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
190 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
191 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
193 return reciprocal_normalized;
199 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
201 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
204 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
206 s32 coverage_x, coverage_y;
211 coverage_x = x2 >> 6;
212 coverage_y = y2 >> 8;
220 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
223 mask_down_right |= mask_down_right << 16;
225 coverage_x = x1 >> 6;
227 mask_up_left = 0xFFFF0000 << coverage_x;
229 mask_up_left = 0xFFFF0000;
231 coverage_y = y1 >> 8;
233 mask_up_left |= mask_up_left >> 16;
235 return mask_up_left & mask_down_right;
238 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
241 u32 mask = texture_region_mask(x1, y1, x2, y2);
243 psx_gpu->dirty_textures_4bpp_mask |= mask;
244 psx_gpu->dirty_textures_8bpp_mask |= mask;
245 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
250 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
251 u32 y1, u32 x2, u32 y2)
253 u32 mask = texture_region_mask(x1, y1, x2, y2) &
254 psx_gpu->viewport_mask;
256 psx_gpu->dirty_textures_4bpp_mask |= mask;
257 psx_gpu->dirty_textures_8bpp_mask |= mask;
258 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
263 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
266 u32 mask = texture_region_mask(x1, y1, x2, y2);
268 u8 *texture_page_ptr;
273 psx_gpu->dirty_textures_8bpp_mask |= mask;
274 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
276 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
277 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
279 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
280 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
281 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
282 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
290 texel_block = *vram_ptr;
292 texture_page_ptr[0] = texel_block & 0xF;
293 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
294 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
295 texture_page_ptr[3] = texel_block >> 12;
298 texture_page_ptr += 4;
312 psx_gpu->dirty_textures_4bpp_mask |= mask;
318 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
320 u32 current_texture_page = psx_gpu->current_texture_page;
321 u8 *texture_page_ptr = psx_gpu->texture_page_base;
322 u16 *vram_ptr = psx_gpu->vram_ptr;
328 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
329 vram_ptr += (current_texture_page & 0xF) * 64;
331 texture_cache_loads++;
338 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
348 texel_block = *vram_ptr;
350 texture_page_ptr[0] = texel_block & 0xF;
351 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
352 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
353 texture_page_ptr[3] = texel_block >> 12;
356 texture_page_ptr += 4;
370 vram_ptr -= (1024 * 16) - 4;
376 vram_ptr += (16 * 1024) - (4 * 16);
381 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
384 u16 *texture_page_ptr = psx_gpu->texture_page_base;
385 u16 *vram_ptr = psx_gpu->vram_ptr;
392 texture_cache_loads++;
394 vram_ptr += (texture_page >> 4) * 256 * 1024;
395 vram_ptr += (texture_page & 0xF) * 64;
397 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
398 texture_page_ptr += (8 * 16) * 8;
411 load_128b(texels, vram_ptr);
412 store_128b(texels, texture_page_ptr);
414 texture_page_ptr += 8;
422 vram_ptr -= (1024 * 16);
431 vram_ptr += (16 * 1024);
433 texture_page_ptr += (8 * 16) * 8;
441 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
443 u32 current_texture_page = psx_gpu->current_texture_page;
444 u32 update_textures =
445 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
447 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
449 if(update_textures & (1 << current_texture_page))
451 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
452 update_textures &= ~(1 << current_texture_page);
457 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
458 (current_texture_page & 0x10);
460 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
464 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
466 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
467 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
469 u32 num_blocks_dest = 0;
470 block_struct *block_src = psx_gpu->blocks;
471 block_struct *block_dest = psx_gpu->blocks;
473 u16 *vram_ptr = psx_gpu->vram_ptr;
476 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
478 for(i = 0; i < psx_gpu->num_blocks; i++)
480 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
481 if(fb_offset & (1 << 11))
483 *block_dest = *block_src;
492 for(i = 0; i < psx_gpu->num_blocks; i++)
494 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
495 if((fb_offset & (1 << 11)) == 0)
497 *block_dest = *block_src;
505 psx_gpu->num_blocks = num_blocks_dest;
508 if(psx_gpu->num_blocks)
510 render_block_handler_struct *render_block_handler =
511 psx_gpu->render_block_handler;
513 render_block_handler->texture_blocks(psx_gpu);
514 render_block_handler->shade_blocks(psx_gpu);
515 render_block_handler->blend_blocks(psx_gpu);
518 span_pixel_blocks += psx_gpu->num_blocks;
519 render_buffer_flushes++;
522 psx_gpu->num_blocks = 0;
529 #define setup_gradient_calculation_input(set, vertex) \
530 /* First type is: uvrg bxxx xxxx */\
531 /* Second type is: yyyy ybyy uvrg */\
532 /* Since x_a and y_c are the same the same variable is used for both. */\
533 x##set##_a_y##set##_c.e[0] = vertex->u; \
534 x##set##_a_y##set##_c.e[1] = vertex->v; \
535 x##set##_a_y##set##_c.e[2] = vertex->r; \
536 x##set##_a_y##set##_c.e[3] = vertex->g; \
537 dup_4x16b(x##set##_b, vertex->x); \
538 dup_4x16b(x##set##_c, vertex->x); \
539 dup_4x16b(y##set##_a, vertex->y); \
540 dup_4x16b(y##set##_b, vertex->y); \
541 x##set##_b.e[0] = vertex->b; \
542 y##set##_b.e[1] = vertex->b \
545 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
546 vertex_struct *b, vertex_struct *c)
548 u32 triangle_area = psx_gpu->triangle_area;
549 u32 winding_mask_scalar;
551 u32 triangle_area_shift;
552 u64 triangle_area_reciprocal =
553 fixed_reciprocal(triangle_area, &triangle_area_shift);
554 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
556 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
557 // ( d0 * d1 ) - ( d2 * d3 ) =
558 // ( m0 ) - ( m1 ) = gradient
560 // This is split to do 12 elements at a time over three sets: a, b, and c.
561 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
562 // two of the slots are unused.
564 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
567 vec_4x16s x0_a_y0_c, x0_b, x0_c;
568 vec_4x16s y0_a, y0_b;
569 vec_4x16s x1_a_y1_c, x1_b, x1_c;
570 vec_4x16s y1_a, y1_b;
571 vec_4x16s x2_a_y2_c, x2_b, x2_c;
572 vec_4x16s y2_a, y2_b;
576 vec_4x32u uvrgb_phase;
578 vec_4x16s d0_a_d3_c, d0_b, d0_c;
579 vec_4x16s d1_a, d1_b, d1_c_d2_a;
580 vec_4x16s d2_b, d2_c;
581 vec_4x16s d3_a, d3_b;
583 vec_4x32s m0_a, m0_b, m0_c;
584 vec_4x32s m1_a, m1_b, m1_c;
586 vec_4x32u gradient_area_a, gradient_area_c;
587 vec_2x32u gradient_area_b;
589 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
590 vec_2x32u gradient_area_sign_b;
591 vec_4x32u winding_mask;
593 vec_2x64u gradient_wide_a0, gradient_wide_a1;
594 vec_2x64u gradient_wide_c0, gradient_wide_c1;
595 vec_2x64u gradient_wide_b;
597 vec_4x32u gradient_a, gradient_c;
598 vec_2x32u gradient_b;
599 vec_16x8s gradient_shift;
601 setup_gradient_calculation_input(0, a);
602 setup_gradient_calculation_input(1, b);
603 setup_gradient_calculation_input(2, c);
605 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
606 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
607 shl_long_4x16b(b_base, x0_b, 16);
609 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
610 add_4x32b(b_base, b_base, uvrgb_phase);
612 // Can probably pair these, but it'll require careful register allocation
613 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
614 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
616 sub_4x16b(d0_b, x1_b, x0_b);
617 sub_4x16b(d0_c, x1_c, x0_c);
619 sub_4x16b(d1_a, y2_a, y1_a);
620 sub_4x16b(d1_b, y2_b, y1_b);
622 sub_4x16b(d2_b, x2_b, x1_b);
623 sub_4x16b(d2_c, x2_c, x1_c);
625 sub_4x16b(d3_a, y1_a, y0_a);
626 sub_4x16b(d3_b, y1_b, y0_b);
628 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
629 mul_long_4x16b(m0_b, d0_b, d1_b);
630 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
632 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
633 mul_long_4x16b(m1_b, d2_b, d3_b);
634 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
636 sub_4x32b(gradient_area_a, m0_a, m1_a);
637 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
638 sub_4x32b(gradient_area_c, m0_c, m1_c);
640 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
641 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
642 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
644 abs_4x32b(gradient_area_a, gradient_area_a);
645 abs_2x32b(gradient_area_b, gradient_area_b);
646 abs_4x32b(gradient_area_c, gradient_area_c);
648 winding_mask_scalar = -psx_gpu->triangle_winding;
650 dup_4x32b(winding_mask, winding_mask_scalar);
651 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
652 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
653 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
655 mul_scalar_long_2x32b(gradient_wide_a0,
656 vector_cast(vec_2x32s, gradient_area_a.low),
657 (s64)triangle_area_reciprocal);
658 mul_scalar_long_2x32b(gradient_wide_a1,
659 vector_cast(vec_2x32s, gradient_area_a.high),
660 (s64)triangle_area_reciprocal);
661 mul_scalar_long_2x32b(gradient_wide_b,
662 vector_cast(vec_2x32s, gradient_area_b),
663 (s64)triangle_area_reciprocal);
664 mul_scalar_long_2x32b(gradient_wide_c0,
665 vector_cast(vec_2x32s, gradient_area_c.low),
666 (s64)triangle_area_reciprocal);
667 mul_scalar_long_2x32b(gradient_wide_c1,
668 vector_cast(vec_2x32s, gradient_area_c.high),
669 (s64)triangle_area_reciprocal);
671 dup_16x8b(gradient_shift, triangle_area_shift);
672 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
673 vector_cast(vec_2x64u, gradient_shift));
674 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
675 vector_cast(vec_2x64u, gradient_shift));
676 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
677 vector_cast(vec_2x64u, gradient_shift));
678 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
679 vector_cast(vec_2x64u, gradient_shift));
680 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
681 vector_cast(vec_2x64u, gradient_shift));
683 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
684 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
685 mov_narrow_2x64b(gradient_b, gradient_wide_b);
686 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
687 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
689 shl_4x32b(gradient_a, gradient_a, 4);
690 shl_2x32b(gradient_b, gradient_b, 4);
691 shl_4x32b(gradient_c, gradient_c, 4);
693 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
694 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
695 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
697 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
698 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
699 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
701 u32 left_adjust = a->x;
702 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
703 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
713 eor_4x32b(zero, zero, zero);
714 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
715 add_2x32b(b_dx2, gradient_b, gradient_b);
716 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
717 add_2x32b(b_dx3, gradient_b, b_dx2);
719 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
721 psx_gpu->u_block_span.e[0] = zero.e[0];
722 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
723 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
724 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
726 psx_gpu->v_block_span.e[0] = zero.e[1];
727 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
728 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
729 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
731 psx_gpu->r_block_span.e[0] = zero.e[2];
732 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
733 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
734 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
736 psx_gpu->g_block_span.e[0] = zero.e[3];
737 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
738 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
739 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
741 psx_gpu->b_block_span.e[0] = zero.e[0];
742 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
743 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
744 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
746 psx_gpu->uvrg = uvrg_base;
747 psx_gpu->b = b_base.e[0];
749 psx_gpu->uvrg_dx = gradient_a;
750 psx_gpu->uvrg_dy = gradient_c;
751 psx_gpu->b_dy = gradient_b.e[1];
755 #define vector_check(_a, _b) \
756 if(memcmp(&_a, &_b, sizeof(_b))) \
758 if(sizeof(_b) == 8) \
760 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
761 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
765 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
766 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
771 #define scalar_check(_a, _b) \
773 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
776 #if !defined(NEON_BUILD) && !defined(NDEBUG)
777 static void setup_spans_debug_check(psx_gpu_struct *psx_gpu,
778 edge_data_struct *span_edge_data_element)
780 u32 _num_spans = span_edge_data_element - psx_gpu->span_edge_data;
781 if (_num_spans > MAX_SPANS)
782 *(volatile int *)0 = 1;
783 if (_num_spans < psx_gpu->num_spans)
785 if(span_edge_data_element->num_blocks > MAX_BLOCKS_PER_ROW)
786 *(volatile int *)0 = 2;
787 if(span_edge_data_element->y >= 2048)
788 *(volatile int *)0 = 3;
792 #define setup_spans_debug_check(psx_gpu, span_edge_data_element)
795 #define setup_spans_prologue_alternate_yes() \
796 vec_2x64s alternate_x; \
797 vec_2x64s alternate_dx_dy; \
798 vec_4x32s alternate_x_32; \
799 vec_4x16u alternate_x_16; \
801 vec_4x16u alternate_select; \
802 vec_4x16s y_mid_point; \
806 s32 edge_dx_dy_alt; \
809 #define setup_spans_prologue_alternate_no() \
811 #define setup_spans_prologue(alternate_active) \
812 edge_data_struct *span_edge_data; \
813 vec_4x32u *span_uvrg_offset; \
814 u32 *span_b_offset; \
818 vec_2x64s edges_xy; \
819 vec_2x32s edges_dx_dy; \
820 vec_2x32u edge_shifts; \
822 vec_2x64s left_x, right_x; \
823 vec_2x64s left_dx_dy, right_dx_dy; \
824 vec_4x32s left_x_32, right_x_32; \
825 vec_8x16s left_right_x_16; \
827 vec_8x16s left_edge; \
828 vec_8x16s right_edge; \
829 vec_4x16u span_shift; \
833 vec_4x16u c_0xFFFE; \
836 vec_2x32s x_starts; \
845 vec_4x32u uvrg = psx_gpu->uvrg; \
846 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
847 u32 b = psx_gpu->b; \
848 u32 b_dy = psx_gpu->b_dy; \
850 dup_2x32b(c_0x01, 0x01); \
851 setup_spans_prologue_alternate_##alternate_active() \
853 #define setup_spans_prologue_b() \
854 span_edge_data = psx_gpu->span_edge_data; \
855 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
856 span_b_offset = psx_gpu->span_b_offset; \
858 vec_8x16u c_0x0001; \
859 vec_4x16u c_max_blocks_per_row; \
861 dup_8x16b(c_0x0001, 0x0001); \
862 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
863 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
864 add_8x16b(right_edge, right_edge, c_0x0001); \
865 dup_4x16b(c_0x04, 0x04); \
866 dup_4x16b(c_0x07, 0x07); \
867 dup_4x16b(c_0xFFFE, 0xFFFE); \
868 dup_4x16b(c_max_blocks_per_row, MAX_BLOCKS_PER_ROW); \
871 #define compute_edge_delta_x2() \
874 vec_2x32s height_reciprocals; \
875 vec_2x32s heights_b; \
878 u32 edge_shift = reciprocal_table[height]; \
880 dup_2x32b(heights, height); \
881 sub_2x32b(widths, x_ends, x_starts); \
883 dup_2x32b(edge_shifts, edge_shift); \
884 sub_2x32b(heights_b, heights, c_0x01); \
885 shr_2x32b(height_reciprocals, edge_shifts, 10); \
887 mla_2x32b(heights_b, x_starts, heights); \
888 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
889 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
890 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
893 #define compute_edge_delta_x3(start_c, height_a, height_b) \
896 vec_2x32s height_reciprocals; \
897 vec_2x32s heights_b; \
902 u32 height_reciprocal_alt; \
904 heights.e[0] = height_a; \
905 heights.e[1] = height_b; \
907 edge_shifts.e[0] = reciprocal_table[height_a]; \
908 edge_shifts.e[1] = reciprocal_table[height_b]; \
909 edge_shift_alt = reciprocal_table[height_minor_b]; \
911 sub_2x32b(widths, x_ends, x_starts); \
912 width_alt = x_c - start_c; \
914 shr_2x32b(height_reciprocals, edge_shifts, 10); \
915 height_reciprocal_alt = edge_shift_alt >> 10; \
917 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
918 edge_shift_alt &= 0x1F; \
920 sub_2x32b(heights_b, heights, c_0x01); \
921 height_b_alt = height_minor_b - 1; \
923 mla_2x32b(heights_b, x_starts, heights); \
924 height_b_alt += height_minor_b * start_c; \
926 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
927 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
929 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
930 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
934 #define setup_spans_adjust_y_up() \
935 sub_4x32b(y_x4, y_x4, c_0x04) \
937 #define setup_spans_adjust_y_down() \
938 add_4x32b(y_x4, y_x4, c_0x04) \
940 #define setup_spans_adjust_interpolants_up() \
941 sub_4x32b(uvrg, uvrg, uvrg_dy); \
944 #define setup_spans_adjust_interpolants_down() \
945 add_4x32b(uvrg, uvrg, uvrg_dy); \
949 #define setup_spans_clip_interpolants_increment() \
950 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
953 #define setup_spans_clip_interpolants_decrement() \
954 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
957 #define setup_spans_clip_alternate_yes() \
958 edge_alt += edge_dx_dy_alt * (s64)(clip) \
960 #define setup_spans_clip_alternate_no() \
962 #define setup_spans_clip(direction, alternate_active) \
964 clipped_triangles++; \
965 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
966 setup_spans_clip_alternate_##alternate_active(); \
967 setup_spans_clip_interpolants_##direction(); \
971 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
973 vec_2x64u edge_shifts_64; \
974 vec_2x64s edges_dx_dy_64; \
976 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
977 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
979 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
980 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
982 left_x.e[0] = edges_xy.e[left_index]; \
983 right_x.e[0] = edges_xy.e[right_index]; \
985 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
986 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
987 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
988 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
990 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
991 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
993 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
994 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
997 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
999 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
1000 s64 edge_dx_dy_alt_64; \
1002 dup_4x16b(y_mid_point, y_b); \
1004 edge_alt <<= edge_shift_alt; \
1005 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
1007 alternate_x.e[0] = edge_alt; \
1008 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
1009 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1011 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1012 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1016 #define setup_spans_y_select_up() \
1017 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1019 #define setup_spans_y_select_down() \
1020 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1022 #define setup_spans_y_select_alternate_yes(direction) \
1023 setup_spans_y_select_##direction() \
1025 #define setup_spans_y_select_alternate_no(direction) \
1027 #define setup_spans_alternate_select_left() \
1028 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1030 #define setup_spans_alternate_select_right() \
1031 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1033 #define setup_spans_alternate_select_none() \
1035 #define setup_spans_increment_alternate_yes() \
1036 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1037 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1038 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1039 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1040 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1042 #define setup_spans_increment_alternate_no() \
1044 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1046 span_uvrg_offset[0] = uvrg; \
1047 span_b_offset[0] = b; \
1048 setup_spans_adjust_interpolants_##direction(); \
1050 span_uvrg_offset[1] = uvrg; \
1051 span_b_offset[1] = b; \
1052 setup_spans_adjust_interpolants_##direction(); \
1054 span_uvrg_offset[2] = uvrg; \
1055 span_b_offset[2] = b; \
1056 setup_spans_adjust_interpolants_##direction(); \
1058 span_uvrg_offset[3] = uvrg; \
1059 span_b_offset[3] = b; \
1060 setup_spans_adjust_interpolants_##direction(); \
1062 span_uvrg_offset += 4; \
1063 span_b_offset += 4; \
1065 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1066 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1068 add_2x64b(left_x, left_x, left_dx_dy); \
1069 add_2x64b(right_x, right_x, right_dx_dy); \
1071 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1072 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1074 add_2x64b(left_x, left_x, left_dx_dy); \
1075 add_2x64b(right_x, right_x, right_dx_dy); \
1077 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1078 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1080 setup_spans_increment_alternate_##alternate_active(); \
1081 setup_spans_y_select_alternate_##alternate_active(direction); \
1082 setup_spans_alternate_select_##alternate(); \
1084 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1085 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1087 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1088 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1089 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1090 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1091 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1092 min_4x16b(left_right_x_16.high, left_right_x_16.high, c_max_blocks_per_row); \
1095 for(i = 0; i < 4; i++) \
1097 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1098 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1099 span_edge_data[i].right_mask = span_shift.e[i]; \
1100 span_edge_data[i].y = y_x4.e[i]; \
1101 setup_spans_debug_check(psx_gpu, &span_edge_data[i]); \
1104 span_edge_data += 4; \
1106 setup_spans_adjust_y_##direction(); \
1110 #define setup_spans_alternate_adjust_yes() \
1111 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1113 #define setup_spans_alternate_adjust_no() \
1116 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1117 setup_spans_alternate_adjust_##alternate_active(); \
1118 if(y_c > psx_gpu->viewport_end_y) \
1119 height -= y_c - psx_gpu->viewport_end_y - 1; \
1121 clip = psx_gpu->viewport_start_y - y_a; \
1126 setup_spans_clip(increment, alternate_active); \
1129 setup_spans_prologue_b(); \
1136 y_x4.e[1] = y_a + 1; \
1137 y_x4.e[2] = y_a + 2; \
1138 y_x4.e[3] = y_a + 3; \
1139 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1142 psx_gpu->num_spans = height; \
1145 setup_spans_set_x4(alternate, down, alternate_active); \
1147 } while(height > 0); \
1151 #define setup_spans_alternate_pre_increment_yes() \
1152 edge_alt += edge_dx_dy_alt \
1154 #define setup_spans_alternate_pre_increment_no() \
1156 #define setup_spans_up_decrement_height_yes() \
1159 #define setup_spans_up_decrement_height_no() \
1162 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1163 setup_spans_alternate_adjust_##alternate_active(); \
1166 if(y_c < psx_gpu->viewport_start_y) \
1167 height -= psx_gpu->viewport_start_y - y_c; \
1169 setup_spans_up_decrement_height_##alternate_active(); \
1171 clip = y_a - psx_gpu->viewport_end_y; \
1176 setup_spans_clip(decrement, alternate_active); \
1179 setup_spans_prologue_b(); \
1186 y_x4.e[1] = y_a - 1; \
1187 y_x4.e[2] = y_a - 2; \
1188 y_x4.e[3] = y_a - 3; \
1189 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1190 setup_spans_alternate_pre_increment_##alternate_active(); \
1191 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1193 setup_spans_adjust_interpolants_up(); \
1195 psx_gpu->num_spans = height; \
1198 setup_spans_set_x4(alternate, up, alternate_active); \
1203 #define index_left 0
1204 #define index_right 1
1206 #define setup_spans_up_up(minor, major) \
1207 setup_spans_prologue(yes); \
1208 s32 height_minor_a = y_a - y_b; \
1209 s32 height_minor_b = y_b - y_c; \
1210 s32 height = y_a - y_c; \
1212 dup_2x32b(x_starts, x_a); \
1213 x_ends.e[0] = x_c; \
1214 x_ends.e[1] = x_b; \
1216 compute_edge_delta_x3(x_b, height, height_minor_a); \
1217 setup_spans_up(index_##major, index_##minor, minor, yes) \
1222 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1223 vertex_struct *v_b, vertex_struct *v_c)
1225 setup_spans_up_up(left, right);
1228 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1229 vertex_struct *v_b, vertex_struct *v_c)
1231 setup_spans_up_up(right, left);
1234 #define setup_spans_down_down(minor, major) \
1235 setup_spans_prologue(yes); \
1236 s32 height_minor_a = y_b - y_a; \
1237 s32 height_minor_b = y_c - y_b; \
1238 s32 height = y_c - y_a; \
1240 dup_2x32b(x_starts, x_a); \
1241 x_ends.e[0] = x_c; \
1242 x_ends.e[1] = x_b; \
1244 compute_edge_delta_x3(x_b, height, height_minor_a); \
1245 setup_spans_down(index_##major, index_##minor, minor, yes) \
1247 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1248 vertex_struct *v_b, vertex_struct *v_c)
1250 setup_spans_down_down(left, right);
1253 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1254 vertex_struct *v_b, vertex_struct *v_c)
1256 setup_spans_down_down(right, left);
1259 #define setup_spans_up_flat() \
1260 s32 height = y_a - y_c; \
1263 compute_edge_delta_x2(); \
1264 setup_spans_up(index_left, index_right, none, no) \
1266 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1267 vertex_struct *v_b, vertex_struct *v_c)
1269 setup_spans_prologue(no);
1270 x_starts.e[0] = x_a;
1271 x_starts.e[1] = x_b;
1272 dup_2x32b(x_ends, x_c);
1274 setup_spans_up_flat();
1277 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1278 vertex_struct *v_b, vertex_struct *v_c)
1280 setup_spans_prologue(no);
1281 dup_2x32b(x_starts, x_a);
1285 setup_spans_up_flat();
1288 #define setup_spans_down_flat() \
1289 s32 height = y_c - y_a; \
1292 compute_edge_delta_x2(); \
1293 setup_spans_down(index_left, index_right, none, no) \
1295 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1296 vertex_struct *v_b, vertex_struct *v_c)
1298 setup_spans_prologue(no);
1299 x_starts.e[0] = x_a;
1300 x_starts.e[1] = x_b;
1301 dup_2x32b(x_ends, x_c);
1303 setup_spans_down_flat();
1306 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1307 vertex_struct *v_b, vertex_struct *v_c)
1309 setup_spans_prologue(no);
1310 dup_2x32b(x_starts, x_a);
1314 setup_spans_down_flat();
1317 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1318 vertex_struct *v_b, vertex_struct *v_c)
1320 setup_spans_prologue(no);
1328 s32 height_minor_a = y_a - y_b;
1329 s32 height_minor_b = y_c - y_a;
1330 s32 height_major = y_c - y_b;
1332 vec_2x64s edges_xy_b;
1333 vec_2x32s edges_dx_dy_b;
1334 vec_2x32u edge_shifts_b;
1336 vec_2x32s height_increment;
1338 x_starts.e[0] = x_a;
1339 x_starts.e[1] = x_c;
1340 dup_2x32b(x_ends, x_b);
1342 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1344 height_increment.e[0] = 0;
1345 height_increment.e[1] = height_minor_b;
1347 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1349 edges_xy_b.e[0] = edge_alt;
1350 edges_xy_b.e[1] = edges_xy.e[1];
1352 edge_shifts_b = edge_shifts;
1353 edge_shifts_b.e[0] = edge_shift_alt;
1355 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1356 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1360 if(y_b < psx_gpu->viewport_start_y)
1361 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1363 clip = y_a - psx_gpu->viewport_end_y;
1366 height_minor_a -= clip;
1368 setup_spans_clip(decrement, no);
1371 setup_spans_prologue_b();
1373 if (height_minor_a > 512)
1374 height_minor_a = 512;
1375 if (height_minor_a > 0)
1378 y_x4.e[1] = y_a - 1;
1379 y_x4.e[2] = y_a - 2;
1380 y_x4.e[3] = y_a - 3;
1381 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1382 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1383 setup_spans_adjust_interpolants_up();
1385 psx_gpu->num_spans = height_minor_a;
1386 while(height_minor_a > 0)
1388 setup_spans_set_x4(none, up, no);
1389 height_minor_a -= 4;
1392 span_edge_data += height_minor_a;
1393 span_uvrg_offset += height_minor_a;
1394 span_b_offset += height_minor_a;
1397 edges_xy = edges_xy_b;
1398 edges_dx_dy = edges_dx_dy_b;
1399 edge_shifts = edge_shifts_b;
1401 uvrg = psx_gpu->uvrg;
1406 if(y_c > psx_gpu->viewport_end_y)
1407 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1409 clip = psx_gpu->viewport_start_y - y_a;
1412 height_minor_b -= clip;
1414 setup_spans_clip(increment, no);
1417 if (height_minor_b > 512)
1418 height_minor_b = 512;
1419 if (height_minor_b > 0)
1422 y_x4.e[1] = y_a + 1;
1423 y_x4.e[2] = y_a + 2;
1424 y_x4.e[3] = y_a + 3;
1425 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1427 // FIXME: overflow corner case
1428 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1429 height_minor_b &= ~3;
1431 psx_gpu->num_spans += height_minor_b;
1432 while(height_minor_b > 0)
1434 setup_spans_set_x4(none, down, no);
1435 height_minor_b -= 4;
1439 left_split_triangles++;
1445 #define dither_table_entry_normal(value) \
1449 #define setup_blocks_load_msb_mask_indirect() \
1451 #define setup_blocks_load_msb_mask_direct() \
1452 vec_8x16u msb_mask; \
1453 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1456 #define setup_blocks_variables_shaded_textured(target) \
1457 vec_4x32u u_block; \
1458 vec_4x32u v_block; \
1459 vec_4x32u r_block; \
1460 vec_4x32u g_block; \
1461 vec_4x32u b_block; \
1462 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1463 vec_4x32u uvrg_dx4; \
1464 vec_4x32u uvrg_dx8; \
1466 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1467 u32 b_dx4 = b_dx << 2; \
1468 u32 b_dx8 = b_dx << 3; \
1471 vec_16x8u texture_mask; \
1472 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1473 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1474 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1475 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1477 #define setup_blocks_variables_shaded_untextured(target) \
1478 vec_4x32u r_block; \
1479 vec_4x32u g_block; \
1480 vec_4x32u b_block; \
1482 vec_4x32u rgb_dx4; \
1483 vec_4x32u rgb_dx8; \
1486 vec_8x8u d64_0x07; \
1491 dup_8x8b(d64_0x07, 0x07); \
1492 dup_8x8b(d64_1, 1); \
1493 dup_8x8b(d64_4, 4); \
1494 dup_8x8b(d64_128, 128); \
1496 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1497 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1498 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1499 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1501 #define setup_blocks_variables_unshaded_textured(target) \
1502 vec_4x32u u_block; \
1503 vec_4x32u v_block; \
1504 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1507 vec_2x32u uv = psx_gpu->uvrg.low; \
1509 vec_16x8u texture_mask; \
1510 shl_2x32b(uv_dx4, uv_dx, 2); \
1511 shl_2x32b(uv_dx8, uv_dx, 3); \
1512 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1513 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1516 #define setup_blocks_variables_unshaded_untextured_direct() \
1517 or_8x16b(colors, colors, msb_mask) \
1519 #define setup_blocks_variables_unshaded_untextured_indirect() \
1521 #define setup_blocks_variables_unshaded_untextured(target) \
1522 u32 color = psx_gpu->triangle_color; \
1525 u32 color_r = color & 0xFF; \
1526 u32 color_g = (color >> 8) & 0xFF; \
1527 u32 color_b = (color >> 16) & 0xFF; \
1529 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1530 ((color_b >> 3) << 10); \
1531 dup_8x16b(colors, color); \
1532 setup_blocks_variables_unshaded_untextured_##target() \
1534 #define setup_blocks_span_initialize_dithered_textured() \
1535 vec_8x16u dither_offsets; \
1536 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1538 #define setup_blocks_span_initialize_dithered_untextured() \
1539 vec_8x8u dither_offsets; \
1540 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1542 #define setup_blocks_span_initialize_dithered(texturing) \
1543 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1544 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1545 vec_8x8s dither_offsets_short; \
1548 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1549 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1550 setup_blocks_span_initialize_dithered_##texturing() \
1552 #define setup_blocks_span_initialize_undithered(texturing) \
1555 #define setup_blocks_span_initialize_shaded_textured() \
1557 vec_4x32u block_span; \
1558 u32 offset = span_edge_data->left_x; \
1560 uvrg = *span_uvrg_offset; \
1561 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1562 b = *span_b_offset; \
1563 b += b_dx * offset; \
1565 dup_4x32b(u_block, uvrg.e[0]); \
1566 dup_4x32b(v_block, uvrg.e[1]); \
1567 dup_4x32b(r_block, uvrg.e[2]); \
1568 dup_4x32b(g_block, uvrg.e[3]); \
1569 dup_4x32b(b_block, b); \
1571 block_span = psx_gpu->u_block_span; \
1572 add_4x32b(u_block, u_block, block_span); \
1573 block_span = psx_gpu->v_block_span; \
1574 add_4x32b(v_block, v_block, block_span); \
1575 block_span = psx_gpu->r_block_span; \
1576 add_4x32b(r_block, r_block, block_span); \
1577 block_span = psx_gpu->g_block_span; \
1578 add_4x32b(g_block, g_block, block_span); \
1579 block_span = psx_gpu->b_block_span; \
1580 add_4x32b(b_block, b_block, block_span); \
1583 #define setup_blocks_span_initialize_shaded_untextured() \
1585 vec_4x32u block_span; \
1586 u32 offset = span_edge_data->left_x; \
1588 rgb.low = span_uvrg_offset->high; \
1589 rgb.high.e[0] = *span_b_offset; \
1590 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1592 dup_4x32b(r_block, rgb.e[0]); \
1593 dup_4x32b(g_block, rgb.e[1]); \
1594 dup_4x32b(b_block, rgb.e[2]); \
1596 block_span = psx_gpu->r_block_span; \
1597 add_4x32b(r_block, r_block, block_span); \
1598 block_span = psx_gpu->g_block_span; \
1599 add_4x32b(g_block, g_block, block_span); \
1600 block_span = psx_gpu->b_block_span; \
1601 add_4x32b(b_block, b_block, block_span); \
1604 #define setup_blocks_span_initialize_unshaded_textured() \
1606 vec_4x32u block_span; \
1607 u32 offset = span_edge_data->left_x; \
1609 uv = span_uvrg_offset->low; \
1610 mla_scalar_2x32b(uv, uv_dx, offset); \
1612 dup_4x32b(u_block, uv.e[0]); \
1613 dup_4x32b(v_block, uv.e[1]); \
1615 block_span = psx_gpu->u_block_span; \
1616 add_4x32b(u_block, u_block, block_span); \
1617 block_span = psx_gpu->v_block_span; \
1618 add_4x32b(v_block, v_block, block_span); \
1621 #define setup_blocks_span_initialize_unshaded_untextured() \
1624 #define setup_blocks_texture_swizzled() \
1626 vec_8x8u u_saved = u; \
1627 sli_8x8b(u, v, 4); \
1628 sri_8x8b(v, u_saved, 4); \
1631 #define setup_blocks_texture_unswizzled() \
1633 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1636 vec_8x16u u_whole; \
1637 vec_8x16u v_whole; \
1638 vec_8x16u r_whole; \
1639 vec_8x16u g_whole; \
1640 vec_8x16u b_whole; \
1652 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1653 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1654 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1655 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1656 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1658 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1659 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1660 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1661 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1662 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1663 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1664 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1665 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1666 dup_4x32b(dx4, b_dx4); \
1667 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1669 mov_narrow_8x16b(u, u_whole); \
1670 mov_narrow_8x16b(v, v_whole); \
1671 mov_narrow_8x16b(r, r_whole); \
1672 mov_narrow_8x16b(g, g_whole); \
1673 mov_narrow_8x16b(b, b_whole); \
1675 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1676 add_4x32b(u_block, u_block, dx8); \
1677 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1678 add_4x32b(v_block, v_block, dx8); \
1679 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1680 add_4x32b(r_block, r_block, dx8); \
1681 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1682 add_4x32b(g_block, g_block, dx8); \
1683 dup_4x32b(dx8, b_dx8); \
1684 add_4x32b(b_block, b_block, dx8); \
1686 and_8x8b(u, u, texture_mask.low); \
1687 and_8x8b(v, v, texture_mask.high); \
1688 setup_blocks_texture_##swizzling(); \
1690 zip_8x16b(uv, u, v); \
1695 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1696 block->fb_ptr = fb_ptr; \
1699 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1702 vec_8x16u u_whole; \
1703 vec_8x16u v_whole; \
1712 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1713 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1715 dup_4x32b(dx4, uv_dx4.e[0]); \
1716 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1717 dup_4x32b(dx4, uv_dx4.e[1]); \
1718 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1720 mov_narrow_8x16b(u, u_whole); \
1721 mov_narrow_8x16b(v, v_whole); \
1723 dup_4x32b(dx8, uv_dx8.e[0]); \
1724 add_4x32b(u_block, u_block, dx8); \
1725 dup_4x32b(dx8, uv_dx8.e[1]); \
1726 add_4x32b(v_block, v_block, dx8); \
1728 and_8x8b(u, u, texture_mask.low); \
1729 and_8x8b(v, v, texture_mask.high); \
1730 setup_blocks_texture_##swizzling(); \
1732 zip_8x16b(uv, u, v); \
1734 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1735 block->fb_ptr = fb_ptr; \
1738 #define setup_blocks_store_shaded_untextured_dithered() \
1739 addq_8x8b(r, r, dither_offsets); \
1740 addq_8x8b(g, g, dither_offsets); \
1741 addq_8x8b(b, b, dither_offsets); \
1743 subq_8x8b(r, r, d64_4); \
1744 subq_8x8b(g, g, d64_4); \
1745 subq_8x8b(b, b, d64_4) \
1747 #define setup_blocks_store_shaded_untextured_undithered() \
1750 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1751 block->pixels = _pixels; \
1752 block->fb_ptr = fb_ptr \
1754 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1755 block->pixels = _pixels; \
1756 block->fb_ptr = fb_ptr \
1758 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1759 mul_long_8x8b(pixels, r, d64_1) \
1762 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1763 store_8x16b(_pixels, fb_ptr) \
1765 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1767 vec_8x16u fb_pixels; \
1768 vec_8x16u draw_mask; \
1769 vec_8x16u test_mask = psx_gpu->test_mask; \
1771 load_8x16b(fb_pixels, fb_ptr); \
1772 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1773 tst_8x16b(draw_mask, draw_mask, test_mask); \
1774 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1775 store_8x16b(fb_pixels, fb_ptr); \
1778 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1779 pixels = msb_mask; \
1780 mla_long_8x8b(pixels, r, d64_1) \
1783 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1786 vec_8x16u r_whole; \
1787 vec_8x16u g_whole; \
1788 vec_8x16u b_whole; \
1799 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1800 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1801 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1803 dup_4x32b(dx4, rgb_dx4.e[0]); \
1804 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1805 dup_4x32b(dx4, rgb_dx4.e[1]); \
1806 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1807 dup_4x32b(dx4, rgb_dx4.e[2]); \
1808 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1810 mov_narrow_8x16b(r, r_whole); \
1811 mov_narrow_8x16b(g, g_whole); \
1812 mov_narrow_8x16b(b, b_whole); \
1814 dup_4x32b(dx8, rgb_dx8.e[0]); \
1815 add_4x32b(r_block, r_block, dx8); \
1816 dup_4x32b(dx8, rgb_dx8.e[1]); \
1817 add_4x32b(g_block, g_block, dx8); \
1818 dup_4x32b(dx8, rgb_dx8.e[2]); \
1819 add_4x32b(b_block, b_block, dx8); \
1821 setup_blocks_store_shaded_untextured_##dithering(); \
1823 shr_8x8b(r, r, 3); \
1824 bic_8x8b(g, g, d64_0x07); \
1825 bic_8x8b(b, b, d64_0x07); \
1827 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1828 mla_long_8x8b(pixels, g, d64_4); \
1829 mla_long_8x8b(pixels, b, d64_128) \
1831 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1834 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1836 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1839 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1840 (_block)->draw_mask_bits = bits \
1842 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1844 vec_8x16u bits_mask; \
1845 vec_8x16u test_mask = psx_gpu->test_mask; \
1846 dup_8x16b(bits_mask, bits); \
1847 tst_8x16b(bits_mask, bits_mask, test_mask); \
1848 (_block)->draw_mask = bits_mask; \
1851 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1854 #define setup_blocks_add_blocks_indirect() \
1855 num_blocks += span_num_blocks; \
1857 if(num_blocks > MAX_BLOCKS) \
1859 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1860 flush_render_block_buffer(psx_gpu); \
1861 num_blocks = span_num_blocks; \
1862 block = psx_gpu->blocks; \
1865 #define setup_blocks_add_blocks_direct() \
1866 stats_add(texel_blocks_untextured, span_num_blocks); \
1867 span_pixel_blocks += span_num_blocks \
1870 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1871 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1872 psx_gpu_struct *psx_gpu) \
1874 setup_blocks_load_msb_mask_##target(); \
1875 setup_blocks_variables_##shading##_##texturing(target); \
1877 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1878 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1879 u32 *span_b_offset = psx_gpu->span_b_offset; \
1881 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1883 u32 num_spans = psx_gpu->num_spans; \
1888 u32 num_blocks = psx_gpu->num_blocks; \
1889 u32 span_num_blocks; \
1893 span_num_blocks = span_edge_data->num_blocks; \
1894 if(span_num_blocks) \
1896 y = span_edge_data->y; \
1897 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1899 setup_blocks_span_initialize_##shading##_##texturing(); \
1900 setup_blocks_span_initialize_##dithering(texturing); \
1902 setup_blocks_add_blocks_##target(); \
1904 s32 pixel_span = span_num_blocks * 8; \
1905 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1906 span_pixels += pixel_span; \
1908 span_num_blocks--; \
1909 while(span_num_blocks) \
1911 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1913 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1917 span_num_blocks--; \
1920 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1921 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1922 span_edge_data->right_mask); \
1928 zero_block_spans++; \
1933 span_uvrg_offset++; \
1937 psx_gpu->num_blocks = num_blocks; \
1941 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1945 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1946 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1948 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1949 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1951 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1952 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1953 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1954 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1956 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1957 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1959 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1961 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1962 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1965 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1967 block_struct *block = psx_gpu->blocks;
1968 u32 num_blocks = psx_gpu->num_blocks;
1969 stats_add(texel_blocks_4bpp, num_blocks);
1971 vec_8x8u texels_low;
1972 vec_8x8u texels_high;
1979 vec_16x8u clut_high;
1981 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1982 u16 *clut_ptr = psx_gpu->clut_ptr;
1984 // Can be done with one deinterleaving load on NEON
1985 load_8x16b(clut_a, clut_ptr);
1986 load_8x16b(clut_b, clut_ptr + 8);
1987 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
1989 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
1990 update_texture_4bpp_cache(psx_gpu);
1994 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
1995 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
1996 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
1997 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
1998 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
1999 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
2000 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
2001 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
2003 tbl_16(texels_low, texels, clut_low);
2004 tbl_16(texels_high, texels, clut_high);
2006 // Can be done with an interleaving store on NEON
2007 zip_8x16b(pixels, texels_low, texels_high);
2009 block->texels = pixels;
2016 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2018 block_struct *block = psx_gpu->blocks;
2019 u32 num_blocks = psx_gpu->num_blocks;
2021 stats_add(texel_blocks_8bpp, num_blocks);
2023 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2024 update_texture_8bpp_cache(psx_gpu);
2027 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2035 for(i = 0; i < 8; i++)
2037 offset = block->uv.e[i];
2039 texel = texture_ptr_8bpp[offset];
2040 texels.e[i] = psx_gpu->clut_ptr[texel];
2043 block->texels = texels;
2050 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2052 block_struct *block = psx_gpu->blocks;
2053 u32 num_blocks = psx_gpu->num_blocks;
2055 stats_add(texel_blocks_16bpp, num_blocks);
2059 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2065 for(i = 0; i < 8; i++)
2067 offset = block->uv.e[i];
2068 offset += ((offset & 0xFF00) * 3);
2070 texels.e[i] = texture_ptr_16bpp[offset];
2073 block->texels = texels;
2083 #define shade_blocks_load_msb_mask_indirect() \
2085 #define shade_blocks_load_msb_mask_direct() \
2086 vec_8x16u msb_mask; \
2087 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2089 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2090 block->draw_mask = _draw_mask; \
2091 block->pixels = _pixels \
2093 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2095 vec_8x16u fb_pixels; \
2096 or_8x16b(_pixels, _pixels, msb_mask); \
2097 load_8x16b(fb_pixels, block->fb_ptr); \
2098 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2099 store_8x16b(fb_pixels, block->fb_ptr); \
2103 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2104 if(psx_gpu->triangle_color == 0x808080) \
2106 false_modulated_blocks += num_blocks; \
2109 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2110 if(psx_gpu->triangle_color == 0x808080) \
2113 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2114 false_modulated_blocks += num_blocks; \
2119 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2122 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2125 u32 color = psx_gpu->triangle_color; \
2126 dup_8x8b(colors_r, color); \
2127 dup_8x8b(colors_g, color >> 8); \
2128 dup_8x8b(colors_b, color >> 16); \
2129 shade_blocks_textured_false_modulated_check_##dithering(target); \
2132 #define shade_blocks_textured_modulated_shaded_block_load() \
2133 colors_r = block->r; \
2134 colors_g = block->g; \
2135 colors_b = block->b \
2137 #define shade_blocks_textured_modulated_unshaded_block_load() \
2139 #define shade_blocks_textured_modulate_dithered(component) \
2140 pixels_##component = block->dither_offsets; \
2141 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2143 #define shade_blocks_textured_modulate_undithered(component) \
2144 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2146 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2147 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2148 psx_gpu_struct *psx_gpu) \
2150 block_struct *block = psx_gpu->blocks; \
2151 u32 num_blocks = psx_gpu->num_blocks; \
2154 vec_8x8u texels_r; \
2155 vec_8x8u texels_g; \
2156 vec_8x8u texels_b; \
2158 vec_8x8u colors_r; \
2159 vec_8x8u colors_g; \
2160 vec_8x8u colors_b; \
2162 vec_8x8u pixels_r_low; \
2163 vec_8x8u pixels_g_low; \
2164 vec_8x8u pixels_b_low; \
2167 vec_8x16u pixels_r; \
2168 vec_8x16u pixels_g; \
2169 vec_8x16u pixels_b; \
2171 vec_8x16u draw_mask; \
2172 vec_8x16u zero_mask; \
2174 vec_8x8u d64_0x07; \
2175 vec_8x8u d64_0x1F; \
2180 vec_8x16u d128_0x8000; \
2182 vec_8x16u test_mask = psx_gpu->test_mask; \
2183 u32 draw_mask_bits; \
2184 shade_blocks_load_msb_mask_##target(); \
2186 dup_8x8b(d64_0x07, 0x07); \
2187 dup_8x8b(d64_0x1F, 0x1F); \
2188 dup_8x8b(d64_1, 1); \
2189 dup_8x8b(d64_4, 4); \
2190 dup_8x8b(d64_128, 128); \
2192 dup_8x16b(d128_0x8000, 0x8000); \
2194 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2199 draw_mask_bits = block->draw_mask_bits; \
2200 dup_8x16b(draw_mask, draw_mask_bits); \
2201 tst_8x16b(draw_mask, draw_mask, test_mask); \
2203 shade_blocks_textured_modulated_##shading##_block_load(); \
2205 texels = block->texels; \
2207 mov_narrow_8x16b(texels_r, texels); \
2208 shr_narrow_8x16b(texels_g, texels, 5); \
2209 shr_narrow_8x16b(texels_b, texels, 7); \
2211 and_8x8b(texels_r, texels_r, d64_0x1F); \
2212 and_8x8b(texels_g, texels_g, d64_0x1F); \
2213 shr_8x8b(texels_b, texels_b, 3); \
2215 shade_blocks_textured_modulate_##dithering(r); \
2216 shade_blocks_textured_modulate_##dithering(g); \
2217 shade_blocks_textured_modulate_##dithering(b); \
2219 cmpeqz_8x16b(zero_mask, texels); \
2220 and_8x16b(pixels, texels, d128_0x8000); \
2222 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2223 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2224 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2226 or_8x16b(zero_mask, draw_mask, zero_mask); \
2228 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2229 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2230 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2232 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2233 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2234 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2236 shade_blocks_store_##target(zero_mask, pixels); \
2245 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2246 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2247 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2248 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2250 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2251 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2252 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2253 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2258 #define shade_blocks_textured_unmodulated_builder(target) \
2259 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2261 block_struct *block = psx_gpu->blocks; \
2262 u32 num_blocks = psx_gpu->num_blocks; \
2263 vec_8x16u draw_mask; \
2264 vec_8x16u test_mask = psx_gpu->test_mask; \
2265 u32 draw_mask_bits; \
2268 shade_blocks_load_msb_mask_##target(); \
2272 vec_8x16u zero_mask; \
2274 draw_mask_bits = block->draw_mask_bits; \
2275 dup_8x16b(draw_mask, draw_mask_bits); \
2276 tst_8x16b(draw_mask, draw_mask, test_mask); \
2278 pixels = block->texels; \
2280 cmpeqz_8x16b(zero_mask, pixels); \
2281 or_8x16b(zero_mask, draw_mask, zero_mask); \
2283 shade_blocks_store_##target(zero_mask, pixels); \
2290 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2291 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2294 block_struct *block = psx_gpu->blocks; \
2295 u32 num_blocks = psx_gpu->num_blocks; \
2296 vec_8x16u draw_mask; \
2297 vec_8x16u test_mask = psx_gpu->test_mask; \
2298 u32 draw_mask_bits; \
2301 shade_blocks_load_msb_mask_##target(); \
2305 vec_8x16u zero_mask; \
2307 draw_mask_bits = block->draw_mask_bits; \
2308 dup_8x16b(draw_mask, draw_mask_bits); \
2309 tst_8x16b(draw_mask, draw_mask, test_mask); \
2311 pixels = block->texels; \
2313 cmpeqz_8x16b(zero_mask, pixels); \
2314 or_8x16b(zero_mask, draw_mask, zero_mask); \
2316 shade_blocks_store_##target(zero_mask, pixels); \
2325 shade_blocks_textured_unmodulated_builder(indirect)
2326 shade_blocks_textured_unmodulated_builder(direct)
2328 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2332 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2334 block_struct *block = psx_gpu->blocks;
2335 u32 num_blocks = psx_gpu->num_blocks;
2337 vec_8x16u pixels = block->pixels;
2338 shade_blocks_load_msb_mask_direct();
2342 shade_blocks_store_direct(block->draw_mask, pixels);
2351 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2356 #define blend_blocks_mask_evaluate_on() \
2357 vec_8x16u mask_pixels; \
2358 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2359 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2361 #define blend_blocks_mask_evaluate_off() \
2363 #define blend_blocks_average() \
2365 vec_8x16u pixels_no_msb; \
2366 vec_8x16u fb_pixels_no_msb; \
2368 vec_8x16u d128_0x0421; \
2369 vec_8x16u d128_0x8000; \
2371 dup_8x16b(d128_0x0421, 0x0421); \
2372 dup_8x16b(d128_0x8000, 0x8000); \
2374 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2375 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2376 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2377 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2378 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2379 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2382 #define blend_blocks_add() \
2384 vec_8x16u pixels_rb, pixels_g; \
2385 vec_8x16u fb_rb, fb_g; \
2387 vec_8x16u d128_0x7C1F; \
2388 vec_8x16u d128_0x03E0; \
2390 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2391 dup_8x16b(d128_0x03E0, 0x03E0); \
2393 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2394 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2396 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2397 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2399 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2400 add_8x16b(fb_g, fb_g, pixels_g); \
2402 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2403 vector_cast(vec_16x8u, d128_0x7C1F)); \
2404 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2406 or_8x16b(blend_pixels, fb_rb, fb_g); \
2409 #define blend_blocks_subtract() \
2411 vec_8x16u pixels_rb, pixels_g; \
2412 vec_8x16u fb_rb, fb_g; \
2414 vec_8x16u d128_0x7C1F; \
2415 vec_8x16u d128_0x03E0; \
2417 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2418 dup_8x16b(d128_0x03E0, 0x03E0); \
2420 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2421 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2423 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2424 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2426 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2427 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2428 subs_8x16b(fb_g, fb_g, pixels_g); \
2430 or_8x16b(blend_pixels, fb_rb, fb_g); \
2433 #define blend_blocks_add_fourth() \
2435 vec_8x16u pixels_rb, pixels_g; \
2436 vec_8x16u pixels_fourth; \
2437 vec_8x16u fb_rb, fb_g; \
2439 vec_8x16u d128_0x7C1F; \
2440 vec_8x16u d128_0x1C07; \
2441 vec_8x16u d128_0x03E0; \
2442 vec_8x16u d128_0x00E0; \
2444 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2445 dup_8x16b(d128_0x1C07, 0x1C07); \
2446 dup_8x16b(d128_0x03E0, 0x03E0); \
2447 dup_8x16b(d128_0x00E0, 0x00E0); \
2449 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2451 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2452 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2454 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2455 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2457 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2458 add_8x16b(fb_g, fb_g, pixels_g); \
2460 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2461 vector_cast(vec_16x8u, d128_0x7C1F)); \
2462 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2464 or_8x16b(blend_pixels, fb_rb, fb_g); \
2467 #define blend_blocks_blended_combine_textured() \
2469 vec_8x16u blend_mask; \
2470 cmpltz_8x16b(blend_mask, pixels); \
2472 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2473 bif_8x16b(blend_pixels, pixels, blend_mask); \
2476 #define blend_blocks_blended_combine_untextured() \
2479 #define blend_blocks_body_blend(blend_mode, texturing) \
2481 blend_blocks_##blend_mode(); \
2482 blend_blocks_blended_combine_##texturing(); \
2485 #define blend_blocks_body_average(texturing) \
2486 blend_blocks_body_blend(average, texturing) \
2488 #define blend_blocks_body_add(texturing) \
2489 blend_blocks_body_blend(add, texturing) \
2491 #define blend_blocks_body_subtract(texturing) \
2492 blend_blocks_body_blend(subtract, texturing) \
2494 #define blend_blocks_body_add_fourth(texturing) \
2495 blend_blocks_body_blend(add_fourth, texturing) \
2497 #define blend_blocks_body_unblended(texturing) \
2498 blend_pixels = pixels \
2501 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2503 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2506 block_struct *block = psx_gpu->blocks; \
2507 u32 num_blocks = psx_gpu->num_blocks; \
2508 vec_8x16u draw_mask; \
2510 vec_8x16u blend_pixels; \
2511 vec_8x16u framebuffer_pixels; \
2512 vec_8x16u msb_mask; \
2516 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2520 pixels = block->pixels; \
2521 draw_mask = block->draw_mask; \
2522 fb_ptr = block->fb_ptr; \
2524 load_8x16b(framebuffer_pixels, fb_ptr); \
2526 blend_blocks_mask_evaluate_##mask_evaluate(); \
2527 blend_blocks_body_##blend_mode(texturing); \
2529 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2530 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2531 store_8x16b(framebuffer_pixels, fb_ptr); \
2541 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2545 blend_blocks_builder(textured, average, off);
2546 blend_blocks_builder(textured, average, on);
2547 blend_blocks_builder(textured, add, off);
2548 blend_blocks_builder(textured, add, on);
2549 blend_blocks_builder(textured, subtract, off);
2550 blend_blocks_builder(textured, subtract, on);
2551 blend_blocks_builder(textured, add_fourth, off);
2552 blend_blocks_builder(textured, add_fourth, on);
2554 blend_blocks_builder(untextured, average, off);
2555 blend_blocks_builder(untextured, average, on);
2556 blend_blocks_builder(untextured, add, off);
2557 blend_blocks_builder(untextured, add, on);
2558 blend_blocks_builder(untextured, subtract, off);
2559 blend_blocks_builder(untextured, subtract, on);
2560 blend_blocks_builder(untextured, add_fourth, off);
2561 blend_blocks_builder(untextured, add_fourth, on);
2563 blend_blocks_builder(textured, unblended, on);
2568 #define vertex_swap(_a, _b) \
2570 vertex_struct *temp_vertex = _a; \
2573 triangle_winding ^= 1; \
2577 // Setup blocks parametric-variables:
2578 // SHADE TEXTURE_MAP SWIZZLING
2585 // 8 inputs, 6 combinations
2587 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2588 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2590 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2591 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2593 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2595 setup_blocks_switch_untextured_##shading(dithering, target) \
2597 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2598 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2600 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2601 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2603 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2604 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2606 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2607 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2609 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2610 dithering, mask_evaluate) \
2611 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2613 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2615 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2617 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2619 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2621 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2622 dithering, mask_evaluate) \
2623 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2624 texture_mode, dithering) \
2626 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2627 blending, mask_evaluate) \
2628 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2629 dithering, mask_evaluate) \
2634 #define texture_blocks_switch_untextured(texture_mode) \
2635 texture_blocks_untextured \
2637 #define texture_blocks_switch_textured(texture_mode) \
2638 texture_blocks_##texture_mode \
2640 #define texture_blocks_switch(texturing, texture_mode) \
2641 texture_blocks_switch_##texturing(texture_mode) \
2644 // Shade blocks parametric-variables:
2645 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2654 // 16 inputs, 8 combinations
2656 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2657 shade_blocks_unshaded_untextured_##target \
2659 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2660 shade_blocks_textured_unmodulated_##target \
2662 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2663 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2665 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2666 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2668 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2669 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2671 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2672 shade_blocks_shaded_untextured \
2674 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2675 shade_blocks_textured_unmodulated_##target \
2677 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2678 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2680 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2681 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2683 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2684 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2686 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2688 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2690 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2692 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2694 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2696 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2698 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2699 dithering, mask_evaluate) \
2700 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2703 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2704 blending, mask_evaluate) \
2705 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2709 // Blend blocks parametric-variables:
2710 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2729 // 32 inputs, 18 combinations
2731 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2732 blend_blocks_textured_unblended_##mask_evaluate \
2734 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2735 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2737 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2738 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2741 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2742 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2744 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2746 texture_blocks_switch(texturing, texture_mode), \
2747 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2749 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2752 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2753 mask_evaluate, shading, dithering, texturing, blending) \
2754 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2755 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2756 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2757 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2759 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2760 mask_evaluate, shading, dithering, texturing) \
2761 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2762 mask_evaluate, shading, dithering, texturing, unblended), \
2763 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2764 mask_evaluate, shading, dithering, texturing, blended) \
2766 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2767 mask_evaluate, shading, dithering) \
2768 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2769 mask_evaluate, shading, dithering, untextured), \
2770 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2771 mask_evaluate, shading, dithering, textured) \
2773 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2774 mask_evaluate, shading) \
2775 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2776 mask_evaluate, shading, undithered), \
2777 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2778 mask_evaluate, shading, dithered) \
2780 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2782 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2784 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2787 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2788 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2789 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2791 #define render_blocks_switch_block_texture_mode(texture_mode) \
2792 render_blocks_switch_block_blend_mode(texture_mode, average), \
2793 render_blocks_switch_block_blend_mode(texture_mode, add), \
2794 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2795 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2797 #define render_blocks_switch_block() \
2798 render_blocks_switch_block_texture_mode(4bpp), \
2799 render_blocks_switch_block_texture_mode(8bpp), \
2800 render_blocks_switch_block_texture_mode(16bpp), \
2801 render_blocks_switch_block_texture_mode(16bpp) \
2804 render_block_handler_struct render_triangle_block_handlers[] =
2806 render_blocks_switch_block()
2809 #undef render_blocks_switch_block_modulation
2811 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2812 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2814 "texture mode: " #texture_mode "\n" \
2815 "blend mode: " #blend_mode "\n" \
2816 "mask evaluation: " #mask_evaluate "\n" \
2823 char *render_block_flag_strings[] =
2825 render_blocks_switch_block()
2829 #define triangle_y_direction_up 1
2830 #define triangle_y_direction_flat 2
2831 #define triangle_y_direction_down 0
2833 #define triangle_winding_positive 0
2834 #define triangle_winding_negative 1
2836 #define triangle_set_direction(direction_variable, value) \
2837 u32 direction_variable = (u32)(value) >> 31; \
2839 direction_variable = 2 \
2841 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2842 case (triangle_y_direction_##direction_a | \
2843 (triangle_y_direction_##direction_b << 2) | \
2844 (triangle_y_direction_##direction_c << 4) | \
2845 (triangle_winding_##winding << 6)) \
2847 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2848 vertex_struct *vertexes_out[3])
2850 s32 y_top, y_bottom;
2852 u32 triangle_winding = 0;
2854 vertex_struct *a = &(vertexes[0]);
2855 vertex_struct *b = &(vertexes[1]);
2856 vertex_struct *c = &(vertexes[2]);
2858 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2864 if(triangle_area == 0)
2886 if((y_bottom - y_top) >= 512)
2894 if(triangle_area < 0)
2896 triangle_area = -triangle_area;
2897 triangle_winding ^= 1;
2912 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2920 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2929 psx_gpu->triangle_area = triangle_area;
2930 psx_gpu->triangle_winding = triangle_winding;
2932 vertexes_out[0] = a;
2933 vertexes_out[1] = b;
2934 vertexes_out[2] = c;
2939 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2940 vertex_struct *vertex_ptrs[3], u32 flags)
2942 psx_gpu->num_spans = 0;
2944 vertex_struct *a = vertex_ptrs[0];
2945 vertex_struct *b = vertex_ptrs[1];
2946 vertex_struct *c = vertex_ptrs[2];
2948 s32 y_delta_a = b->y - a->y;
2949 s32 y_delta_b = c->y - b->y;
2950 s32 y_delta_c = c->y - a->y;
2952 triangle_set_direction(y_direction_a, y_delta_a);
2953 triangle_set_direction(y_direction_b, y_delta_b);
2954 triangle_set_direction(y_direction_c, y_delta_c);
2956 compute_all_gradients(psx_gpu, a, b, c);
2958 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2959 (psx_gpu->triangle_winding << 6))
2961 triangle_case(up, up, up, negative):
2962 triangle_case(up, up, flat, negative):
2963 triangle_case(up, up, down, negative):
2964 setup_spans_up_right(psx_gpu, a, b, c);
2967 triangle_case(flat, up, up, negative):
2968 triangle_case(flat, up, flat, negative):
2969 triangle_case(flat, up, down, negative):
2970 setup_spans_up_a(psx_gpu, a, b, c);
2973 triangle_case(down, up, up, negative):
2974 setup_spans_up_down(psx_gpu, a, c, b);
2977 triangle_case(down, up, flat, negative):
2978 setup_spans_down_a(psx_gpu, a, c, b);
2981 triangle_case(down, up, down, negative):
2982 setup_spans_down_right(psx_gpu, a, c, b);
2985 triangle_case(down, flat, up, negative):
2986 triangle_case(down, flat, flat, negative):
2987 triangle_case(down, flat, down, negative):
2988 setup_spans_down_b(psx_gpu, a, b, c);
2991 triangle_case(down, down, up, negative):
2992 triangle_case(down, down, flat, negative):
2993 triangle_case(down, down, down, negative):
2994 setup_spans_down_left(psx_gpu, a, b, c);
2997 triangle_case(up, up, up, positive):
2998 triangle_case(up, up, flat, positive):
2999 triangle_case(up, up, down, positive):
3000 setup_spans_up_left(psx_gpu, a, b, c);
3003 triangle_case(up, flat, up, positive):
3004 triangle_case(up, flat, flat, positive):
3005 triangle_case(up, flat, down, positive):
3006 setup_spans_up_b(psx_gpu, a, b, c);
3009 triangle_case(up, down, up, positive):
3010 setup_spans_up_right(psx_gpu, a, c, b);
3013 triangle_case(up, down, flat, positive):
3014 setup_spans_up_a(psx_gpu, a, c, b);
3017 triangle_case(up, down, down, positive):
3018 setup_spans_up_down(psx_gpu, a, b, c);
3021 triangle_case(flat, down, up, positive):
3022 triangle_case(flat, down, flat, positive):
3023 triangle_case(flat, down, down, positive):
3024 setup_spans_down_a(psx_gpu, a, b, c);
3027 triangle_case(down, down, up, positive):
3028 triangle_case(down, down, flat, positive):
3029 triangle_case(down, down, down, positive):
3030 setup_spans_down_right(psx_gpu, a, b, c);
3035 spans += psx_gpu->num_spans;
3038 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3042 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3044 for(i = 0; i < psx_gpu->num_spans; i++)
3046 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3047 psx_gpu->span_edge_data[i].num_blocks = 0;
3052 for(i = 0; i < psx_gpu->num_spans; i++)
3054 if(psx_gpu->span_edge_data[i].y & 1)
3055 psx_gpu->span_edge_data[i].num_blocks = 0;
3059 assert(psx_gpu->span_edge_data[0].y < 1024u);
3061 u32 render_state = flags &
3062 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3063 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3064 render_state |= psx_gpu->render_state_base;
3066 if((psx_gpu->render_state != render_state) ||
3067 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3069 psx_gpu->render_state = render_state;
3070 flush_render_block_buffer(psx_gpu);
3076 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3078 psx_gpu->render_block_handler =
3079 &(render_triangle_block_handlers[render_state]);
3080 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3084 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3087 vertex_struct *vertex_ptrs[3];
3088 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3089 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3092 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3094 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3096 block_struct *block = psx_gpu->blocks;
3097 u32 num_blocks = psx_gpu->num_blocks;
3100 vec_8x8u texel_indexes;
3102 u16 *clut_ptr = psx_gpu->clut_ptr;
3107 texel_indexes = block->r;
3109 for(i = 0; i < 8; i++)
3111 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3114 block->texels = texels;
3124 #define setup_sprite_tiled_initialize_4bpp_clut() \
3125 u16 *clut_ptr = psx_gpu->clut_ptr; \
3126 vec_8x16u clut_a, clut_b; \
3127 vec_16x8u clut_low, clut_high; \
3129 load_8x16b(clut_a, clut_ptr); \
3130 load_8x16b(clut_b, clut_ptr + 8); \
3131 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3133 #define setup_sprite_tiled_initialize_4bpp() \
3134 setup_sprite_tiled_initialize_4bpp_clut(); \
3136 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3137 update_texture_4bpp_cache(psx_gpu) \
3139 #define setup_sprite_tiled_initialize_8bpp() \
3140 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3141 update_texture_8bpp_cache(psx_gpu) \
3144 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3145 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3146 ((texture_offset + offset) & texture_mask); \
3148 load_64b(texels, texture_block_ptr) \
3151 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3152 num_blocks += tile_num_blocks; \
3153 sprite_blocks += tile_num_blocks; \
3155 if(num_blocks > MAX_BLOCKS) \
3157 flush_render_block_buffer(psx_gpu); \
3158 num_blocks = tile_num_blocks; \
3159 block = psx_gpu->blocks; \
3162 #define setup_sprite_tile_full_4bpp(edge) \
3164 vec_8x8u texels_low, texels_high; \
3166 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3168 while(sub_tile_height) \
3170 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3171 tbl_16(texels_low, texels, clut_low); \
3172 tbl_16(texels_high, texels, clut_high); \
3173 zip_8x16b(pixels, texels_low, texels_high); \
3175 block->texels = pixels; \
3176 block->draw_mask_bits = left_mask_bits; \
3177 block->fb_ptr = fb_ptr; \
3180 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3181 tbl_16(texels_low, texels, clut_low); \
3182 tbl_16(texels_high, texels, clut_high); \
3183 zip_8x16b(pixels, texels_low, texels_high); \
3185 block->texels = pixels; \
3186 block->draw_mask_bits = right_mask_bits; \
3187 block->fb_ptr = fb_ptr + 8; \
3191 texture_offset += 0x10; \
3192 sub_tile_height--; \
3194 texture_offset += 0xF00; \
3195 psx_gpu->num_blocks = num_blocks; \
3198 #define setup_sprite_tile_half_4bpp(edge) \
3200 vec_8x8u texels_low, texels_high; \
3202 setup_sprite_tile_add_blocks(sub_tile_height); \
3204 while(sub_tile_height) \
3206 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3207 tbl_16(texels_low, texels, clut_low); \
3208 tbl_16(texels_high, texels, clut_high); \
3209 zip_8x16b(pixels, texels_low, texels_high); \
3211 block->texels = pixels; \
3212 block->draw_mask_bits = edge##_mask_bits; \
3213 block->fb_ptr = fb_ptr; \
3217 texture_offset += 0x10; \
3218 sub_tile_height--; \
3220 texture_offset += 0xF00; \
3221 psx_gpu->num_blocks = num_blocks; \
3225 #define setup_sprite_tile_full_8bpp(edge) \
3227 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3229 while(sub_tile_height) \
3231 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3232 block->r = texels; \
3233 block->draw_mask_bits = left_mask_bits; \
3234 block->fb_ptr = fb_ptr; \
3237 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3238 block->r = texels; \
3239 block->draw_mask_bits = right_mask_bits; \
3240 block->fb_ptr = fb_ptr + 8; \
3244 texture_offset += 0x10; \
3245 sub_tile_height--; \
3247 texture_offset += 0xF00; \
3248 psx_gpu->num_blocks = num_blocks; \
3251 #define setup_sprite_tile_half_8bpp(edge) \
3253 setup_sprite_tile_add_blocks(sub_tile_height); \
3255 while(sub_tile_height) \
3257 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3258 block->r = texels; \
3259 block->draw_mask_bits = edge##_mask_bits; \
3260 block->fb_ptr = fb_ptr; \
3264 texture_offset += 0x10; \
3265 sub_tile_height--; \
3267 texture_offset += 0xF00; \
3268 psx_gpu->num_blocks = num_blocks; \
3272 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3273 texture_offset = texture_offset_base + 8; \
3276 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3277 texture_offset = texture_offset_base \
3279 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3280 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3282 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3283 texture_offset = texture_offset_base \
3285 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3288 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3290 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3291 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3293 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3296 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3300 sub_tile_height = column_data; \
3301 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3302 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3303 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3306 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3310 u32 tiles_remaining = column_data >> 16; \
3311 sub_tile_height = column_data & 0xFF; \
3312 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3313 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3314 tiles_remaining -= 1; \
3316 while(tiles_remaining) \
3318 sub_tile_height = 16; \
3319 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3320 tiles_remaining--; \
3323 sub_tile_height = (column_data >> 8) & 0xFF; \
3324 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3325 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3329 #define setup_sprite_column_data_single() \
3330 column_data = height \
3332 #define setup_sprite_column_data_multi() \
3333 column_data = 16 - offset_v; \
3334 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3335 column_data |= (tile_height - 1) << 16 \
3338 #define RIGHT_MASK_BIT_SHIFT 8
3339 #define RIGHT_MASK_BIT_SHIFT_4x 16
3341 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3342 edge_mode, edge, x4mode) \
3344 setup_sprite_column_data_##multi_height(); \
3345 left_mask_bits = left_block_mask | right_block_mask; \
3346 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3348 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3349 texture_mode, x4mode); \
3352 #define setup_sprite_tiled_advance_column() \
3353 texture_offset_base += 0x100; \
3354 if((texture_offset_base & 0xF00) == 0) \
3355 texture_offset_base -= (0x100 + 0xF00) \
3357 #define FB_PTR_MULTIPLIER 1
3358 #define FB_PTR_MULTIPLIER_4x 2
3360 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3361 left_mode, right_mode, x4mode) \
3363 setup_sprite_column_data_##multi_height(); \
3364 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3365 * FB_PTR_MULTIPLIER##x4mode; \
3368 left_mask_bits = left_block_mask; \
3369 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3371 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3372 texture_mode, x4mode); \
3373 fb_ptr += fb_ptr_advance_column; \
3375 left_mask_bits = 0x00; \
3376 right_mask_bits = 0x00; \
3380 setup_sprite_tiled_advance_column(); \
3381 setup_sprite_tile_column_height_##multi_height(full, none, \
3382 texture_mode, x4mode); \
3383 fb_ptr += fb_ptr_advance_column; \
3387 left_mask_bits = right_block_mask; \
3388 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3390 setup_sprite_tiled_advance_column(); \
3391 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3392 texture_mode, x4mode); \
3397 #define setup_sprite_tiled_initialize_4bpp_4x() \
3398 setup_sprite_tiled_initialize_4bpp_clut() \
3400 #define setup_sprite_tiled_initialize_8bpp_4x() \
3403 #define setup_sprite_tile_full_4bpp_4x(edge) \
3405 vec_8x8u texels_low, texels_high; \
3406 vec_8x16u pixels, pixels_wide; \
3407 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3408 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3409 u32 left_mask_bits_b = left_mask_bits >> 8; \
3410 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3411 u32 right_mask_bits_b = right_mask_bits >> 8; \
3413 while(sub_tile_height) \
3415 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3416 tbl_16(texels_low, texels, clut_low); \
3417 tbl_16(texels_high, texels, clut_high); \
3418 zip_8x16b(pixels, texels_low, texels_high); \
3420 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3421 block->texels = pixels_wide; \
3422 block->draw_mask_bits = left_mask_bits_a; \
3423 block->fb_ptr = fb_ptr; \
3426 block->texels = pixels_wide; \
3427 block->draw_mask_bits = left_mask_bits_a; \
3428 block->fb_ptr = fb_ptr + 1024; \
3431 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3432 block->texels = pixels_wide; \
3433 block->draw_mask_bits = left_mask_bits_b; \
3434 block->fb_ptr = fb_ptr + 8; \
3437 block->texels = pixels_wide; \
3438 block->draw_mask_bits = left_mask_bits_b; \
3439 block->fb_ptr = fb_ptr + 1024 + 8; \
3442 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3443 tbl_16(texels_low, texels, clut_low); \
3444 tbl_16(texels_high, texels, clut_high); \
3445 zip_8x16b(pixels, texels_low, texels_high); \
3447 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3448 block->texels = pixels_wide; \
3449 block->draw_mask_bits = right_mask_bits_a; \
3450 block->fb_ptr = fb_ptr + 16; \
3453 block->texels = pixels_wide; \
3454 block->draw_mask_bits = right_mask_bits_a; \
3455 block->fb_ptr = fb_ptr + 1024 + 16; \
3458 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3459 block->texels = pixels_wide; \
3460 block->draw_mask_bits = right_mask_bits_b; \
3461 block->fb_ptr = fb_ptr + 24; \
3464 block->texels = pixels_wide; \
3465 block->draw_mask_bits = right_mask_bits_b; \
3466 block->fb_ptr = fb_ptr + 1024 + 24; \
3470 texture_offset += 0x10; \
3471 sub_tile_height--; \
3473 texture_offset += 0xF00; \
3474 psx_gpu->num_blocks = num_blocks; \
3477 #define setup_sprite_tile_half_4bpp_4x(edge) \
3479 vec_8x8u texels_low, texels_high; \
3480 vec_8x16u pixels, pixels_wide; \
3481 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3482 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3483 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3485 while(sub_tile_height) \
3487 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3488 tbl_16(texels_low, texels, clut_low); \
3489 tbl_16(texels_high, texels, clut_high); \
3490 zip_8x16b(pixels, texels_low, texels_high); \
3492 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3493 block->texels = pixels_wide; \
3494 block->draw_mask_bits = edge##_mask_bits_a; \
3495 block->fb_ptr = fb_ptr; \
3498 block->texels = pixels_wide; \
3499 block->draw_mask_bits = edge##_mask_bits_a; \
3500 block->fb_ptr = fb_ptr + 1024; \
3503 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3504 block->texels = pixels_wide; \
3505 block->draw_mask_bits = edge##_mask_bits_b; \
3506 block->fb_ptr = fb_ptr + 8; \
3509 block->texels = pixels_wide; \
3510 block->draw_mask_bits = edge##_mask_bits_b; \
3511 block->fb_ptr = fb_ptr + 1024 + 8; \
3515 texture_offset += 0x10; \
3516 sub_tile_height--; \
3518 texture_offset += 0xF00; \
3519 psx_gpu->num_blocks = num_blocks; \
3523 #define setup_sprite_tile_full_8bpp_4x(edge) \
3525 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3526 vec_16x8u texels_wide; \
3527 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3528 u32 left_mask_bits_b = left_mask_bits >> 8; \
3529 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3530 u32 right_mask_bits_b = right_mask_bits >> 8; \
3532 while(sub_tile_height) \
3534 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3535 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3536 block->r = texels_wide.low; \
3537 block->draw_mask_bits = left_mask_bits_a; \
3538 block->fb_ptr = fb_ptr; \
3541 block->r = texels_wide.low; \
3542 block->draw_mask_bits = left_mask_bits_a; \
3543 block->fb_ptr = fb_ptr + 1024; \
3546 block->r = texels_wide.high; \
3547 block->draw_mask_bits = left_mask_bits_b; \
3548 block->fb_ptr = fb_ptr + 8; \
3551 block->r = texels_wide.high; \
3552 block->draw_mask_bits = left_mask_bits_b; \
3553 block->fb_ptr = fb_ptr + 1024 + 8; \
3556 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3557 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3558 block->r = texels_wide.low; \
3559 block->draw_mask_bits = right_mask_bits_a; \
3560 block->fb_ptr = fb_ptr + 16; \
3563 block->r = texels_wide.low; \
3564 block->draw_mask_bits = right_mask_bits_a; \
3565 block->fb_ptr = fb_ptr + 1024 + 16; \
3568 block->r = texels_wide.high; \
3569 block->draw_mask_bits = right_mask_bits_b; \
3570 block->fb_ptr = fb_ptr + 24; \
3573 block->r = texels_wide.high; \
3574 block->draw_mask_bits = right_mask_bits_b; \
3575 block->fb_ptr = fb_ptr + 24 + 1024; \
3579 texture_offset += 0x10; \
3580 sub_tile_height--; \
3582 texture_offset += 0xF00; \
3583 psx_gpu->num_blocks = num_blocks; \
3586 #define setup_sprite_tile_half_8bpp_4x(edge) \
3588 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3589 vec_16x8u texels_wide; \
3590 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3591 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3593 while(sub_tile_height) \
3595 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3596 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3597 block->r = texels_wide.low; \
3598 block->draw_mask_bits = edge##_mask_bits_a; \
3599 block->fb_ptr = fb_ptr; \
3602 block->r = texels_wide.low; \
3603 block->draw_mask_bits = edge##_mask_bits_a; \
3604 block->fb_ptr = fb_ptr + 1024; \
3607 block->r = texels_wide.high; \
3608 block->draw_mask_bits = edge##_mask_bits_b; \
3609 block->fb_ptr = fb_ptr + 8; \
3612 block->r = texels_wide.high; \
3613 block->draw_mask_bits = edge##_mask_bits_b; \
3614 block->fb_ptr = fb_ptr + 8 + 1024; \
3618 texture_offset += 0x10; \
3619 sub_tile_height--; \
3621 texture_offset += 0xF00; \
3622 psx_gpu->num_blocks = num_blocks; \
3626 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3627 texture_offset = texture_offset_base + 8; \
3630 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3631 texture_offset = texture_offset_base \
3633 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3634 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3636 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3637 texture_offset = texture_offset_base \
3639 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3642 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3644 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3645 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3647 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3650 #define setup_sprite_offset_u_adjust() \
3652 #define setup_sprite_comapre_left_block_mask() \
3653 ((left_block_mask & 0xFF) == 0xFF) \
3655 #define setup_sprite_comapre_right_block_mask() \
3656 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3659 #define setup_sprite_offset_u_adjust_4x() \
3661 offset_u_right = offset_u_right * 2 + 1 \
3663 #define setup_sprite_comapre_left_block_mask_4x() \
3664 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3666 #define setup_sprite_comapre_right_block_mask_4x() \
3667 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3670 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3671 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3672 s32 u, s32 v, s32 width, s32 height, u32 color) \
3674 s32 offset_u = u & 0xF; \
3675 s32 offset_v = v & 0xF; \
3677 s32 width_rounded = offset_u + width + 15; \
3678 s32 height_rounded = offset_v + height + 15; \
3679 s32 tile_height = height_rounded / 16; \
3680 s32 tile_width = width_rounded / 16; \
3681 u32 offset_u_right = width_rounded & 0xF; \
3683 setup_sprite_offset_u_adjust##x4mode(); \
3685 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3686 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3688 u32 left_mask_bits; \
3689 u32 right_mask_bits; \
3691 u32 sub_tile_height; \
3694 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3695 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3696 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3697 ((psx_gpu->texture_mask_height >> 4) << 12); \
3698 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3699 ((v & 0xF0) << 8); \
3700 u32 texture_offset_base = texture_offset; \
3703 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3704 u32 num_blocks = psx_gpu->num_blocks; \
3705 block_struct *block = psx_gpu->blocks + num_blocks; \
3707 u8 *texture_block_ptr; \
3710 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3712 control_mask = tile_width == 1; \
3713 control_mask |= (tile_height == 1) << 1; \
3714 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3715 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3717 sprites_##texture_mode++; \
3719 switch(control_mask) \
3723 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3728 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3733 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3738 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3743 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3748 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3753 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3758 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3763 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3768 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3773 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3778 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3783 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3788 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3795 setup_sprite_tiled_builder(4bpp,);
3796 setup_sprite_tiled_builder(8bpp,);
3798 setup_sprite_tiled_builder(4bpp,_4x);
3799 setup_sprite_tiled_builder(8bpp,_4x);
3802 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3804 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3805 s32 v, s32 width, s32 height, u32 color)
3807 u32 left_offset = u & 0x7;
3808 u32 width_rounded = width + left_offset + 7;
3810 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3811 u32 right_width = width_rounded & 0x7;
3812 u32 block_width = width_rounded / 8;
3813 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3815 u32 left_mask_bits = ~(0xFF << left_offset);
3816 u32 right_mask_bits = 0xFE << right_width;
3818 u32 texture_offset_base = u + (v * 1024);
3820 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3822 u32 blocks_remaining;
3823 u32 num_blocks = psx_gpu->num_blocks;
3824 block_struct *block = psx_gpu->blocks + num_blocks;
3826 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3827 u16 *texture_block_ptr;
3829 texture_offset_base &= ~0x7;
3831 stats_add(sprites_16bpp, 1);
3833 if(block_width == 1)
3835 u32 mask_bits = left_mask_bits | right_mask_bits;
3842 if(num_blocks > MAX_BLOCKS)
3844 flush_render_block_buffer(psx_gpu);
3846 block = psx_gpu->blocks;
3850 texture_page_ptr + (texture_offset_base & texture_mask);
3852 block->texels = *(vec_8x16u *)texture_block_ptr;
3853 block->draw_mask_bits = mask_bits;
3854 block->fb_ptr = fb_ptr;
3858 texture_offset_base += 1024;
3862 psx_gpu->num_blocks = num_blocks;
3871 blocks_remaining = block_width - 2;
3872 num_blocks += block_width;
3873 sprite_blocks += block_width;
3875 if(num_blocks > MAX_BLOCKS)
3877 flush_render_block_buffer(psx_gpu);
3878 num_blocks = block_width;
3879 block = psx_gpu->blocks;
3882 texture_offset = texture_offset_base;
3883 texture_offset_base += 1024;
3885 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3886 block->texels = *(vec_8x16u *)texture_block_ptr;
3888 block->draw_mask_bits = left_mask_bits;
3889 block->fb_ptr = fb_ptr;
3891 texture_offset += 8;
3895 while(blocks_remaining)
3897 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3898 block->texels = *(vec_8x16u *)texture_block_ptr;
3900 block->draw_mask_bits = 0;
3901 block->fb_ptr = fb_ptr;
3903 texture_offset += 8;
3910 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3911 block->texels = *(vec_8x16u *)texture_block_ptr;
3913 block->draw_mask_bits = right_mask_bits;
3914 block->fb_ptr = fb_ptr;
3916 fb_ptr += fb_ptr_pitch;
3920 psx_gpu->num_blocks = num_blocks;
3929 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3930 s32 v, s32 width, s32 height, u32 color)
3932 u32 right_width = ((width - 1) & 0x7) + 1;
3933 u32 right_mask_bits = (0xFF << right_width);
3934 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3935 u32 block_width = (width + 7) / 8;
3936 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3937 u32 blocks_remaining;
3938 u32 num_blocks = psx_gpu->num_blocks;
3939 block_struct *block = psx_gpu->blocks + num_blocks;
3941 u32 color_r = color & 0xFF;
3942 u32 color_g = (color >> 8) & 0xFF;
3943 u32 color_b = (color >> 16) & 0xFF;
3945 vec_8x16u right_mask;
3946 vec_8x16u test_mask = psx_gpu->test_mask;
3947 vec_8x16u zero_mask;
3949 sprites_untextured++;
3951 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3953 dup_8x16b(colors, color);
3954 dup_8x16b(zero_mask, 0x00);
3955 dup_8x16b(right_mask, right_mask_bits);
3956 tst_8x16b(right_mask, right_mask, test_mask);
3960 blocks_remaining = block_width - 1;
3961 num_blocks += block_width;
3964 sprite_blocks += block_width;
3967 if(num_blocks > MAX_BLOCKS)
3969 flush_render_block_buffer(psx_gpu);
3970 num_blocks = block_width;
3971 block = psx_gpu->blocks;
3974 while(blocks_remaining)
3976 block->pixels = colors;
3977 block->draw_mask = zero_mask;
3978 block->fb_ptr = fb_ptr;
3985 block->pixels = colors;
3986 block->draw_mask = right_mask;
3987 block->fb_ptr = fb_ptr;
3990 fb_ptr += fb_ptr_pitch;
3993 psx_gpu->num_blocks = num_blocks;
3999 static void __attribute__((noinline))
4000 setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4001 s32 v, s32 width, s32 height, u32 color)
4003 u32 r = color & 0xFF;
4004 u32 g = (color >> 8) & 0xFF;
4005 u32 b = (color >> 16) & 0xFF;
4006 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4008 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4010 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4015 if(psx_gpu->num_blocks)
4017 flush_render_block_buffer(psx_gpu);
4024 vram_ptr = (void *)vram_ptr16;
4025 if((uintptr_t)vram_ptr16 & 2)
4027 *vram_ptr16 = color_32bpp;
4028 vram_ptr = (void *)(vram_ptr16 + 1);
4032 while(num_width >= 4 * 2)
4034 vram_ptr[0] = color_32bpp;
4035 vram_ptr[1] = color_32bpp;
4036 vram_ptr[2] = color_32bpp;
4037 vram_ptr[3] = color_32bpp;
4043 while(num_width >= 2)
4045 *vram_ptr++ = color_32bpp;
4051 *(u16 *)vram_ptr = color_32bpp;
4059 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4060 s32 v, s32 width, s32 height, u32 color);
4062 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4063 s32 v, s32 width, s32 height, u32 color)
4065 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
4066 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
4067 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
4069 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
4075 s32 w1 = width > 512 ? 512 : width;
4076 setup_sprite_untextured_512(psx_gpu, x, y, 0, 0, w1, height, color);
4083 #define setup_sprite_blocks_switch_textured(texture_mode) \
4084 setup_sprite_##texture_mode \
4086 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4087 setup_sprite_untextured \
4089 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4090 setup_sprite_blocks_switch_##texturing(texture_mode) \
4093 #define texture_sprite_blocks_switch_4bpp() \
4094 texture_blocks_untextured \
4096 #define texture_sprite_blocks_switch_8bpp() \
4097 texture_sprite_blocks_8bpp \
4099 #define texture_sprite_blocks_switch_16bpp() \
4100 texture_blocks_untextured \
4102 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4103 texture_blocks_untextured \
4105 #define texture_sprite_blocks_switch_textured(texture_mode) \
4106 texture_sprite_blocks_switch_##texture_mode() \
4108 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4109 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4111 setup_sprite_blocks_switch(texturing, texture_mode), \
4112 texture_sprite_blocks_switch_##texturing(texture_mode), \
4113 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4115 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4118 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4119 mask_evaluate, shading, dithering, texturing, blending) \
4120 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4121 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4122 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4123 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4125 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4126 mask_evaluate, shading, dithering, texturing) \
4127 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4128 mask_evaluate, shading, dithering, texturing, unblended), \
4129 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4130 mask_evaluate, shading, dithering, texturing, blended) \
4132 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4133 mask_evaluate, shading, dithering) \
4134 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4135 mask_evaluate, shading, dithering, untextured), \
4136 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4137 mask_evaluate, shading, dithering, textured) \
4139 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4140 mask_evaluate, shading) \
4141 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4142 mask_evaluate, shading, undithered), \
4143 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4144 mask_evaluate, shading, dithered) \
4146 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4147 blend_mode, mask_evaluate) \
4148 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4149 mask_evaluate, unshaded), \
4150 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4151 mask_evaluate, shaded) \
4153 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4154 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4156 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4159 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4160 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4161 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4162 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4163 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4165 #define render_sprite_blocks_switch_block() \
4166 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4167 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4168 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4169 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4172 render_block_handler_struct render_sprite_block_handlers[] =
4174 render_sprite_blocks_switch_block()
4178 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4179 s32 width, s32 height, u32 flags, u32 color)
4181 s32 x_right = x + width - 1;
4182 s32 y_bottom = y + height - 1;
4188 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4194 if(x < psx_gpu->viewport_start_x)
4196 u32 clip = psx_gpu->viewport_start_x - x;
4202 if(y < psx_gpu->viewport_start_y)
4204 s32 clip = psx_gpu->viewport_start_y - y;
4210 if(x_right > psx_gpu->viewport_end_x)
4211 width -= x_right - psx_gpu->viewport_end_x;
4213 if(y_bottom > psx_gpu->viewport_end_y)
4214 height -= y_bottom - psx_gpu->viewport_end_y;
4216 if((width <= 0) || (height <= 0))
4220 span_pixels += width * height;
4224 u32 render_state = flags &
4225 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4226 RENDER_FLAGS_TEXTURE_MAP);
4228 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4230 if((psx_gpu->render_state != render_state) ||
4231 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4233 psx_gpu->render_state = render_state;
4234 flush_render_block_buffer(psx_gpu);
4240 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4244 if(psx_gpu->triangle_color != color)
4246 flush_render_block_buffer(psx_gpu);
4247 psx_gpu->triangle_color = color;
4250 if(color == 0x808080)
4251 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4253 render_block_handler_struct *render_block_handler =
4254 &(render_sprite_block_handlers[render_state]);
4255 psx_gpu->render_block_handler = render_block_handler;
4257 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4258 (psx_gpu, x, y, u, v, width, height, color);
4261 #define draw_pixel_line_mask_evaluate_yes() \
4262 if((*vram_ptr & 0x8000) == 0) \
4264 #define draw_pixel_line_mask_evaluate_no() \
4267 #define draw_pixel_line_shaded() \
4269 color_r = fixed_to_int(current_r); \
4270 color_g = fixed_to_int(current_g); \
4271 color_b = fixed_to_int(current_b); \
4273 current_r += gradient_r; \
4274 current_g += gradient_g; \
4275 current_b += gradient_b; \
4278 #define draw_pixel_line_unshaded() \
4280 color_r = color & 0xFF; \
4281 color_g = (color >> 8) & 0xFF; \
4282 color_b = (color >> 16) & 0xFF; \
4286 #define draw_pixel_line_dithered(_x, _y) \
4288 u32 dither_xor = _x ^ _y; \
4289 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4290 dither_offset |= (_y & 0x1) << 1; \
4291 dither_offset |= (dither_xor & 0x1) << 2; \
4292 dither_offset -= 4; \
4294 color_r += dither_offset; \
4295 color_g += dither_offset; \
4296 color_b += dither_offset; \
4317 #define draw_pixel_line_undithered(_x, _y) \
4320 #define draw_pixel_line_average() \
4321 color_r = (color_r + fb_r) / 2; \
4322 color_g = (color_g + fb_g) / 2; \
4323 color_b = (color_b + fb_b) / 2 \
4325 #define draw_pixel_line_add() \
4340 #define draw_pixel_line_subtract() \
4341 color_r = fb_r - color_r; \
4342 color_g = fb_g - color_g; \
4343 color_b = fb_b - color_b; \
4354 #define draw_pixel_line_add_fourth() \
4355 color_r = fb_r + (color_r / 4); \
4356 color_g = fb_g + (color_g / 4); \
4357 color_b = fb_b + (color_b / 4); \
4369 #define draw_pixel_line_blended(blend_mode) \
4370 s32 fb_pixel = *vram_ptr; \
4371 s32 fb_r = fb_pixel & 0x1F; \
4372 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4373 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4375 draw_pixel_line_##blend_mode() \
4377 #define draw_pixel_line_unblended(blend_mode) \
4380 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4382 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4383 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4385 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4387 draw_pixel_line_##shading(); \
4388 draw_pixel_line_##dithering(_x, _y); \
4394 draw_pixel_line_##blending(blend_mode); \
4396 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4397 psx_gpu->mask_msb; \
4401 #define update_increment(value) \
4404 #define update_decrement(value) \
4407 #define update_vram_row_increment(value) \
4410 #define update_vram_row_decrement(value) \
4413 #define compare_increment(a, b) \
4416 #define compare_decrement(a, b) \
4419 #define set_line_gradients(minor) \
4421 s32 gradient_divisor = delta_##minor; \
4422 if(gradient_divisor != 0) \
4424 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4425 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4426 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4434 current_r = fixed_center(vertex_a->r); \
4435 current_g = fixed_center(vertex_a->g); \
4436 current_b = fixed_center(vertex_a->b); \
4439 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4440 mask_evaluate, blend_mode) \
4443 error_step = delta_y * 2; \
4444 error_wrap = delta_x * 2; \
4448 set_line_gradients(x); \
4450 for(current_x = x_a; current_x <= x_b; current_x++) \
4452 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4453 mask_evaluate, blend_mode); \
4454 error += error_step; \
4457 if(error >= error_wrap) \
4459 update_##direction(current_y); \
4460 update_vram_row_##direction(); \
4461 error -= error_wrap; \
4466 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4467 mask_evaluate, blend_mode) \
4470 error_step = delta_x * 2; \
4471 error_wrap = delta_y * 2; \
4475 set_line_gradients(y); \
4477 for(current_y = y_a; compare_##direction(current_y, y_b); \
4478 update_##direction(current_y)) \
4480 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4481 mask_evaluate, blend_mode); \
4482 error += error_step; \
4483 update_vram_row_##direction(); \
4485 if(error > error_wrap) \
4489 error -= error_wrap; \
4495 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4501 if(delta_x > delta_y) \
4503 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4504 mask_evaluate, blend_mode); \
4508 draw_line_span_vertical(decrement, shading, blending, dithering, \
4509 mask_evaluate, blend_mode); \
4514 if(delta_x > delta_y) \
4516 draw_line_span_horizontal(increment, shading, blending, dithering, \
4517 mask_evaluate, blend_mode); \
4521 draw_line_span_vertical(increment, shading, blending, dithering, \
4522 mask_evaluate, blend_mode); \
4527 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4528 u32 color, int double_resolution)
4530 s32 color_r, color_g, color_b;
4531 u32 triangle_winding = 0;
4533 fixed_type gradient_r = 0;
4534 fixed_type gradient_g = 0;
4535 fixed_type gradient_b = 0;
4536 fixed_type current_r = 0;
4537 fixed_type current_g = 0;
4538 fixed_type current_b = 0;
4543 s32 delta_x, delta_y;
4554 flush_render_block_buffer(psx_gpu);
4555 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4557 vertex_struct *vertex_a = &(vertexes[0]);
4558 vertex_struct *vertex_b = &(vertexes[1]);
4566 if(vertex_a->x >= vertex_b->x)
4568 vertex_swap(vertex_a, vertex_b);
4577 delta_x = x_b - x_a;
4578 delta_y = y_b - y_a;
4580 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4583 if(double_resolution)
4593 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4595 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4599 if(flags & RENDER_FLAGS_SHADE)
4600 control_mask |= 0x1;
4602 if(flags & RENDER_FLAGS_BLEND)
4604 control_mask |= 0x2;
4605 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4608 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4609 control_mask |= 0x4;
4611 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4612 control_mask |= 0x8;
4614 switch(control_mask)
4617 render_line_body(unshaded, unblended, undithered, no, none);
4621 render_line_body(shaded, unblended, undithered, no, none);
4625 render_line_body(unshaded, blended, undithered, no, average);
4629 render_line_body(shaded, blended, undithered, no, average);
4633 render_line_body(unshaded, unblended, dithered, no, none);
4637 render_line_body(shaded, unblended, dithered, no, none);
4641 render_line_body(unshaded, blended, dithered, no, average);
4645 render_line_body(shaded, blended, dithered, no, average);
4649 render_line_body(unshaded, unblended, undithered, yes, none);
4653 render_line_body(shaded, unblended, undithered, yes, none);
4657 render_line_body(unshaded, blended, undithered, yes, average);
4661 render_line_body(shaded, blended, undithered, yes, average);
4665 render_line_body(unshaded, unblended, dithered, yes, none);
4669 render_line_body(shaded, unblended, dithered, yes, none);
4673 render_line_body(unshaded, blended, dithered, yes, average);
4677 render_line_body(shaded, blended, dithered, yes, average);
4681 render_line_body(unshaded, blended, undithered, no, add);
4685 render_line_body(shaded, blended, undithered, no, add);
4689 render_line_body(unshaded, blended, dithered, no, add);
4693 render_line_body(shaded, blended, dithered, no, add);
4697 render_line_body(unshaded, blended, undithered, yes, add);
4701 render_line_body(shaded, blended, undithered, yes, add);
4705 render_line_body(unshaded, blended, dithered, yes, add);
4709 render_line_body(shaded, blended, dithered, yes, add);
4713 render_line_body(unshaded, blended, undithered, no, subtract);
4717 render_line_body(shaded, blended, undithered, no, subtract);
4721 render_line_body(unshaded, blended, dithered, no, subtract);
4725 render_line_body(shaded, blended, dithered, no, subtract);
4729 render_line_body(unshaded, blended, undithered, yes, subtract);
4733 render_line_body(shaded, blended, undithered, yes, subtract);
4737 render_line_body(unshaded, blended, dithered, yes, subtract);
4741 render_line_body(shaded, blended, dithered, yes, subtract);
4745 render_line_body(unshaded, blended, undithered, no, add_fourth);
4749 render_line_body(shaded, blended, undithered, no, add_fourth);
4753 render_line_body(unshaded, blended, dithered, no, add_fourth);
4757 render_line_body(shaded, blended, dithered, no, add_fourth);
4761 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4765 render_line_body(shaded, blended, undithered, yes, add_fourth);
4769 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4773 render_line_body(shaded, blended, dithered, yes, add_fourth);
4779 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4780 u32 width, u32 height)
4782 if((width == 0) || (height == 0))
4785 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4787 u32 r = color & 0xFF;
4788 u32 g = (color >> 8) & 0xFF;
4789 u32 b = (color >> 16) & 0xFF;
4790 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4792 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4794 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4796 u32 pitch = 512 - (width / 2);
4799 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4804 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4813 vram_ptr[0] = color_32bpp;
4814 vram_ptr[1] = color_32bpp;
4815 vram_ptr[2] = color_32bpp;
4816 vram_ptr[3] = color_32bpp;
4817 vram_ptr[4] = color_32bpp;
4818 vram_ptr[5] = color_32bpp;
4819 vram_ptr[6] = color_32bpp;
4820 vram_ptr[7] = color_32bpp;
4831 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4832 u32 width, u32 height)
4834 if((width == 0) || (height == 0))
4840 u32 r = color & 0xFF;
4841 u32 g = (color >> 8) & 0xFF;
4842 u32 b = (color >> 16) & 0xFF;
4843 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4845 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4847 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4849 u32 pitch = 1024 / 2 - (width / 2);
4857 vram_ptr[0] = color_32bpp;
4858 vram_ptr[1] = color_32bpp;
4859 vram_ptr[2] = color_32bpp;
4860 vram_ptr[3] = color_32bpp;
4861 vram_ptr[4] = color_32bpp;
4862 vram_ptr[5] = color_32bpp;
4863 vram_ptr[6] = color_32bpp;
4864 vram_ptr[7] = color_32bpp;
4875 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4876 u32 width, u32 height, u32 pitch)
4878 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4880 u32 mask_msb = psx_gpu->mask_msb;
4882 if((width == 0) || (height == 0))
4885 flush_render_block_buffer(psx_gpu);
4886 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4888 for(draw_y = 0; draw_y < height; draw_y++)
4890 for(draw_x = 0; draw_x < width; draw_x++)
4892 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4900 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4901 u32 dest_x, u32 dest_y, u32 width, u32 height)
4903 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4904 dest_x, dest_y, width, height, 1024);
4908 void initialize_reciprocal_table(void)
4911 u32 height_normalized;
4912 u32 height_reciprocal;
4915 for(height = 1; height < sizeof(reciprocal_table)
4916 / sizeof(reciprocal_table[0]); height++)
4918 shift = __builtin_clz(height);
4919 height_normalized = height << shift;
4920 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4923 shift = 32 - (51 - shift);
4925 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4930 #define dither_table_row(a, b, c, d) \
4931 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4933 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4935 vec_8x16u test_mask =
4936 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4938 psx_gpu->test_mask = test_mask;
4940 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4941 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4942 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4943 psx_gpu->viewport_mask = 0;
4944 psx_gpu->current_texture_page = 0;
4945 psx_gpu->current_texture_mask = 0;
4946 psx_gpu->last_8bpp_texture_page = 0;
4948 psx_gpu->clut_settings = 0;
4949 psx_gpu->texture_settings = 0;
4950 psx_gpu->render_state = 0;
4951 psx_gpu->render_state_base = 0;
4952 psx_gpu->num_blocks = 0;
4953 psx_gpu->uvrgb_phase = 0x8000;
4955 psx_gpu->vram_ptr = vram;
4956 psx_gpu->vram_out_ptr = vram;
4958 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4959 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4960 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4962 psx_gpu->mask_msb = 0;
4964 psx_gpu->texture_window_x = 0;
4965 psx_gpu->texture_window_y = 0;
4966 psx_gpu->texture_mask_width = 0xFF;
4967 psx_gpu->texture_mask_height = 0xFF;
4969 psx_gpu->render_mode = 0;
4971 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4973 initialize_reciprocal_table();
4974 psx_gpu->reciprocal_table_ptr = reciprocal_table;
4981 // (minus ones(4) * 4)
4983 // d0: (1 3 5 7): x1 ^ y1
4984 // d1: (2 3 6 7): y0
4985 // d2: (4 5 6 7): x0 ^ y0
4987 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
4988 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
4989 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
4990 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
4992 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
4994 psx_gpu->saved_hres = 256;
5000 gettimeofday(&tv, NULL);
5002 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
5005 #if 0 //def NEON_BUILD
5010 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
5015 void init_counter(void)
5018 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
5019 value |= 5; // master enable, ccnt reset
5020 value &= ~8; // ccnt divider 0
5021 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
5022 // enable cycle counter
5023 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
5026 void triangle_benchmark(psx_gpu_struct *psx_gpu)
5033 const u32 iterations = 500000;
5035 psx_gpu->num_blocks = 64;
5036 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5038 for(i = 0; i < 64; i++)
5040 memset(&(psx_gpu->blocks[i].r), 0, 16);
5045 ticks = get_counter();
5047 for(i = 0; i < iterations; i++)
5049 texture_sprite_blocks_8bpp(psx_gpu);
5052 ticks_elapsed = get_counter() - ticks;
5054 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5059 #include "psx_gpu_4x.c"