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.
21 #include "vector_ops.h"
23 #include "psx_gpu_simd.h"
26 u32 span_pixel_blocks = 0;
32 u32 sprites_16bpp = 0;
33 u32 sprite_blocks = 0;
34 u32 sprites_untextured = 0;
36 u32 trivial_rejects = 0;
40 u32 texel_blocks_4bpp = 0;
41 u32 texel_blocks_8bpp = 0;
42 u32 texel_blocks_16bpp = 0;
43 u32 texel_blocks_untextured = 0;
45 u32 render_buffer_flushes = 0;
46 u32 state_changes = 0;
47 u32 left_split_triangles = 0;
48 u32 flat_triangles = 0;
49 u32 clipped_triangles = 0;
50 u32 zero_block_spans = 0;
51 u32 texture_cache_loads = 0;
52 u32 false_modulated_blocks = 0;
54 #define stats_add(stat, count) // stat += count
56 /* double size for enhancement */
57 u32 reciprocal_table[512 * 2];
60 typedef s32 fixed_type;
62 #define EDGE_STEP_BITS 32
65 #define fixed_center(value) \
66 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
68 #define int_to_fixed(value) \
69 (((fixed_type)(value)) << FIXED_BITS) \
71 #define fixed_to_int(value) \
72 ((value) >> FIXED_BITS) \
74 #define fixed_to_double(value) \
75 ((value) / (double)(1 << FIXED_BITS)) \
77 #define double_to_fixed(value) \
78 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
80 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
81 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
82 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
83 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
85 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
86 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
88 struct render_block_handler_struct
91 texture_blocks_function_type *texture_blocks;
92 shade_blocks_function_type *shade_blocks;
93 blend_blocks_function_type *blend_blocks;
98 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
100 u32 shift = __builtin_clz(denominator);
101 u32 denominator_normalized = denominator << shift;
103 double numerator = (1ULL << 62) + denominator_normalized;
106 double denominator_normalized_dp_b;
107 u64 denominator_normalized_dp_u64;
110 double reciprocal_dp;
112 u64 numerator_u64 = (denominator_normalized >> 10) |
113 ((u64)(62 + 1023) << 52);
114 *((u64 *)(&numerator_b)) = numerator_u64;
116 denominator_normalized_dp_u64 =
117 (u64)(denominator_normalized << 21) |
118 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
119 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
121 // Implement with a DP divide
122 reciprocal_dp = numerator / denominator_normalized_dp_b;
123 reciprocal = reciprocal_dp;
125 if(reciprocal == 0x80000001)
126 reciprocal = 0x80000000;
128 *_shift = 62 - shift;
132 double reciprocal_estimate(double a)
137 q = (int)(a * 512.0);
138 /* a in units of 1/512 rounded down */
139 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
140 s = (int)(256.0 * r + 0.5);
142 /* r in units of 1/256 rounded to nearest */
144 return (double)s / 256.0;
147 u32 reciprocal_estimate_u32(u32 value)
150 volatile double dp_value;
151 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
153 if((value >> 31) == 0)
156 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
158 *dp_value_ptr = dp_value_u64;
160 dp_value = reciprocal_estimate(dp_value);
161 dp_value_u64 = *dp_value_ptr;
163 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
166 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
168 u32 shift = __builtin_clz(value);
169 u32 value_normalized = value << shift;
171 *_shift = 62 - shift;
173 value_normalized -= 2;
175 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
177 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
178 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
179 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
180 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
181 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
182 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
184 return reciprocal_normalized;
190 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
192 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
195 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
197 s32 coverage_x, coverage_y;
202 coverage_x = x2 >> 6;
203 coverage_y = y2 >> 8;
211 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
214 mask_down_right |= mask_down_right << 16;
216 coverage_x = x1 >> 6;
218 mask_up_left = 0xFFFF0000 << coverage_x;
220 mask_up_left = 0xFFFF0000;
222 coverage_y = y1 >> 8;
224 mask_up_left |= mask_up_left >> 16;
226 return mask_up_left & mask_down_right;
229 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
232 u32 mask = texture_region_mask(x1, y1, x2, y2);
234 psx_gpu->dirty_textures_4bpp_mask |= mask;
235 psx_gpu->dirty_textures_8bpp_mask |= mask;
236 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
241 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
242 u32 y1, u32 x2, u32 y2)
244 u32 mask = texture_region_mask(x1, y1, x2, y2) &
245 psx_gpu->viewport_mask;
247 psx_gpu->dirty_textures_4bpp_mask |= mask;
248 psx_gpu->dirty_textures_8bpp_mask |= mask;
249 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
254 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
257 u32 mask = texture_region_mask(x1, y1, x2, y2);
259 u8 *texture_page_ptr;
264 psx_gpu->dirty_textures_8bpp_mask |= mask;
265 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
267 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
268 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
270 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
271 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
272 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
273 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
281 texel_block = *vram_ptr;
283 texture_page_ptr[0] = texel_block & 0xF;
284 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
285 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
286 texture_page_ptr[3] = texel_block >> 12;
289 texture_page_ptr += 4;
303 psx_gpu->dirty_textures_4bpp_mask |= mask;
309 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
311 u32 current_texture_page = psx_gpu->current_texture_page;
312 u8 *texture_page_ptr = psx_gpu->texture_page_base;
313 u16 *vram_ptr = psx_gpu->vram_ptr;
319 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
320 vram_ptr += (current_texture_page & 0xF) * 64;
322 texture_cache_loads++;
329 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
339 texel_block = *vram_ptr;
341 texture_page_ptr[0] = texel_block & 0xF;
342 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
343 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
344 texture_page_ptr[3] = texel_block >> 12;
347 texture_page_ptr += 4;
361 vram_ptr -= (1024 * 16) - 4;
367 vram_ptr += (16 * 1024) - (4 * 16);
372 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
375 u16 *texture_page_ptr = psx_gpu->texture_page_base;
376 u16 *vram_ptr = psx_gpu->vram_ptr;
383 texture_cache_loads++;
385 vram_ptr += (texture_page >> 4) * 256 * 1024;
386 vram_ptr += (texture_page & 0xF) * 64;
388 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
389 texture_page_ptr += (8 * 16) * 8;
402 load_128b(texels, vram_ptr);
403 store_128b(texels, texture_page_ptr);
405 texture_page_ptr += 8;
413 vram_ptr -= (1024 * 16);
422 vram_ptr += (16 * 1024);
424 texture_page_ptr += (8 * 16) * 8;
432 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
434 u32 current_texture_page = psx_gpu->current_texture_page;
435 u32 update_textures =
436 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
438 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
440 if(update_textures & (1 << current_texture_page))
442 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
443 update_textures &= ~(1 << current_texture_page);
448 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
449 (current_texture_page & 0x10);
451 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
455 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
457 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
458 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
460 u32 num_blocks_dest = 0;
461 block_struct *block_src = psx_gpu->blocks;
462 block_struct *block_dest = psx_gpu->blocks;
464 u16 *vram_ptr = psx_gpu->vram_ptr;
467 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
469 for(i = 0; i < psx_gpu->num_blocks; i++)
471 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
472 if(fb_offset & (1 << 11))
474 *block_dest = *block_src;
483 for(i = 0; i < psx_gpu->num_blocks; i++)
485 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
486 if((fb_offset & (1 << 11)) == 0)
488 *block_dest = *block_src;
496 psx_gpu->num_blocks = num_blocks_dest;
499 if(psx_gpu->num_blocks)
501 render_block_handler_struct *render_block_handler =
502 psx_gpu->render_block_handler;
504 render_block_handler->texture_blocks(psx_gpu);
505 render_block_handler->shade_blocks(psx_gpu);
506 render_block_handler->blend_blocks(psx_gpu);
509 span_pixel_blocks += psx_gpu->num_blocks;
510 render_buffer_flushes++;
513 psx_gpu->num_blocks = 0;
520 #define setup_gradient_calculation_input(set, vertex) \
521 /* First type is: uvrg bxxx xxxx */\
522 /* Second type is: yyyy ybyy uvrg */\
523 /* Since x_a and y_c are the same the same variable is used for both. */\
524 x##set##_a_y##set##_c.e[0] = vertex->u; \
525 x##set##_a_y##set##_c.e[1] = vertex->v; \
526 x##set##_a_y##set##_c.e[2] = vertex->r; \
527 x##set##_a_y##set##_c.e[3] = vertex->g; \
528 dup_4x16b(x##set##_b, vertex->x); \
529 dup_4x16b(x##set##_c, vertex->x); \
530 dup_4x16b(y##set##_a, vertex->y); \
531 dup_4x16b(y##set##_b, vertex->y); \
532 x##set##_b.e[0] = vertex->b; \
533 y##set##_b.e[1] = vertex->b \
536 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
537 vertex_struct *b, vertex_struct *c)
539 u32 triangle_area = psx_gpu->triangle_area;
540 u32 winding_mask_scalar;
542 u32 triangle_area_shift;
543 u64 triangle_area_reciprocal =
544 fixed_reciprocal(triangle_area, &triangle_area_shift);
545 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
547 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
548 // ( d0 * d1 ) - ( d2 * d3 ) =
549 // ( m0 ) - ( m1 ) = gradient
551 // This is split to do 12 elements at a time over three sets: a, b, and c.
552 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
553 // two of the slots are unused.
555 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
558 vec_4x16s x0_a_y0_c, x0_b, x0_c;
559 vec_4x16s y0_a, y0_b;
560 vec_4x16s x1_a_y1_c, x1_b, x1_c;
561 vec_4x16s y1_a, y1_b;
562 vec_4x16s x2_a_y2_c, x2_b, x2_c;
563 vec_4x16s y2_a, y2_b;
567 vec_4x32u uvrgb_phase;
569 vec_4x16s d0_a_d3_c, d0_b, d0_c;
570 vec_4x16s d1_a, d1_b, d1_c_d2_a;
571 vec_4x16s d2_b, d2_c;
572 vec_4x16s d3_a, d3_b;
574 vec_4x32s m0_a, m0_b, m0_c;
575 vec_4x32s m1_a, m1_b, m1_c;
577 vec_4x32u gradient_area_a, gradient_area_c;
578 vec_2x32u gradient_area_b;
580 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
581 vec_2x32u gradient_area_sign_b;
582 vec_4x32u winding_mask;
584 vec_2x64u gradient_wide_a0, gradient_wide_a1;
585 vec_2x64u gradient_wide_c0, gradient_wide_c1;
586 vec_2x64u gradient_wide_b;
588 vec_4x32u gradient_a, gradient_c;
589 vec_2x32u gradient_b;
590 vec_16x8s gradient_shift;
592 setup_gradient_calculation_input(0, a);
593 setup_gradient_calculation_input(1, b);
594 setup_gradient_calculation_input(2, c);
596 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
597 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
598 shl_long_4x16b(b_base, x0_b, 16);
600 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
601 add_4x32b(b_base, b_base, uvrgb_phase);
603 // Can probably pair these, but it'll require careful register allocation
604 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
605 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
607 sub_4x16b(d0_b, x1_b, x0_b);
608 sub_4x16b(d0_c, x1_c, x0_c);
610 sub_4x16b(d1_a, y2_a, y1_a);
611 sub_4x16b(d1_b, y2_b, y1_b);
613 sub_4x16b(d2_b, x2_b, x1_b);
614 sub_4x16b(d2_c, x2_c, x1_c);
616 sub_4x16b(d3_a, y1_a, y0_a);
617 sub_4x16b(d3_b, y1_b, y0_b);
619 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
620 mul_long_4x16b(m0_b, d0_b, d1_b);
621 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
623 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
624 mul_long_4x16b(m1_b, d2_b, d3_b);
625 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
627 sub_4x32b(gradient_area_a, m0_a, m1_a);
628 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
629 sub_4x32b(gradient_area_c, m0_c, m1_c);
631 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
632 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
633 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
635 abs_4x32b(gradient_area_a, gradient_area_a);
636 abs_2x32b(gradient_area_b, gradient_area_b);
637 abs_4x32b(gradient_area_c, gradient_area_c);
639 winding_mask_scalar = -psx_gpu->triangle_winding;
641 dup_4x32b(winding_mask, winding_mask_scalar);
642 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
643 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
644 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
646 mul_scalar_long_2x32b(gradient_wide_a0,
647 vector_cast(vec_2x32s, gradient_area_a.low),
648 (s64)triangle_area_reciprocal);
649 mul_scalar_long_2x32b(gradient_wide_a1,
650 vector_cast(vec_2x32s, gradient_area_a.high),
651 (s64)triangle_area_reciprocal);
652 mul_scalar_long_2x32b(gradient_wide_b,
653 vector_cast(vec_2x32s, gradient_area_b),
654 (s64)triangle_area_reciprocal);
655 mul_scalar_long_2x32b(gradient_wide_c0,
656 vector_cast(vec_2x32s, gradient_area_c.low),
657 (s64)triangle_area_reciprocal);
658 mul_scalar_long_2x32b(gradient_wide_c1,
659 vector_cast(vec_2x32s, gradient_area_c.high),
660 (s64)triangle_area_reciprocal);
662 dup_16x8b(gradient_shift, triangle_area_shift);
663 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
664 vector_cast(vec_2x64u, gradient_shift));
665 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
666 vector_cast(vec_2x64u, gradient_shift));
667 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
668 vector_cast(vec_2x64u, gradient_shift));
669 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
670 vector_cast(vec_2x64u, gradient_shift));
671 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
672 vector_cast(vec_2x64u, gradient_shift));
674 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
675 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
676 mov_narrow_2x64b(gradient_b, gradient_wide_b);
677 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
678 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
680 shl_4x32b(gradient_a, gradient_a, 4);
681 shl_2x32b(gradient_b, gradient_b, 4);
682 shl_4x32b(gradient_c, gradient_c, 4);
684 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
685 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
686 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
688 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
689 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
690 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
692 u32 left_adjust = a->x;
693 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
694 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
704 eor_4x32b(zero, zero, zero);
705 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
706 add_2x32b(b_dx2, gradient_b, gradient_b);
707 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
708 add_2x32b(b_dx3, gradient_b, b_dx2);
710 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
712 psx_gpu->u_block_span.e[0] = zero.e[0];
713 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
714 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
715 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
717 psx_gpu->v_block_span.e[0] = zero.e[1];
718 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
719 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
720 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
722 psx_gpu->r_block_span.e[0] = zero.e[2];
723 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
724 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
725 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
727 psx_gpu->g_block_span.e[0] = zero.e[3];
728 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
729 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
730 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
732 psx_gpu->b_block_span.e[0] = zero.e[0];
733 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
734 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
735 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
737 psx_gpu->uvrg = uvrg_base;
738 psx_gpu->b = b_base.e[0];
740 psx_gpu->uvrg_dx = gradient_a;
741 psx_gpu->uvrg_dy = gradient_c;
742 psx_gpu->b_dy = gradient_b.e[1];
746 #define vector_check(_a, _b) \
747 if(memcmp(&_a, &_b, sizeof(_b))) \
749 if(sizeof(_b) == 8) \
751 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
752 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
756 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
757 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
762 #define scalar_check(_a, _b) \
764 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
768 #define setup_spans_debug_check(span_edge_data_element) \
770 u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data; \
771 if (_num_spans > MAX_SPANS) \
773 if (_num_spans < psx_gpu->num_spans) \
775 if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW) \
777 if(span_edge_data_element.y > 2048) \
783 #define setup_spans_debug_check(span_edge_data_element) \
787 #define setup_spans_prologue_alternate_yes() \
788 vec_2x64s alternate_x; \
789 vec_2x64s alternate_dx_dy; \
790 vec_4x32s alternate_x_32; \
791 vec_2x32s alternate_x_16; \
793 vec_4x16u alternate_select; \
794 vec_4x16s y_mid_point; \
798 s32 edge_dx_dy_alt; \
801 #define setup_spans_prologue_alternate_no() \
803 #define setup_spans_prologue(alternate_active) \
804 edge_data_struct *span_edge_data; \
805 vec_4x32u *span_uvrg_offset; \
806 u32 *span_b_offset; \
810 vec_2x64s edges_xy; \
811 vec_2x32s edges_dx_dy; \
812 vec_2x32u edge_shifts; \
814 vec_2x64s left_x, right_x; \
815 vec_2x64s left_dx_dy, right_dx_dy; \
816 vec_4x32s left_x_32, right_x_32; \
817 vec_8x16s left_right_x_16; \
819 vec_8x16s left_edge; \
820 vec_8x16s right_edge; \
821 vec_4x16u span_shift; \
825 vec_4x16u c_0xFFFE; \
828 vec_2x32s x_starts; \
837 vec_4x32u uvrg = psx_gpu->uvrg; \
838 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
839 u32 b = psx_gpu->b; \
840 u32 b_dy = psx_gpu->b_dy; \
842 dup_2x32b(c_0x01, 0x01); \
843 setup_spans_prologue_alternate_##alternate_active() \
845 #define setup_spans_prologue_b() \
846 span_edge_data = psx_gpu->span_edge_data; \
847 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
848 span_b_offset = psx_gpu->span_b_offset; \
850 vec_8x16u c_0x0001; \
852 dup_8x16b(c_0x0001, 0x0001); \
853 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
854 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
855 add_8x16b(right_edge, right_edge, c_0x0001); \
856 dup_4x16b(c_0x04, 0x04); \
857 dup_4x16b(c_0x07, 0x07); \
858 dup_4x16b(c_0xFFFE, 0xFFFE); \
861 #define compute_edge_delta_x2() \
864 vec_2x32s height_reciprocals; \
865 vec_2x32s heights_b; \
868 u32 edge_shift = reciprocal_table[height]; \
870 dup_2x32b(heights, height); \
871 sub_2x32b(widths, x_ends, x_starts); \
873 dup_2x32b(edge_shifts, edge_shift); \
874 sub_2x32b(heights_b, heights, c_0x01); \
875 shr_2x32b(height_reciprocals, edge_shifts, 10); \
877 mla_2x32b(heights_b, x_starts, heights); \
878 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
879 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
880 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
883 #define compute_edge_delta_x3(start_c, height_a, height_b) \
886 vec_2x32s height_reciprocals; \
887 vec_2x32s heights_b; \
892 u32 height_reciprocal_alt; \
894 heights.e[0] = height_a; \
895 heights.e[1] = height_b; \
897 edge_shifts.e[0] = reciprocal_table[height_a]; \
898 edge_shifts.e[1] = reciprocal_table[height_b]; \
899 edge_shift_alt = reciprocal_table[height_minor_b]; \
901 sub_2x32b(widths, x_ends, x_starts); \
902 width_alt = x_c - start_c; \
904 shr_2x32b(height_reciprocals, edge_shifts, 10); \
905 height_reciprocal_alt = edge_shift_alt >> 10; \
907 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
908 edge_shift_alt &= 0x1F; \
910 sub_2x32b(heights_b, heights, c_0x01); \
911 height_b_alt = height_minor_b - 1; \
913 mla_2x32b(heights_b, x_starts, heights); \
914 height_b_alt += height_minor_b * start_c; \
916 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
917 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
919 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
920 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
924 #define setup_spans_adjust_y_up() \
925 sub_4x32b(y_x4, y_x4, c_0x04) \
927 #define setup_spans_adjust_y_down() \
928 add_4x32b(y_x4, y_x4, c_0x04) \
930 #define setup_spans_adjust_interpolants_up() \
931 sub_4x32b(uvrg, uvrg, uvrg_dy); \
934 #define setup_spans_adjust_interpolants_down() \
935 add_4x32b(uvrg, uvrg, uvrg_dy); \
939 #define setup_spans_clip_interpolants_increment() \
940 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
943 #define setup_spans_clip_interpolants_decrement() \
944 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
947 #define setup_spans_clip_alternate_yes() \
948 edge_alt += edge_dx_dy_alt * (s64)(clip) \
950 #define setup_spans_clip_alternate_no() \
952 #define setup_spans_clip(direction, alternate_active) \
954 clipped_triangles++; \
955 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
956 setup_spans_clip_alternate_##alternate_active(); \
957 setup_spans_clip_interpolants_##direction(); \
961 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
963 vec_2x64u edge_shifts_64; \
964 vec_2x64s edges_dx_dy_64; \
966 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
967 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
969 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
970 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
972 left_x.e[0] = edges_xy.e[left_index]; \
973 right_x.e[0] = edges_xy.e[right_index]; \
975 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
976 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
977 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
978 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
980 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
981 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
983 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
984 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
987 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
989 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
990 s64 edge_dx_dy_alt_64; \
992 dup_4x16b(y_mid_point, y_b); \
994 edge_alt <<= edge_shift_alt; \
995 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
997 alternate_x.e[0] = edge_alt; \
998 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
999 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1001 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1002 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1006 #define setup_spans_y_select_up() \
1007 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1009 #define setup_spans_y_select_down() \
1010 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1012 #define setup_spans_y_select_alternate_yes(direction) \
1013 setup_spans_y_select_##direction() \
1015 #define setup_spans_y_select_alternate_no(direction) \
1017 #define setup_spans_alternate_select_left() \
1018 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1020 #define setup_spans_alternate_select_right() \
1021 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1023 #define setup_spans_alternate_select_none() \
1025 #define setup_spans_increment_alternate_yes() \
1026 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1027 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1028 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1029 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1030 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1032 #define setup_spans_increment_alternate_no() \
1034 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1036 span_uvrg_offset[0] = uvrg; \
1037 span_b_offset[0] = b; \
1038 setup_spans_adjust_interpolants_##direction(); \
1040 span_uvrg_offset[1] = uvrg; \
1041 span_b_offset[1] = b; \
1042 setup_spans_adjust_interpolants_##direction(); \
1044 span_uvrg_offset[2] = uvrg; \
1045 span_b_offset[2] = b; \
1046 setup_spans_adjust_interpolants_##direction(); \
1048 span_uvrg_offset[3] = uvrg; \
1049 span_b_offset[3] = b; \
1050 setup_spans_adjust_interpolants_##direction(); \
1052 span_uvrg_offset += 4; \
1053 span_b_offset += 4; \
1055 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1056 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1058 add_2x64b(left_x, left_x, left_dx_dy); \
1059 add_2x64b(right_x, right_x, right_dx_dy); \
1061 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1062 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1064 add_2x64b(left_x, left_x, left_dx_dy); \
1065 add_2x64b(right_x, right_x, right_dx_dy); \
1067 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1068 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1070 setup_spans_increment_alternate_##alternate_active(); \
1071 setup_spans_y_select_alternate_##alternate_active(direction); \
1072 setup_spans_alternate_select_##alternate(); \
1074 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1075 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1077 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1078 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1079 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1080 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1081 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1084 for(i = 0; i < 4; i++) \
1086 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1087 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1088 span_edge_data[i].right_mask = span_shift.e[i]; \
1089 span_edge_data[i].y = y_x4.e[i]; \
1090 setup_spans_debug_check(span_edge_data[i]); \
1093 span_edge_data += 4; \
1095 setup_spans_adjust_y_##direction(); \
1099 #define setup_spans_alternate_adjust_yes() \
1100 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1102 #define setup_spans_alternate_adjust_no() \
1105 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1106 setup_spans_alternate_adjust_##alternate_active(); \
1107 if(y_c > psx_gpu->viewport_end_y) \
1108 height -= y_c - psx_gpu->viewport_end_y - 1; \
1110 clip = psx_gpu->viewport_start_y - y_a; \
1115 setup_spans_clip(increment, alternate_active); \
1118 setup_spans_prologue_b(); \
1123 y_x4.e[1] = y_a + 1; \
1124 y_x4.e[2] = y_a + 2; \
1125 y_x4.e[3] = y_a + 3; \
1126 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1129 psx_gpu->num_spans = height; \
1132 setup_spans_set_x4(alternate, down, alternate_active); \
1134 } while(height > 0); \
1138 #define setup_spans_alternate_pre_increment_yes() \
1139 edge_alt += edge_dx_dy_alt \
1141 #define setup_spans_alternate_pre_increment_no() \
1143 #define setup_spans_up_decrement_height_yes() \
1146 #define setup_spans_up_decrement_height_no() \
1149 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1150 setup_spans_alternate_adjust_##alternate_active(); \
1153 if(y_c < psx_gpu->viewport_start_y) \
1154 height -= psx_gpu->viewport_start_y - y_c; \
1156 setup_spans_up_decrement_height_##alternate_active(); \
1158 clip = y_a - psx_gpu->viewport_end_y; \
1163 setup_spans_clip(decrement, alternate_active); \
1166 setup_spans_prologue_b(); \
1171 y_x4.e[1] = y_a - 1; \
1172 y_x4.e[2] = y_a - 2; \
1173 y_x4.e[3] = y_a - 3; \
1174 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1175 setup_spans_alternate_pre_increment_##alternate_active(); \
1176 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1178 setup_spans_adjust_interpolants_up(); \
1180 psx_gpu->num_spans = height; \
1183 setup_spans_set_x4(alternate, up, alternate_active); \
1188 #define index_left 0
1189 #define index_right 1
1191 #define setup_spans_up_up(minor, major) \
1192 setup_spans_prologue(yes); \
1193 s32 height_minor_a = y_a - y_b; \
1194 s32 height_minor_b = y_b - y_c; \
1195 s32 height = y_a - y_c; \
1197 dup_2x32b(x_starts, x_a); \
1198 x_ends.e[0] = x_c; \
1199 x_ends.e[1] = x_b; \
1201 compute_edge_delta_x3(x_b, height, height_minor_a); \
1202 setup_spans_up(index_##major, index_##minor, minor, yes) \
1207 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1208 vertex_struct *v_b, vertex_struct *v_c)
1210 setup_spans_up_up(left, right);
1213 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1214 vertex_struct *v_b, vertex_struct *v_c)
1216 setup_spans_up_up(right, left);
1219 #define setup_spans_down_down(minor, major) \
1220 setup_spans_prologue(yes); \
1221 s32 height_minor_a = y_b - y_a; \
1222 s32 height_minor_b = y_c - y_b; \
1223 s32 height = y_c - y_a; \
1225 dup_2x32b(x_starts, x_a); \
1226 x_ends.e[0] = x_c; \
1227 x_ends.e[1] = x_b; \
1229 compute_edge_delta_x3(x_b, height, height_minor_a); \
1230 setup_spans_down(index_##major, index_##minor, minor, yes) \
1232 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1233 vertex_struct *v_b, vertex_struct *v_c)
1235 setup_spans_down_down(left, right);
1238 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1239 vertex_struct *v_b, vertex_struct *v_c)
1241 setup_spans_down_down(right, left);
1244 #define setup_spans_up_flat() \
1245 s32 height = y_a - y_c; \
1248 compute_edge_delta_x2(); \
1249 setup_spans_up(index_left, index_right, none, no) \
1251 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1252 vertex_struct *v_b, vertex_struct *v_c)
1254 setup_spans_prologue(no);
1255 x_starts.e[0] = x_a;
1256 x_starts.e[1] = x_b;
1257 dup_2x32b(x_ends, x_c);
1259 setup_spans_up_flat();
1262 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1263 vertex_struct *v_b, vertex_struct *v_c)
1265 setup_spans_prologue(no);
1266 dup_2x32b(x_starts, x_a);
1270 setup_spans_up_flat();
1273 #define setup_spans_down_flat() \
1274 s32 height = y_c - y_a; \
1277 compute_edge_delta_x2(); \
1278 setup_spans_down(index_left, index_right, none, no) \
1280 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1281 vertex_struct *v_b, vertex_struct *v_c)
1283 setup_spans_prologue(no);
1284 x_starts.e[0] = x_a;
1285 x_starts.e[1] = x_b;
1286 dup_2x32b(x_ends, x_c);
1288 setup_spans_down_flat();
1291 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1292 vertex_struct *v_b, vertex_struct *v_c)
1294 setup_spans_prologue(no);
1295 dup_2x32b(x_starts, x_a);
1299 setup_spans_down_flat();
1302 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1303 vertex_struct *v_b, vertex_struct *v_c)
1305 setup_spans_prologue(no);
1313 s32 height_minor_a = y_a - y_b;
1314 s32 height_minor_b = y_c - y_a;
1315 s32 height_major = y_c - y_b;
1317 vec_2x64s edges_xy_b;
1318 vec_2x32s edges_dx_dy_b;
1319 vec_2x32u edge_shifts_b;
1321 vec_2x32s height_increment;
1323 x_starts.e[0] = x_a;
1324 x_starts.e[1] = x_c;
1325 dup_2x32b(x_ends, x_b);
1327 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1329 height_increment.e[0] = 0;
1330 height_increment.e[1] = height_minor_b;
1332 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1334 edges_xy_b.e[0] = edge_alt;
1335 edges_xy_b.e[1] = edges_xy.e[1];
1337 edge_shifts_b = edge_shifts;
1338 edge_shifts_b.e[0] = edge_shift_alt;
1340 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1341 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1345 if(y_b < psx_gpu->viewport_start_y)
1346 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1348 clip = y_a - psx_gpu->viewport_end_y;
1351 height_minor_a -= clip;
1353 setup_spans_clip(decrement, no);
1356 setup_spans_prologue_b();
1358 if(height_minor_a > 0)
1361 y_x4.e[1] = y_a - 1;
1362 y_x4.e[2] = y_a - 2;
1363 y_x4.e[3] = y_a - 3;
1364 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1365 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1366 setup_spans_adjust_interpolants_up();
1368 psx_gpu->num_spans = height_minor_a;
1369 while(height_minor_a > 0)
1371 setup_spans_set_x4(none, up, no);
1372 height_minor_a -= 4;
1375 span_edge_data += height_minor_a;
1376 span_uvrg_offset += height_minor_a;
1377 span_b_offset += height_minor_a;
1380 edges_xy = edges_xy_b;
1381 edges_dx_dy = edges_dx_dy_b;
1382 edge_shifts = edge_shifts_b;
1384 uvrg = psx_gpu->uvrg;
1389 if(y_c > psx_gpu->viewport_end_y)
1390 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1392 clip = psx_gpu->viewport_start_y - y_a;
1395 height_minor_b -= clip;
1397 setup_spans_clip(increment, no);
1400 if(height_minor_b > 0)
1403 y_x4.e[1] = y_a + 1;
1404 y_x4.e[2] = y_a + 2;
1405 y_x4.e[3] = y_a + 3;
1406 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1408 // FIXME: overflow corner case
1409 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1410 height_minor_b &= ~3;
1412 psx_gpu->num_spans += height_minor_b;
1413 while(height_minor_b > 0)
1415 setup_spans_set_x4(none, down, no);
1416 height_minor_b -= 4;
1420 left_split_triangles++;
1426 #define dither_table_entry_normal(value) \
1430 #define setup_blocks_load_msb_mask_indirect() \
1432 #define setup_blocks_load_msb_mask_direct() \
1433 vec_8x16u msb_mask; \
1434 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1437 #define setup_blocks_variables_shaded_textured(target) \
1438 vec_4x32u u_block; \
1439 vec_4x32u v_block; \
1440 vec_4x32u r_block; \
1441 vec_4x32u g_block; \
1442 vec_4x32u b_block; \
1443 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1444 vec_4x32u uvrg_dx4; \
1445 vec_4x32u uvrg_dx8; \
1447 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1448 u32 b_dx4 = b_dx << 2; \
1449 u32 b_dx8 = b_dx << 3; \
1452 vec_16x8u texture_mask; \
1453 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1454 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1455 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1456 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1458 #define setup_blocks_variables_shaded_untextured(target) \
1459 vec_4x32u r_block; \
1460 vec_4x32u g_block; \
1461 vec_4x32u b_block; \
1463 vec_4x32u rgb_dx4; \
1464 vec_4x32u rgb_dx8; \
1467 vec_8x8u d64_0x07; \
1472 dup_8x8b(d64_0x07, 0x07); \
1473 dup_8x8b(d64_1, 1); \
1474 dup_8x8b(d64_4, 4); \
1475 dup_8x8b(d64_128, 128); \
1477 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1478 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1479 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1480 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1482 #define setup_blocks_variables_unshaded_textured(target) \
1483 vec_4x32u u_block; \
1484 vec_4x32u v_block; \
1485 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1488 vec_2x32u uv = psx_gpu->uvrg.low; \
1490 vec_16x8u texture_mask; \
1491 shl_2x32b(uv_dx4, uv_dx, 2); \
1492 shl_2x32b(uv_dx8, uv_dx, 3); \
1493 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1494 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1497 #define setup_blocks_variables_unshaded_untextured_direct() \
1498 or_8x16b(colors, colors, msb_mask) \
1500 #define setup_blocks_variables_unshaded_untextured_indirect() \
1502 #define setup_blocks_variables_unshaded_untextured(target) \
1503 u32 color = psx_gpu->triangle_color; \
1506 u32 color_r = color & 0xFF; \
1507 u32 color_g = (color >> 8) & 0xFF; \
1508 u32 color_b = (color >> 16) & 0xFF; \
1510 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1511 ((color_b >> 3) << 10); \
1512 dup_8x16b(colors, color); \
1513 setup_blocks_variables_unshaded_untextured_##target() \
1515 #define setup_blocks_span_initialize_dithered_textured() \
1516 vec_8x16u dither_offsets; \
1517 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1519 #define setup_blocks_span_initialize_dithered_untextured() \
1520 vec_8x8u dither_offsets; \
1521 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1523 #define setup_blocks_span_initialize_dithered(texturing) \
1524 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1525 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1526 vec_8x8s dither_offsets_short; \
1529 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1530 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1531 setup_blocks_span_initialize_dithered_##texturing() \
1533 #define setup_blocks_span_initialize_undithered(texturing) \
1536 #define setup_blocks_span_initialize_shaded_textured() \
1538 vec_4x32u block_span; \
1539 u32 offset = span_edge_data->left_x; \
1541 uvrg = *span_uvrg_offset; \
1542 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1543 b = *span_b_offset; \
1544 b += b_dx * offset; \
1546 dup_4x32b(u_block, uvrg.e[0]); \
1547 dup_4x32b(v_block, uvrg.e[1]); \
1548 dup_4x32b(r_block, uvrg.e[2]); \
1549 dup_4x32b(g_block, uvrg.e[3]); \
1550 dup_4x32b(b_block, b); \
1552 block_span = psx_gpu->u_block_span; \
1553 add_4x32b(u_block, u_block, block_span); \
1554 block_span = psx_gpu->v_block_span; \
1555 add_4x32b(v_block, v_block, block_span); \
1556 block_span = psx_gpu->r_block_span; \
1557 add_4x32b(r_block, r_block, block_span); \
1558 block_span = psx_gpu->g_block_span; \
1559 add_4x32b(g_block, g_block, block_span); \
1560 block_span = psx_gpu->b_block_span; \
1561 add_4x32b(b_block, b_block, block_span); \
1564 #define setup_blocks_span_initialize_shaded_untextured() \
1566 vec_4x32u block_span; \
1567 u32 offset = span_edge_data->left_x; \
1569 rgb.low = span_uvrg_offset->high; \
1570 rgb.high.e[0] = *span_b_offset; \
1571 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1573 dup_4x32b(r_block, rgb.e[0]); \
1574 dup_4x32b(g_block, rgb.e[1]); \
1575 dup_4x32b(b_block, rgb.e[2]); \
1577 block_span = psx_gpu->r_block_span; \
1578 add_4x32b(r_block, r_block, block_span); \
1579 block_span = psx_gpu->g_block_span; \
1580 add_4x32b(g_block, g_block, block_span); \
1581 block_span = psx_gpu->b_block_span; \
1582 add_4x32b(b_block, b_block, block_span); \
1585 #define setup_blocks_span_initialize_unshaded_textured() \
1587 vec_4x32u block_span; \
1588 u32 offset = span_edge_data->left_x; \
1590 uv = span_uvrg_offset->low; \
1591 mla_scalar_2x32b(uv, uv_dx, offset); \
1593 dup_4x32b(u_block, uv.e[0]); \
1594 dup_4x32b(v_block, uv.e[1]); \
1596 block_span = psx_gpu->u_block_span; \
1597 add_4x32b(u_block, u_block, block_span); \
1598 block_span = psx_gpu->v_block_span; \
1599 add_4x32b(v_block, v_block, block_span); \
1602 #define setup_blocks_span_initialize_unshaded_untextured() \
1605 #define setup_blocks_texture_swizzled() \
1607 vec_8x8u u_saved = u; \
1608 sli_8x8b(u, v, 4); \
1609 sri_8x8b(v, u_saved, 4); \
1612 #define setup_blocks_texture_unswizzled() \
1614 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1617 vec_8x16u u_whole; \
1618 vec_8x16u v_whole; \
1619 vec_8x16u r_whole; \
1620 vec_8x16u g_whole; \
1621 vec_8x16u b_whole; \
1633 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1634 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1635 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1636 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1637 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1639 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1640 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1641 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1642 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1643 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1644 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1645 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1646 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1647 dup_4x32b(dx4, b_dx4); \
1648 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1650 mov_narrow_8x16b(u, u_whole); \
1651 mov_narrow_8x16b(v, v_whole); \
1652 mov_narrow_8x16b(r, r_whole); \
1653 mov_narrow_8x16b(g, g_whole); \
1654 mov_narrow_8x16b(b, b_whole); \
1656 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1657 add_4x32b(u_block, u_block, dx8); \
1658 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1659 add_4x32b(v_block, v_block, dx8); \
1660 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1661 add_4x32b(r_block, r_block, dx8); \
1662 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1663 add_4x32b(g_block, g_block, dx8); \
1664 dup_4x32b(dx8, b_dx8); \
1665 add_4x32b(b_block, b_block, dx8); \
1667 and_8x8b(u, u, texture_mask.low); \
1668 and_8x8b(v, v, texture_mask.high); \
1669 setup_blocks_texture_##swizzling(); \
1671 zip_8x16b(uv, u, v); \
1676 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1677 block->fb_ptr = fb_ptr; \
1680 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1683 vec_8x16u u_whole; \
1684 vec_8x16u v_whole; \
1693 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1694 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1696 dup_4x32b(dx4, uv_dx4.e[0]); \
1697 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1698 dup_4x32b(dx4, uv_dx4.e[1]); \
1699 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1701 mov_narrow_8x16b(u, u_whole); \
1702 mov_narrow_8x16b(v, v_whole); \
1704 dup_4x32b(dx8, uv_dx8.e[0]); \
1705 add_4x32b(u_block, u_block, dx8); \
1706 dup_4x32b(dx8, uv_dx8.e[1]); \
1707 add_4x32b(v_block, v_block, dx8); \
1709 and_8x8b(u, u, texture_mask.low); \
1710 and_8x8b(v, v, texture_mask.high); \
1711 setup_blocks_texture_##swizzling(); \
1713 zip_8x16b(uv, u, v); \
1715 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1716 block->fb_ptr = fb_ptr; \
1719 #define setup_blocks_store_shaded_untextured_dithered() \
1720 addq_8x8b(r, r, dither_offsets); \
1721 addq_8x8b(g, g, dither_offsets); \
1722 addq_8x8b(b, b, dither_offsets); \
1724 subq_8x8b(r, r, d64_4); \
1725 subq_8x8b(g, g, d64_4); \
1726 subq_8x8b(b, b, d64_4) \
1728 #define setup_blocks_store_shaded_untextured_undithered() \
1731 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1732 block->pixels = _pixels; \
1733 block->fb_ptr = fb_ptr \
1735 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1736 block->pixels = _pixels; \
1737 block->fb_ptr = fb_ptr \
1739 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1740 mul_long_8x8b(pixels, r, d64_1) \
1743 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1744 store_8x16b(_pixels, fb_ptr) \
1746 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1748 vec_8x16u fb_pixels; \
1749 vec_8x16u draw_mask; \
1750 vec_8x16u test_mask = psx_gpu->test_mask; \
1752 load_8x16b(fb_pixels, fb_ptr); \
1753 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1754 tst_8x16b(draw_mask, draw_mask, test_mask); \
1755 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1756 store_8x16b(fb_pixels, fb_ptr); \
1759 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1760 pixels = msb_mask; \
1761 mla_long_8x8b(pixels, r, d64_1) \
1764 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1767 vec_8x16u r_whole; \
1768 vec_8x16u g_whole; \
1769 vec_8x16u b_whole; \
1780 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1781 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1782 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1784 dup_4x32b(dx4, rgb_dx4.e[0]); \
1785 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1786 dup_4x32b(dx4, rgb_dx4.e[1]); \
1787 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1788 dup_4x32b(dx4, rgb_dx4.e[2]); \
1789 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1791 mov_narrow_8x16b(r, r_whole); \
1792 mov_narrow_8x16b(g, g_whole); \
1793 mov_narrow_8x16b(b, b_whole); \
1795 dup_4x32b(dx8, rgb_dx8.e[0]); \
1796 add_4x32b(r_block, r_block, dx8); \
1797 dup_4x32b(dx8, rgb_dx8.e[1]); \
1798 add_4x32b(g_block, g_block, dx8); \
1799 dup_4x32b(dx8, rgb_dx8.e[2]); \
1800 add_4x32b(b_block, b_block, dx8); \
1802 setup_blocks_store_shaded_untextured_##dithering(); \
1804 shr_8x8b(r, r, 3); \
1805 bic_8x8b(g, g, d64_0x07); \
1806 bic_8x8b(b, b, d64_0x07); \
1808 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1809 mla_long_8x8b(pixels, g, d64_4); \
1810 mla_long_8x8b(pixels, b, d64_128) \
1812 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1815 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1817 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1820 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1821 (_block)->draw_mask_bits = bits \
1823 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1825 vec_8x16u bits_mask; \
1826 vec_8x16u test_mask = psx_gpu->test_mask; \
1827 dup_8x16b(bits_mask, bits); \
1828 tst_8x16b(bits_mask, bits_mask, test_mask); \
1829 (_block)->draw_mask = bits_mask; \
1832 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1835 #define setup_blocks_add_blocks_indirect() \
1836 num_blocks += span_num_blocks; \
1838 if(num_blocks > MAX_BLOCKS) \
1840 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1841 flush_render_block_buffer(psx_gpu); \
1842 num_blocks = span_num_blocks; \
1843 block = psx_gpu->blocks; \
1846 #define setup_blocks_add_blocks_direct() \
1847 stats_add(texel_blocks_untextured, span_num_blocks); \
1848 span_pixel_blocks += span_num_blocks \
1851 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1852 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1853 psx_gpu_struct *psx_gpu) \
1855 setup_blocks_load_msb_mask_##target(); \
1856 setup_blocks_variables_##shading##_##texturing(target); \
1858 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1859 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1860 u32 *span_b_offset = psx_gpu->span_b_offset; \
1862 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1864 u32 num_spans = psx_gpu->num_spans; \
1869 u32 num_blocks = psx_gpu->num_blocks; \
1870 u32 span_num_blocks; \
1874 span_num_blocks = span_edge_data->num_blocks; \
1875 if(span_num_blocks) \
1877 y = span_edge_data->y; \
1878 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1880 setup_blocks_span_initialize_##shading##_##texturing(); \
1881 setup_blocks_span_initialize_##dithering(texturing); \
1883 setup_blocks_add_blocks_##target(); \
1885 s32 pixel_span = span_num_blocks * 8; \
1886 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1887 span_pixels += pixel_span; \
1889 span_num_blocks--; \
1890 while(span_num_blocks) \
1892 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1894 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1898 span_num_blocks--; \
1901 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1902 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1903 span_edge_data->right_mask); \
1909 zero_block_spans++; \
1914 span_uvrg_offset++; \
1918 psx_gpu->num_blocks = num_blocks; \
1922 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1926 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1927 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1929 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1930 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1932 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1933 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1934 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1935 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1937 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1938 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1940 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1942 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1943 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1946 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1948 block_struct *block = psx_gpu->blocks;
1949 u32 num_blocks = psx_gpu->num_blocks;
1950 stats_add(texel_blocks_4bpp, num_blocks);
1952 vec_8x8u texels_low;
1953 vec_8x8u texels_high;
1960 vec_16x8u clut_high;
1962 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1963 u16 *clut_ptr = psx_gpu->clut_ptr;
1965 // Can be done with one deinterleaving load on NEON
1966 load_8x16b(clut_a, clut_ptr);
1967 load_8x16b(clut_b, clut_ptr + 8);
1968 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
1970 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
1971 update_texture_4bpp_cache(psx_gpu);
1975 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
1976 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
1977 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
1978 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
1979 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
1980 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
1981 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
1982 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
1984 tbl_16(texels_low, texels, clut_low);
1985 tbl_16(texels_high, texels, clut_high);
1987 // Can be done with an interleaving store on NEON
1988 zip_8x16b(pixels, texels_low, texels_high);
1990 block->texels = pixels;
1997 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
1999 block_struct *block = psx_gpu->blocks;
2000 u32 num_blocks = psx_gpu->num_blocks;
2002 stats_add(texel_blocks_8bpp, num_blocks);
2004 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2005 update_texture_8bpp_cache(psx_gpu);
2008 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2016 for(i = 0; i < 8; i++)
2018 offset = block->uv.e[i];
2020 texel = texture_ptr_8bpp[offset];
2021 texels.e[i] = psx_gpu->clut_ptr[texel];
2024 block->texels = texels;
2031 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2033 block_struct *block = psx_gpu->blocks;
2034 u32 num_blocks = psx_gpu->num_blocks;
2036 stats_add(texel_blocks_16bpp, num_blocks);
2040 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2046 for(i = 0; i < 8; i++)
2048 offset = block->uv.e[i];
2049 offset += ((offset & 0xFF00) * 3);
2051 texels.e[i] = texture_ptr_16bpp[offset];
2054 block->texels = texels;
2064 #define shade_blocks_load_msb_mask_indirect() \
2066 #define shade_blocks_load_msb_mask_direct() \
2067 vec_8x16u msb_mask; \
2068 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2070 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2071 block->draw_mask = _draw_mask; \
2072 block->pixels = _pixels \
2074 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2076 vec_8x16u fb_pixels; \
2077 or_8x16b(_pixels, _pixels, msb_mask); \
2078 load_8x16b(fb_pixels, block->fb_ptr); \
2079 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2080 store_8x16b(fb_pixels, block->fb_ptr); \
2084 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2085 if(psx_gpu->triangle_color == 0x808080) \
2087 false_modulated_blocks += num_blocks; \
2090 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2091 if(psx_gpu->triangle_color == 0x808080) \
2094 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2095 false_modulated_blocks += num_blocks; \
2100 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2103 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2106 u32 color = psx_gpu->triangle_color; \
2107 dup_8x8b(colors_r, color); \
2108 dup_8x8b(colors_g, color >> 8); \
2109 dup_8x8b(colors_b, color >> 16); \
2110 shade_blocks_textured_false_modulated_check_##dithering(target); \
2113 #define shade_blocks_textured_modulated_shaded_block_load() \
2114 colors_r = block->r; \
2115 colors_g = block->g; \
2116 colors_b = block->b \
2118 #define shade_blocks_textured_modulated_unshaded_block_load() \
2120 #define shade_blocks_textured_modulate_dithered(component) \
2121 pixels_##component = block->dither_offsets; \
2122 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2124 #define shade_blocks_textured_modulate_undithered(component) \
2125 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2127 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2128 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2129 psx_gpu_struct *psx_gpu) \
2131 block_struct *block = psx_gpu->blocks; \
2132 u32 num_blocks = psx_gpu->num_blocks; \
2135 vec_8x8u texels_r; \
2136 vec_8x8u texels_g; \
2137 vec_8x8u texels_b; \
2139 vec_8x8u colors_r; \
2140 vec_8x8u colors_g; \
2141 vec_8x8u colors_b; \
2143 vec_8x8u pixels_r_low; \
2144 vec_8x8u pixels_g_low; \
2145 vec_8x8u pixels_b_low; \
2148 vec_8x16u pixels_r; \
2149 vec_8x16u pixels_g; \
2150 vec_8x16u pixels_b; \
2152 vec_8x16u draw_mask; \
2153 vec_8x16u zero_mask; \
2155 vec_8x8u d64_0x07; \
2156 vec_8x8u d64_0x1F; \
2161 vec_8x16u d128_0x8000; \
2163 vec_8x16u test_mask = psx_gpu->test_mask; \
2164 u32 draw_mask_bits; \
2165 shade_blocks_load_msb_mask_##target(); \
2167 dup_8x8b(d64_0x07, 0x07); \
2168 dup_8x8b(d64_0x1F, 0x1F); \
2169 dup_8x8b(d64_1, 1); \
2170 dup_8x8b(d64_4, 4); \
2171 dup_8x8b(d64_128, 128); \
2173 dup_8x16b(d128_0x8000, 0x8000); \
2175 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2180 draw_mask_bits = block->draw_mask_bits; \
2181 dup_8x16b(draw_mask, draw_mask_bits); \
2182 tst_8x16b(draw_mask, draw_mask, test_mask); \
2184 shade_blocks_textured_modulated_##shading##_block_load(); \
2186 texels = block->texels; \
2188 mov_narrow_8x16b(texels_r, texels); \
2189 shr_narrow_8x16b(texels_g, texels, 5); \
2190 shr_narrow_8x16b(texels_b, texels, 7); \
2192 and_8x8b(texels_r, texels_r, d64_0x1F); \
2193 and_8x8b(texels_g, texels_g, d64_0x1F); \
2194 shr_8x8b(texels_b, texels_b, 3); \
2196 shade_blocks_textured_modulate_##dithering(r); \
2197 shade_blocks_textured_modulate_##dithering(g); \
2198 shade_blocks_textured_modulate_##dithering(b); \
2200 cmpeqz_8x16b(zero_mask, texels); \
2201 and_8x16b(pixels, texels, d128_0x8000); \
2203 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2204 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2205 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2207 or_8x16b(zero_mask, draw_mask, zero_mask); \
2209 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2210 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2211 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2213 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2214 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2215 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2217 shade_blocks_store_##target(zero_mask, pixels); \
2226 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2227 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2228 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2229 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2231 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2232 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2233 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2234 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2239 #define shade_blocks_textured_unmodulated_builder(target) \
2240 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2242 block_struct *block = psx_gpu->blocks; \
2243 u32 num_blocks = psx_gpu->num_blocks; \
2244 vec_8x16u draw_mask; \
2245 vec_8x16u test_mask = psx_gpu->test_mask; \
2246 u32 draw_mask_bits; \
2249 shade_blocks_load_msb_mask_##target(); \
2253 vec_8x16u zero_mask; \
2255 draw_mask_bits = block->draw_mask_bits; \
2256 dup_8x16b(draw_mask, draw_mask_bits); \
2257 tst_8x16b(draw_mask, draw_mask, test_mask); \
2259 pixels = block->texels; \
2261 cmpeqz_8x16b(zero_mask, pixels); \
2262 or_8x16b(zero_mask, draw_mask, zero_mask); \
2264 shade_blocks_store_##target(zero_mask, pixels); \
2271 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2272 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2275 block_struct *block = psx_gpu->blocks; \
2276 u32 num_blocks = psx_gpu->num_blocks; \
2277 vec_8x16u draw_mask; \
2278 vec_8x16u test_mask = psx_gpu->test_mask; \
2279 u32 draw_mask_bits; \
2282 shade_blocks_load_msb_mask_##target(); \
2286 vec_8x16u zero_mask; \
2288 draw_mask_bits = block->draw_mask_bits; \
2289 dup_8x16b(draw_mask, draw_mask_bits); \
2290 tst_8x16b(draw_mask, draw_mask, test_mask); \
2292 pixels = block->texels; \
2294 cmpeqz_8x16b(zero_mask, pixels); \
2295 or_8x16b(zero_mask, draw_mask, zero_mask); \
2297 shade_blocks_store_##target(zero_mask, pixels); \
2306 shade_blocks_textured_unmodulated_builder(indirect)
2307 shade_blocks_textured_unmodulated_builder(direct)
2309 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2313 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2315 block_struct *block = psx_gpu->blocks;
2316 u32 num_blocks = psx_gpu->num_blocks;
2318 vec_8x16u pixels = block->pixels;
2319 shade_blocks_load_msb_mask_direct();
2323 shade_blocks_store_direct(block->draw_mask, pixels);
2332 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2337 #define blend_blocks_mask_evaluate_on() \
2338 vec_8x16u mask_pixels; \
2339 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2340 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2342 #define blend_blocks_mask_evaluate_off() \
2344 #define blend_blocks_average() \
2346 vec_8x16u pixels_no_msb; \
2347 vec_8x16u fb_pixels_no_msb; \
2349 vec_8x16u d128_0x0421; \
2350 vec_8x16u d128_0x8000; \
2352 dup_8x16b(d128_0x0421, 0x0421); \
2353 dup_8x16b(d128_0x8000, 0x8000); \
2355 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2356 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2357 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2358 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2359 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2360 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2363 #define blend_blocks_add() \
2365 vec_8x16u pixels_rb, pixels_g; \
2366 vec_8x16u fb_rb, fb_g; \
2368 vec_8x16u d128_0x7C1F; \
2369 vec_8x16u d128_0x03E0; \
2371 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2372 dup_8x16b(d128_0x03E0, 0x03E0); \
2374 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2375 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2377 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2378 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2380 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2381 add_8x16b(fb_g, fb_g, pixels_g); \
2383 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2384 vector_cast(vec_16x8u, d128_0x7C1F)); \
2385 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2387 or_8x16b(blend_pixels, fb_rb, fb_g); \
2390 #define blend_blocks_subtract() \
2392 vec_8x16u pixels_rb, pixels_g; \
2393 vec_8x16u fb_rb, fb_g; \
2395 vec_8x16u d128_0x7C1F; \
2396 vec_8x16u d128_0x03E0; \
2398 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2399 dup_8x16b(d128_0x03E0, 0x03E0); \
2401 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2402 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2404 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2405 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2407 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2408 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2409 subs_8x16b(fb_g, fb_g, pixels_g); \
2411 or_8x16b(blend_pixels, fb_rb, fb_g); \
2414 #define blend_blocks_add_fourth() \
2416 vec_8x16u pixels_rb, pixels_g; \
2417 vec_8x16u pixels_fourth; \
2418 vec_8x16u fb_rb, fb_g; \
2420 vec_8x16u d128_0x7C1F; \
2421 vec_8x16u d128_0x1C07; \
2422 vec_8x16u d128_0x03E0; \
2423 vec_8x16u d128_0x00E0; \
2425 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2426 dup_8x16b(d128_0x1C07, 0x1C07); \
2427 dup_8x16b(d128_0x03E0, 0x03E0); \
2428 dup_8x16b(d128_0x00E0, 0x00E0); \
2430 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2432 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2433 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2435 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2436 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2438 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2439 add_8x16b(fb_g, fb_g, pixels_g); \
2441 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2442 vector_cast(vec_16x8u, d128_0x7C1F)); \
2443 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2445 or_8x16b(blend_pixels, fb_rb, fb_g); \
2448 #define blend_blocks_blended_combine_textured() \
2450 vec_8x16u blend_mask; \
2451 cmpltz_8x16b(blend_mask, pixels); \
2453 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2454 bif_8x16b(blend_pixels, pixels, blend_mask); \
2457 #define blend_blocks_blended_combine_untextured() \
2460 #define blend_blocks_body_blend(blend_mode, texturing) \
2462 blend_blocks_##blend_mode(); \
2463 blend_blocks_blended_combine_##texturing(); \
2466 #define blend_blocks_body_average(texturing) \
2467 blend_blocks_body_blend(average, texturing) \
2469 #define blend_blocks_body_add(texturing) \
2470 blend_blocks_body_blend(add, texturing) \
2472 #define blend_blocks_body_subtract(texturing) \
2473 blend_blocks_body_blend(subtract, texturing) \
2475 #define blend_blocks_body_add_fourth(texturing) \
2476 blend_blocks_body_blend(add_fourth, texturing) \
2478 #define blend_blocks_body_unblended(texturing) \
2479 blend_pixels = pixels \
2482 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2484 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2487 block_struct *block = psx_gpu->blocks; \
2488 u32 num_blocks = psx_gpu->num_blocks; \
2489 vec_8x16u draw_mask; \
2491 vec_8x16u blend_pixels; \
2492 vec_8x16u framebuffer_pixels; \
2493 vec_8x16u msb_mask; \
2497 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2501 pixels = block->pixels; \
2502 draw_mask = block->draw_mask; \
2503 fb_ptr = block->fb_ptr; \
2505 load_8x16b(framebuffer_pixels, fb_ptr); \
2507 blend_blocks_mask_evaluate_##mask_evaluate(); \
2508 blend_blocks_body_##blend_mode(texturing); \
2510 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2511 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2512 store_8x16b(framebuffer_pixels, fb_ptr); \
2522 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2526 blend_blocks_builder(textured, average, off);
2527 blend_blocks_builder(textured, average, on);
2528 blend_blocks_builder(textured, add, off);
2529 blend_blocks_builder(textured, add, on);
2530 blend_blocks_builder(textured, subtract, off);
2531 blend_blocks_builder(textured, subtract, on);
2532 blend_blocks_builder(textured, add_fourth, off);
2533 blend_blocks_builder(textured, add_fourth, on);
2535 blend_blocks_builder(untextured, average, off);
2536 blend_blocks_builder(untextured, average, on);
2537 blend_blocks_builder(untextured, add, off);
2538 blend_blocks_builder(untextured, add, on);
2539 blend_blocks_builder(untextured, subtract, off);
2540 blend_blocks_builder(untextured, subtract, on);
2541 blend_blocks_builder(untextured, add_fourth, off);
2542 blend_blocks_builder(untextured, add_fourth, on);
2544 blend_blocks_builder(textured, unblended, on);
2549 #define vertex_swap(_a, _b) \
2551 vertex_struct *temp_vertex = _a; \
2554 triangle_winding ^= 1; \
2558 // Setup blocks parametric-variables:
2559 // SHADE TEXTURE_MAP SWIZZLING
2566 // 8 inputs, 6 combinations
2568 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2569 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2571 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2572 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2574 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2576 setup_blocks_switch_untextured_##shading(dithering, target) \
2578 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2579 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2581 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2582 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2584 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2585 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2587 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2588 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2590 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2591 dithering, mask_evaluate) \
2592 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2594 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2596 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2598 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2600 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2602 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2603 dithering, mask_evaluate) \
2604 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2605 texture_mode, dithering) \
2607 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2608 blending, mask_evaluate) \
2609 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2610 dithering, mask_evaluate) \
2615 #define texture_blocks_switch_untextured(texture_mode) \
2616 texture_blocks_untextured \
2618 #define texture_blocks_switch_textured(texture_mode) \
2619 texture_blocks_##texture_mode \
2621 #define texture_blocks_switch(texturing, texture_mode) \
2622 texture_blocks_switch_##texturing(texture_mode) \
2625 // Shade blocks parametric-variables:
2626 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2635 // 16 inputs, 8 combinations
2637 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2638 shade_blocks_unshaded_untextured_##target \
2640 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2641 shade_blocks_textured_unmodulated_##target \
2643 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2644 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2646 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2647 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2649 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2650 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2652 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2653 shade_blocks_shaded_untextured \
2655 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2656 shade_blocks_textured_unmodulated_##target \
2658 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2659 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2661 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2662 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2664 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2665 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2667 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2669 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2671 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2673 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2675 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2677 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2679 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2680 dithering, mask_evaluate) \
2681 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2684 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2685 blending, mask_evaluate) \
2686 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2690 // Blend blocks parametric-variables:
2691 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2710 // 32 inputs, 18 combinations
2712 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2713 blend_blocks_textured_unblended_##mask_evaluate \
2715 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2716 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2718 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2719 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2722 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2723 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2725 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2727 texture_blocks_switch(texturing, texture_mode), \
2728 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2730 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2733 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2734 mask_evaluate, shading, dithering, texturing, blending) \
2735 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2736 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2737 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2738 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2740 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2741 mask_evaluate, shading, dithering, texturing) \
2742 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2743 mask_evaluate, shading, dithering, texturing, unblended), \
2744 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2745 mask_evaluate, shading, dithering, texturing, blended) \
2747 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2748 mask_evaluate, shading, dithering) \
2749 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2750 mask_evaluate, shading, dithering, untextured), \
2751 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2752 mask_evaluate, shading, dithering, textured) \
2754 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2755 mask_evaluate, shading) \
2756 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2757 mask_evaluate, shading, undithered), \
2758 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2759 mask_evaluate, shading, dithered) \
2761 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2763 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2765 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2768 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2769 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2770 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2772 #define render_blocks_switch_block_texture_mode(texture_mode) \
2773 render_blocks_switch_block_blend_mode(texture_mode, average), \
2774 render_blocks_switch_block_blend_mode(texture_mode, add), \
2775 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2776 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2778 #define render_blocks_switch_block() \
2779 render_blocks_switch_block_texture_mode(4bpp), \
2780 render_blocks_switch_block_texture_mode(8bpp), \
2781 render_blocks_switch_block_texture_mode(16bpp), \
2782 render_blocks_switch_block_texture_mode(16bpp) \
2785 render_block_handler_struct render_triangle_block_handlers[] =
2787 render_blocks_switch_block()
2790 #undef render_blocks_switch_block_modulation
2792 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2793 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2795 "texture mode: " #texture_mode "\n" \
2796 "blend mode: " #blend_mode "\n" \
2797 "mask evaluation: " #mask_evaluate "\n" \
2804 char *render_block_flag_strings[] =
2806 render_blocks_switch_block()
2810 #define triangle_y_direction_up 1
2811 #define triangle_y_direction_flat 2
2812 #define triangle_y_direction_down 0
2814 #define triangle_winding_positive 0
2815 #define triangle_winding_negative 1
2817 #define triangle_set_direction(direction_variable, value) \
2818 u32 direction_variable = (u32)(value) >> 31; \
2820 direction_variable = 2 \
2822 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2823 case (triangle_y_direction_##direction_a | \
2824 (triangle_y_direction_##direction_b << 2) | \
2825 (triangle_y_direction_##direction_c << 4) | \
2826 (triangle_winding_##winding << 6)) \
2828 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2829 vertex_struct *vertexes_out[3])
2831 s32 y_top, y_bottom;
2833 u32 triangle_winding = 0;
2835 vertex_struct *a = &(vertexes[0]);
2836 vertex_struct *b = &(vertexes[1]);
2837 vertex_struct *c = &(vertexes[2]);
2839 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2845 if(triangle_area == 0)
2867 if((y_bottom - y_top) >= 512)
2875 if(triangle_area < 0)
2877 triangle_area = -triangle_area;
2878 triangle_winding ^= 1;
2893 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2901 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2910 psx_gpu->triangle_area = triangle_area;
2911 psx_gpu->triangle_winding = triangle_winding;
2913 vertexes_out[0] = a;
2914 vertexes_out[1] = b;
2915 vertexes_out[2] = c;
2920 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2921 vertex_struct *vertex_ptrs[3], u32 flags)
2923 psx_gpu->num_spans = 0;
2925 vertex_struct *a = vertex_ptrs[0];
2926 vertex_struct *b = vertex_ptrs[1];
2927 vertex_struct *c = vertex_ptrs[2];
2929 s32 y_delta_a = b->y - a->y;
2930 s32 y_delta_b = c->y - b->y;
2931 s32 y_delta_c = c->y - a->y;
2933 triangle_set_direction(y_direction_a, y_delta_a);
2934 triangle_set_direction(y_direction_b, y_delta_b);
2935 triangle_set_direction(y_direction_c, y_delta_c);
2937 compute_all_gradients(psx_gpu, a, b, c);
2939 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2940 (psx_gpu->triangle_winding << 6))
2942 triangle_case(up, up, up, negative):
2943 triangle_case(up, up, flat, negative):
2944 triangle_case(up, up, down, negative):
2945 setup_spans_up_right(psx_gpu, a, b, c);
2948 triangle_case(flat, up, up, negative):
2949 triangle_case(flat, up, flat, negative):
2950 triangle_case(flat, up, down, negative):
2951 setup_spans_up_a(psx_gpu, a, b, c);
2954 triangle_case(down, up, up, negative):
2955 setup_spans_up_down(psx_gpu, a, c, b);
2958 triangle_case(down, up, flat, negative):
2959 setup_spans_down_a(psx_gpu, a, c, b);
2962 triangle_case(down, up, down, negative):
2963 setup_spans_down_right(psx_gpu, a, c, b);
2966 triangle_case(down, flat, up, negative):
2967 triangle_case(down, flat, flat, negative):
2968 triangle_case(down, flat, down, negative):
2969 setup_spans_down_b(psx_gpu, a, b, c);
2972 triangle_case(down, down, up, negative):
2973 triangle_case(down, down, flat, negative):
2974 triangle_case(down, down, down, negative):
2975 setup_spans_down_left(psx_gpu, a, b, c);
2978 triangle_case(up, up, up, positive):
2979 triangle_case(up, up, flat, positive):
2980 triangle_case(up, up, down, positive):
2981 setup_spans_up_left(psx_gpu, a, b, c);
2984 triangle_case(up, flat, up, positive):
2985 triangle_case(up, flat, flat, positive):
2986 triangle_case(up, flat, down, positive):
2987 setup_spans_up_b(psx_gpu, a, b, c);
2990 triangle_case(up, down, up, positive):
2991 setup_spans_up_right(psx_gpu, a, c, b);
2994 triangle_case(up, down, flat, positive):
2995 setup_spans_up_a(psx_gpu, a, c, b);
2998 triangle_case(up, down, down, positive):
2999 setup_spans_up_down(psx_gpu, a, b, c);
3002 triangle_case(flat, down, up, positive):
3003 triangle_case(flat, down, flat, positive):
3004 triangle_case(flat, down, down, positive):
3005 setup_spans_down_a(psx_gpu, a, b, c);
3008 triangle_case(down, down, up, positive):
3009 triangle_case(down, down, flat, positive):
3010 triangle_case(down, down, down, positive):
3011 setup_spans_down_right(psx_gpu, a, b, c);
3016 spans += psx_gpu->num_spans;
3019 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3023 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3025 for(i = 0; i < psx_gpu->num_spans; i++)
3027 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3028 psx_gpu->span_edge_data[i].num_blocks = 0;
3033 for(i = 0; i < psx_gpu->num_spans; i++)
3035 if(psx_gpu->span_edge_data[i].y & 1)
3036 psx_gpu->span_edge_data[i].num_blocks = 0;
3041 u32 render_state = flags &
3042 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3043 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3044 render_state |= psx_gpu->render_state_base;
3046 if((psx_gpu->render_state != render_state) ||
3047 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3049 psx_gpu->render_state = render_state;
3050 flush_render_block_buffer(psx_gpu);
3056 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3058 psx_gpu->render_block_handler =
3059 &(render_triangle_block_handlers[render_state]);
3060 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3064 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3067 vertex_struct *vertex_ptrs[3];
3068 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3069 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3072 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3074 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3076 block_struct *block = psx_gpu->blocks;
3077 u32 num_blocks = psx_gpu->num_blocks;
3080 vec_8x8u texel_indexes;
3082 u16 *clut_ptr = psx_gpu->clut_ptr;
3087 texel_indexes = block->r;
3089 for(i = 0; i < 8; i++)
3091 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3094 block->texels = texels;
3104 #define setup_sprite_tiled_initialize_4bpp_clut() \
3105 u16 *clut_ptr = psx_gpu->clut_ptr; \
3106 vec_8x16u clut_a, clut_b; \
3107 vec_16x8u clut_low, clut_high; \
3109 load_8x16b(clut_a, clut_ptr); \
3110 load_8x16b(clut_b, clut_ptr + 8); \
3111 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3113 #define setup_sprite_tiled_initialize_4bpp() \
3114 setup_sprite_tiled_initialize_4bpp_clut(); \
3116 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3117 update_texture_4bpp_cache(psx_gpu) \
3119 #define setup_sprite_tiled_initialize_8bpp() \
3120 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3121 update_texture_8bpp_cache(psx_gpu) \
3124 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3125 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3126 ((texture_offset + offset) & texture_mask); \
3128 load_64b(texels, texture_block_ptr) \
3131 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3132 num_blocks += tile_num_blocks; \
3133 sprite_blocks += tile_num_blocks; \
3135 if(num_blocks > MAX_BLOCKS) \
3137 flush_render_block_buffer(psx_gpu); \
3138 num_blocks = tile_num_blocks; \
3139 block = psx_gpu->blocks; \
3142 #define setup_sprite_tile_full_4bpp(edge) \
3144 vec_8x8u texels_low, texels_high; \
3146 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3148 while(sub_tile_height) \
3150 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3151 tbl_16(texels_low, texels, clut_low); \
3152 tbl_16(texels_high, texels, clut_high); \
3153 zip_8x16b(pixels, texels_low, texels_high); \
3155 block->texels = pixels; \
3156 block->draw_mask_bits = left_mask_bits; \
3157 block->fb_ptr = fb_ptr; \
3160 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3161 tbl_16(texels_low, texels, clut_low); \
3162 tbl_16(texels_high, texels, clut_high); \
3163 zip_8x16b(pixels, texels_low, texels_high); \
3165 block->texels = pixels; \
3166 block->draw_mask_bits = right_mask_bits; \
3167 block->fb_ptr = fb_ptr + 8; \
3171 texture_offset += 0x10; \
3172 sub_tile_height--; \
3174 texture_offset += 0xF00; \
3175 psx_gpu->num_blocks = num_blocks; \
3178 #define setup_sprite_tile_half_4bpp(edge) \
3180 vec_8x8u texels_low, texels_high; \
3182 setup_sprite_tile_add_blocks(sub_tile_height); \
3184 while(sub_tile_height) \
3186 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3187 tbl_16(texels_low, texels, clut_low); \
3188 tbl_16(texels_high, texels, clut_high); \
3189 zip_8x16b(pixels, texels_low, texels_high); \
3191 block->texels = pixels; \
3192 block->draw_mask_bits = edge##_mask_bits; \
3193 block->fb_ptr = fb_ptr; \
3197 texture_offset += 0x10; \
3198 sub_tile_height--; \
3200 texture_offset += 0xF00; \
3201 psx_gpu->num_blocks = num_blocks; \
3205 #define setup_sprite_tile_full_8bpp(edge) \
3207 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3209 while(sub_tile_height) \
3211 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3212 block->r = texels; \
3213 block->draw_mask_bits = left_mask_bits; \
3214 block->fb_ptr = fb_ptr; \
3217 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3218 block->r = texels; \
3219 block->draw_mask_bits = right_mask_bits; \
3220 block->fb_ptr = fb_ptr + 8; \
3224 texture_offset += 0x10; \
3225 sub_tile_height--; \
3227 texture_offset += 0xF00; \
3228 psx_gpu->num_blocks = num_blocks; \
3231 #define setup_sprite_tile_half_8bpp(edge) \
3233 setup_sprite_tile_add_blocks(sub_tile_height); \
3235 while(sub_tile_height) \
3237 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3238 block->r = texels; \
3239 block->draw_mask_bits = edge##_mask_bits; \
3240 block->fb_ptr = fb_ptr; \
3244 texture_offset += 0x10; \
3245 sub_tile_height--; \
3247 texture_offset += 0xF00; \
3248 psx_gpu->num_blocks = num_blocks; \
3252 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3253 texture_offset = texture_offset_base + 8; \
3256 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3257 texture_offset = texture_offset_base \
3259 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3260 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3262 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3263 texture_offset = texture_offset_base \
3265 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3268 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3270 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3271 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3273 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3276 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3280 sub_tile_height = column_data; \
3281 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3282 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3283 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3286 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3290 u32 tiles_remaining = column_data >> 16; \
3291 sub_tile_height = column_data & 0xFF; \
3292 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3293 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3294 tiles_remaining -= 1; \
3296 while(tiles_remaining) \
3298 sub_tile_height = 16; \
3299 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3300 tiles_remaining--; \
3303 sub_tile_height = (column_data >> 8) & 0xFF; \
3304 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3305 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3309 #define setup_sprite_column_data_single() \
3310 column_data = height \
3312 #define setup_sprite_column_data_multi() \
3313 column_data = 16 - offset_v; \
3314 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3315 column_data |= (tile_height - 1) << 16 \
3318 #define RIGHT_MASK_BIT_SHIFT 8
3319 #define RIGHT_MASK_BIT_SHIFT_4x 16
3321 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3322 edge_mode, edge, x4mode) \
3324 setup_sprite_column_data_##multi_height(); \
3325 left_mask_bits = left_block_mask | right_block_mask; \
3326 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3328 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3329 texture_mode, x4mode); \
3332 #define setup_sprite_tiled_advance_column() \
3333 texture_offset_base += 0x100; \
3334 if((texture_offset_base & 0xF00) == 0) \
3335 texture_offset_base -= (0x100 + 0xF00) \
3337 #define FB_PTR_MULTIPLIER 1
3338 #define FB_PTR_MULTIPLIER_4x 2
3340 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3341 left_mode, right_mode, x4mode) \
3343 setup_sprite_column_data_##multi_height(); \
3344 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3345 * FB_PTR_MULTIPLIER##x4mode; \
3348 left_mask_bits = left_block_mask; \
3349 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3351 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3352 texture_mode, x4mode); \
3353 fb_ptr += fb_ptr_advance_column; \
3355 left_mask_bits = 0x00; \
3356 right_mask_bits = 0x00; \
3360 setup_sprite_tiled_advance_column(); \
3361 setup_sprite_tile_column_height_##multi_height(full, none, \
3362 texture_mode, x4mode); \
3363 fb_ptr += fb_ptr_advance_column; \
3367 left_mask_bits = right_block_mask; \
3368 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3370 setup_sprite_tiled_advance_column(); \
3371 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3372 texture_mode, x4mode); \
3377 #define setup_sprite_tiled_initialize_4bpp_4x() \
3378 setup_sprite_tiled_initialize_4bpp_clut() \
3380 #define setup_sprite_tiled_initialize_8bpp_4x() \
3383 #define setup_sprite_tile_full_4bpp_4x(edge) \
3385 vec_8x8u texels_low, texels_high; \
3386 vec_8x16u pixels, pixels_wide; \
3387 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3388 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3389 u32 left_mask_bits_b = left_mask_bits >> 8; \
3390 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3391 u32 right_mask_bits_b = right_mask_bits >> 8; \
3393 while(sub_tile_height) \
3395 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3396 tbl_16(texels_low, texels, clut_low); \
3397 tbl_16(texels_high, texels, clut_high); \
3398 zip_8x16b(pixels, texels_low, texels_high); \
3400 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3401 block->texels = pixels_wide; \
3402 block->draw_mask_bits = left_mask_bits_a; \
3403 block->fb_ptr = fb_ptr; \
3406 block->texels = pixels_wide; \
3407 block->draw_mask_bits = left_mask_bits_a; \
3408 block->fb_ptr = fb_ptr + 1024; \
3411 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3412 block->texels = pixels_wide; \
3413 block->draw_mask_bits = left_mask_bits_b; \
3414 block->fb_ptr = fb_ptr + 8; \
3417 block->texels = pixels_wide; \
3418 block->draw_mask_bits = left_mask_bits_b; \
3419 block->fb_ptr = fb_ptr + 1024 + 8; \
3422 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3423 tbl_16(texels_low, texels, clut_low); \
3424 tbl_16(texels_high, texels, clut_high); \
3425 zip_8x16b(pixels, texels_low, texels_high); \
3427 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3428 block->texels = pixels_wide; \
3429 block->draw_mask_bits = right_mask_bits_a; \
3430 block->fb_ptr = fb_ptr + 16; \
3433 block->texels = pixels_wide; \
3434 block->draw_mask_bits = right_mask_bits_a; \
3435 block->fb_ptr = fb_ptr + 1024 + 16; \
3438 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3439 block->texels = pixels_wide; \
3440 block->draw_mask_bits = right_mask_bits_b; \
3441 block->fb_ptr = fb_ptr + 24; \
3444 block->texels = pixels_wide; \
3445 block->draw_mask_bits = right_mask_bits_b; \
3446 block->fb_ptr = fb_ptr + 1024 + 24; \
3450 texture_offset += 0x10; \
3451 sub_tile_height--; \
3453 texture_offset += 0xF00; \
3454 psx_gpu->num_blocks = num_blocks; \
3457 #define setup_sprite_tile_half_4bpp_4x(edge) \
3459 vec_8x8u texels_low, texels_high; \
3460 vec_8x16u pixels, pixels_wide; \
3461 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3462 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3463 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3465 while(sub_tile_height) \
3467 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3468 tbl_16(texels_low, texels, clut_low); \
3469 tbl_16(texels_high, texels, clut_high); \
3470 zip_8x16b(pixels, texels_low, texels_high); \
3472 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3473 block->texels = pixels_wide; \
3474 block->draw_mask_bits = edge##_mask_bits_a; \
3475 block->fb_ptr = fb_ptr; \
3478 block->texels = pixels_wide; \
3479 block->draw_mask_bits = edge##_mask_bits_a; \
3480 block->fb_ptr = fb_ptr + 1024; \
3483 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3484 block->texels = pixels_wide; \
3485 block->draw_mask_bits = edge##_mask_bits_b; \
3486 block->fb_ptr = fb_ptr + 8; \
3489 block->texels = pixels_wide; \
3490 block->draw_mask_bits = edge##_mask_bits_b; \
3491 block->fb_ptr = fb_ptr + 1024 + 8; \
3495 texture_offset += 0x10; \
3496 sub_tile_height--; \
3498 texture_offset += 0xF00; \
3499 psx_gpu->num_blocks = num_blocks; \
3503 #define setup_sprite_tile_full_8bpp_4x(edge) \
3505 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3506 vec_16x8u texels_wide; \
3507 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3508 u32 left_mask_bits_b = left_mask_bits >> 8; \
3509 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3510 u32 right_mask_bits_b = right_mask_bits >> 8; \
3512 while(sub_tile_height) \
3514 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3515 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3516 block->r = texels_wide.low; \
3517 block->draw_mask_bits = left_mask_bits_a; \
3518 block->fb_ptr = fb_ptr; \
3521 block->r = texels_wide.low; \
3522 block->draw_mask_bits = left_mask_bits_a; \
3523 block->fb_ptr = fb_ptr + 1024; \
3526 block->r = texels_wide.high; \
3527 block->draw_mask_bits = left_mask_bits_b; \
3528 block->fb_ptr = fb_ptr + 8; \
3531 block->r = texels_wide.high; \
3532 block->draw_mask_bits = left_mask_bits_b; \
3533 block->fb_ptr = fb_ptr + 1024 + 8; \
3536 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3537 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3538 block->r = texels_wide.low; \
3539 block->draw_mask_bits = right_mask_bits_a; \
3540 block->fb_ptr = fb_ptr + 16; \
3543 block->r = texels_wide.low; \
3544 block->draw_mask_bits = right_mask_bits_a; \
3545 block->fb_ptr = fb_ptr + 1024 + 16; \
3548 block->r = texels_wide.high; \
3549 block->draw_mask_bits = right_mask_bits_b; \
3550 block->fb_ptr = fb_ptr + 24; \
3553 block->r = texels_wide.high; \
3554 block->draw_mask_bits = right_mask_bits_b; \
3555 block->fb_ptr = fb_ptr + 24 + 1024; \
3559 texture_offset += 0x10; \
3560 sub_tile_height--; \
3562 texture_offset += 0xF00; \
3563 psx_gpu->num_blocks = num_blocks; \
3566 #define setup_sprite_tile_half_8bpp_4x(edge) \
3568 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3569 vec_16x8u texels_wide; \
3570 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3571 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3573 while(sub_tile_height) \
3575 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3576 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3577 block->r = texels_wide.low; \
3578 block->draw_mask_bits = edge##_mask_bits_a; \
3579 block->fb_ptr = fb_ptr; \
3582 block->r = texels_wide.low; \
3583 block->draw_mask_bits = edge##_mask_bits_a; \
3584 block->fb_ptr = fb_ptr + 1024; \
3587 block->r = texels_wide.high; \
3588 block->draw_mask_bits = edge##_mask_bits_b; \
3589 block->fb_ptr = fb_ptr + 8; \
3592 block->r = texels_wide.high; \
3593 block->draw_mask_bits = edge##_mask_bits_b; \
3594 block->fb_ptr = fb_ptr + 8 + 1024; \
3598 texture_offset += 0x10; \
3599 sub_tile_height--; \
3601 texture_offset += 0xF00; \
3602 psx_gpu->num_blocks = num_blocks; \
3606 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3607 texture_offset = texture_offset_base + 8; \
3610 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3611 texture_offset = texture_offset_base \
3613 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3614 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3616 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3617 texture_offset = texture_offset_base \
3619 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3622 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3624 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3625 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3627 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3630 #define setup_sprite_offset_u_adjust() \
3632 #define setup_sprite_comapre_left_block_mask() \
3633 ((left_block_mask & 0xFF) == 0xFF) \
3635 #define setup_sprite_comapre_right_block_mask() \
3636 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3639 #define setup_sprite_offset_u_adjust_4x() \
3641 offset_u_right = offset_u_right * 2 + 1 \
3643 #define setup_sprite_comapre_left_block_mask_4x() \
3644 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3646 #define setup_sprite_comapre_right_block_mask_4x() \
3647 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3650 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3651 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3652 s32 u, s32 v, s32 width, s32 height, u32 color) \
3654 s32 offset_u = u & 0xF; \
3655 s32 offset_v = v & 0xF; \
3657 s32 width_rounded = offset_u + width + 15; \
3658 s32 height_rounded = offset_v + height + 15; \
3659 s32 tile_height = height_rounded / 16; \
3660 s32 tile_width = width_rounded / 16; \
3661 u32 offset_u_right = width_rounded & 0xF; \
3663 setup_sprite_offset_u_adjust##x4mode(); \
3665 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3666 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3668 u32 left_mask_bits; \
3669 u32 right_mask_bits; \
3671 u32 sub_tile_height; \
3674 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3675 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3676 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3677 ((psx_gpu->texture_mask_height >> 4) << 12); \
3678 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3679 ((v & 0xF0) << 8); \
3680 u32 texture_offset_base = texture_offset; \
3683 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3684 u32 num_blocks = psx_gpu->num_blocks; \
3685 block_struct *block = psx_gpu->blocks + num_blocks; \
3687 u8 *texture_block_ptr; \
3690 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3692 control_mask = tile_width == 1; \
3693 control_mask |= (tile_height == 1) << 1; \
3694 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3695 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3697 sprites_##texture_mode++; \
3699 switch(control_mask) \
3703 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3708 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3713 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3718 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3723 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3728 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3733 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3738 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3743 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3748 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3753 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3758 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3763 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3768 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3775 setup_sprite_tiled_builder(4bpp,);
3776 setup_sprite_tiled_builder(8bpp,);
3778 setup_sprite_tiled_builder(4bpp,_4x);
3779 setup_sprite_tiled_builder(8bpp,_4x);
3782 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3784 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3785 s32 v, s32 width, s32 height, u32 color)
3787 u32 left_offset = u & 0x7;
3788 u32 width_rounded = width + left_offset + 7;
3790 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3791 u32 right_width = width_rounded & 0x7;
3792 u32 block_width = width_rounded / 8;
3793 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3795 u32 left_mask_bits = ~(0xFF << left_offset);
3796 u32 right_mask_bits = 0xFE << right_width;
3798 u32 texture_offset_base = u + (v * 1024);
3800 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3802 u32 blocks_remaining;
3803 u32 num_blocks = psx_gpu->num_blocks;
3804 block_struct *block = psx_gpu->blocks + num_blocks;
3806 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3807 u16 *texture_block_ptr;
3809 texture_offset_base &= ~0x7;
3811 stats_add(sprites_16bpp, 1);
3813 if(block_width == 1)
3815 u32 mask_bits = left_mask_bits | right_mask_bits;
3822 if(num_blocks > MAX_BLOCKS)
3824 flush_render_block_buffer(psx_gpu);
3826 block = psx_gpu->blocks;
3830 texture_page_ptr + (texture_offset_base & texture_mask);
3832 block->texels = *(vec_8x16u *)texture_block_ptr;
3833 block->draw_mask_bits = mask_bits;
3834 block->fb_ptr = fb_ptr;
3838 texture_offset_base += 1024;
3842 psx_gpu->num_blocks = num_blocks;
3851 blocks_remaining = block_width - 2;
3852 num_blocks += block_width;
3853 sprite_blocks += block_width;
3855 if(num_blocks > MAX_BLOCKS)
3857 flush_render_block_buffer(psx_gpu);
3858 num_blocks = block_width;
3859 block = psx_gpu->blocks;
3862 texture_offset = texture_offset_base;
3863 texture_offset_base += 1024;
3865 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3866 block->texels = *(vec_8x16u *)texture_block_ptr;
3868 block->draw_mask_bits = left_mask_bits;
3869 block->fb_ptr = fb_ptr;
3871 texture_offset += 8;
3875 while(blocks_remaining)
3877 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3878 block->texels = *(vec_8x16u *)texture_block_ptr;
3880 block->draw_mask_bits = 0;
3881 block->fb_ptr = fb_ptr;
3883 texture_offset += 8;
3890 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3891 block->texels = *(vec_8x16u *)texture_block_ptr;
3893 block->draw_mask_bits = right_mask_bits;
3894 block->fb_ptr = fb_ptr;
3896 fb_ptr += fb_ptr_pitch;
3900 psx_gpu->num_blocks = num_blocks;
3909 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3910 s32 v, s32 width, s32 height, u32 color)
3912 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
3913 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
3914 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
3916 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
3920 u32 right_width = ((width - 1) & 0x7) + 1;
3921 u32 right_mask_bits = (0xFF << right_width);
3922 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3923 u32 block_width = (width + 7) / 8;
3924 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3925 u32 blocks_remaining;
3926 u32 num_blocks = psx_gpu->num_blocks;
3927 block_struct *block = psx_gpu->blocks + num_blocks;
3929 u32 color_r = color & 0xFF;
3930 u32 color_g = (color >> 8) & 0xFF;
3931 u32 color_b = (color >> 16) & 0xFF;
3933 vec_8x16u right_mask;
3934 vec_8x16u test_mask = psx_gpu->test_mask;
3935 vec_8x16u zero_mask;
3937 sprites_untextured++;
3939 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3941 dup_8x16b(colors, color);
3942 dup_8x16b(zero_mask, 0x00);
3943 dup_8x16b(right_mask, right_mask_bits);
3944 tst_8x16b(right_mask, right_mask, test_mask);
3948 blocks_remaining = block_width - 1;
3949 num_blocks += block_width;
3952 sprite_blocks += block_width;
3955 if(num_blocks > MAX_BLOCKS)
3957 flush_render_block_buffer(psx_gpu);
3958 num_blocks = block_width;
3959 block = psx_gpu->blocks;
3962 while(blocks_remaining)
3964 block->pixels = colors;
3965 block->draw_mask = zero_mask;
3966 block->fb_ptr = fb_ptr;
3973 block->pixels = colors;
3974 block->draw_mask = right_mask;
3975 block->fb_ptr = fb_ptr;
3978 fb_ptr += fb_ptr_pitch;
3981 psx_gpu->num_blocks = num_blocks;
3987 void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y,
3988 s32 u, s32 v, s32 width, s32 height, u32 color)
3990 u32 r = color & 0xFF;
3991 u32 g = (color >> 8) & 0xFF;
3992 u32 b = (color >> 16) & 0xFF;
3993 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
3995 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
3997 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4002 if(psx_gpu->num_blocks > MAX_BLOCKS)
4004 flush_render_block_buffer(psx_gpu);
4011 vram_ptr = (void *)vram_ptr16;
4012 if((long)vram_ptr16 & 2)
4014 *vram_ptr16 = color_32bpp;
4015 vram_ptr = (void *)(vram_ptr16 + 1);
4019 while(num_width >= 4 * 2)
4021 vram_ptr[0] = color_32bpp;
4022 vram_ptr[1] = color_32bpp;
4023 vram_ptr[2] = color_32bpp;
4024 vram_ptr[3] = color_32bpp;
4030 while(num_width >= 2)
4032 *vram_ptr++ = color_32bpp;
4038 *(u16 *)vram_ptr = color_32bpp;
4047 #define setup_sprite_blocks_switch_textured(texture_mode) \
4048 setup_sprite_##texture_mode \
4050 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4051 setup_sprite_untextured \
4053 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4054 setup_sprite_blocks_switch_##texturing(texture_mode) \
4057 #define texture_sprite_blocks_switch_4bpp() \
4058 texture_blocks_untextured \
4060 #define texture_sprite_blocks_switch_8bpp() \
4061 texture_sprite_blocks_8bpp \
4063 #define texture_sprite_blocks_switch_16bpp() \
4064 texture_blocks_untextured \
4066 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4067 texture_blocks_untextured \
4069 #define texture_sprite_blocks_switch_textured(texture_mode) \
4070 texture_sprite_blocks_switch_##texture_mode() \
4072 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4073 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4075 setup_sprite_blocks_switch(texturing, texture_mode), \
4076 texture_sprite_blocks_switch_##texturing(texture_mode), \
4077 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4079 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4082 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4083 mask_evaluate, shading, dithering, texturing, blending) \
4084 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4085 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4086 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4087 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4089 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4090 mask_evaluate, shading, dithering, texturing) \
4091 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4092 mask_evaluate, shading, dithering, texturing, unblended), \
4093 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4094 mask_evaluate, shading, dithering, texturing, blended) \
4096 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4097 mask_evaluate, shading, dithering) \
4098 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4099 mask_evaluate, shading, dithering, untextured), \
4100 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4101 mask_evaluate, shading, dithering, textured) \
4103 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4104 mask_evaluate, shading) \
4105 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4106 mask_evaluate, shading, undithered), \
4107 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4108 mask_evaluate, shading, dithered) \
4110 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4111 blend_mode, mask_evaluate) \
4112 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4113 mask_evaluate, unshaded), \
4114 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4115 mask_evaluate, shaded) \
4117 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4118 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4120 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4123 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4124 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4125 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4126 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4127 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4129 #define render_sprite_blocks_switch_block() \
4130 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4131 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4132 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4133 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4136 render_block_handler_struct render_sprite_block_handlers[] =
4138 render_sprite_blocks_switch_block()
4142 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4143 s32 width, s32 height, u32 flags, u32 color)
4145 s32 x_right = x + width - 1;
4146 s32 y_bottom = y + height - 1;
4152 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4158 if(x < psx_gpu->viewport_start_x)
4160 u32 clip = psx_gpu->viewport_start_x - x;
4166 if(y < psx_gpu->viewport_start_y)
4168 s32 clip = psx_gpu->viewport_start_y - y;
4174 if(x_right > psx_gpu->viewport_end_x)
4175 width -= x_right - psx_gpu->viewport_end_x;
4177 if(y_bottom > psx_gpu->viewport_end_y)
4178 height -= y_bottom - psx_gpu->viewport_end_y;
4180 if((width <= 0) || (height <= 0))
4184 span_pixels += width * height;
4188 u32 render_state = flags &
4189 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4190 RENDER_FLAGS_TEXTURE_MAP);
4192 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4194 if((psx_gpu->render_state != render_state) ||
4195 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4197 psx_gpu->render_state = render_state;
4198 flush_render_block_buffer(psx_gpu);
4204 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4208 if(psx_gpu->triangle_color != color)
4210 flush_render_block_buffer(psx_gpu);
4211 psx_gpu->triangle_color = color;
4214 if(color == 0x808080)
4215 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4217 render_block_handler_struct *render_block_handler =
4218 &(render_sprite_block_handlers[render_state]);
4219 psx_gpu->render_block_handler = render_block_handler;
4221 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4222 (psx_gpu, x, y, u, v, width, height, color);
4225 #define draw_pixel_line_mask_evaluate_yes() \
4226 if((*vram_ptr & 0x8000) == 0) \
4228 #define draw_pixel_line_mask_evaluate_no() \
4231 #define draw_pixel_line_shaded() \
4233 color_r = fixed_to_int(current_r); \
4234 color_g = fixed_to_int(current_g); \
4235 color_b = fixed_to_int(current_b); \
4237 current_r += gradient_r; \
4238 current_g += gradient_g; \
4239 current_b += gradient_b; \
4242 #define draw_pixel_line_unshaded() \
4244 color_r = color & 0xFF; \
4245 color_g = (color >> 8) & 0xFF; \
4246 color_b = (color >> 16) & 0xFF; \
4250 #define draw_pixel_line_dithered(_x, _y) \
4252 u32 dither_xor = _x ^ _y; \
4253 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4254 dither_offset |= (_y & 0x1) << 1; \
4255 dither_offset |= (dither_xor & 0x1) << 2; \
4256 dither_offset -= 4; \
4258 color_r += dither_offset; \
4259 color_g += dither_offset; \
4260 color_b += dither_offset; \
4281 #define draw_pixel_line_undithered(_x, _y) \
4284 #define draw_pixel_line_average() \
4285 color_r = (color_r + fb_r) / 2; \
4286 color_g = (color_g + fb_g) / 2; \
4287 color_b = (color_b + fb_b) / 2 \
4289 #define draw_pixel_line_add() \
4304 #define draw_pixel_line_subtract() \
4305 color_r = fb_r - color_r; \
4306 color_g = fb_g - color_g; \
4307 color_b = fb_b - color_b; \
4318 #define draw_pixel_line_add_fourth() \
4319 color_r = fb_r + (color_r / 4); \
4320 color_g = fb_g + (color_g / 4); \
4321 color_b = fb_b + (color_b / 4); \
4333 #define draw_pixel_line_blended(blend_mode) \
4334 s32 fb_pixel = *vram_ptr; \
4335 s32 fb_r = fb_pixel & 0x1F; \
4336 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4337 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4339 draw_pixel_line_##blend_mode() \
4341 #define draw_pixel_line_unblended(blend_mode) \
4344 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4346 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4347 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4349 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4351 draw_pixel_line_##shading(); \
4352 draw_pixel_line_##dithering(_x, _y); \
4358 draw_pixel_line_##blending(blend_mode); \
4360 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4361 psx_gpu->mask_msb; \
4365 #define update_increment(value) \
4368 #define update_decrement(value) \
4371 #define update_vram_row_increment(value) \
4374 #define update_vram_row_decrement(value) \
4377 #define compare_increment(a, b) \
4380 #define compare_decrement(a, b) \
4383 #define set_line_gradients(minor) \
4385 s32 gradient_divisor = delta_##minor; \
4386 if(gradient_divisor != 0) \
4388 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4389 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4390 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4398 current_r = fixed_center(vertex_a->r); \
4399 current_g = fixed_center(vertex_a->g); \
4400 current_b = fixed_center(vertex_a->b); \
4403 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4404 mask_evaluate, blend_mode) \
4407 error_step = delta_y * 2; \
4408 error_wrap = delta_x * 2; \
4412 set_line_gradients(x); \
4414 for(current_x = x_a; current_x <= x_b; current_x++) \
4416 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4417 mask_evaluate, blend_mode); \
4418 error += error_step; \
4421 if(error >= error_wrap) \
4423 update_##direction(current_y); \
4424 update_vram_row_##direction(); \
4425 error -= error_wrap; \
4430 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4431 mask_evaluate, blend_mode) \
4434 error_step = delta_x * 2; \
4435 error_wrap = delta_y * 2; \
4439 set_line_gradients(y); \
4441 for(current_y = y_a; compare_##direction(current_y, y_b); \
4442 update_##direction(current_y)) \
4444 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4445 mask_evaluate, blend_mode); \
4446 error += error_step; \
4447 update_vram_row_##direction(); \
4449 if(error > error_wrap) \
4453 error -= error_wrap; \
4459 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4465 if(delta_x > delta_y) \
4467 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4468 mask_evaluate, blend_mode); \
4472 draw_line_span_vertical(decrement, shading, blending, dithering, \
4473 mask_evaluate, blend_mode); \
4478 if(delta_x > delta_y) \
4480 draw_line_span_horizontal(increment, shading, blending, dithering, \
4481 mask_evaluate, blend_mode); \
4485 draw_line_span_vertical(increment, shading, blending, dithering, \
4486 mask_evaluate, blend_mode); \
4491 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4492 u32 color, int double_resolution)
4494 s32 color_r, color_g, color_b;
4495 u32 triangle_winding = 0;
4497 fixed_type gradient_r = 0;
4498 fixed_type gradient_g = 0;
4499 fixed_type gradient_b = 0;
4500 fixed_type current_r = 0;
4501 fixed_type current_g = 0;
4502 fixed_type current_b = 0;
4507 s32 delta_x, delta_y;
4518 flush_render_block_buffer(psx_gpu);
4519 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4521 vertex_struct *vertex_a = &(vertexes[0]);
4522 vertex_struct *vertex_b = &(vertexes[1]);
4530 if(vertex_a->x >= vertex_b->x)
4532 vertex_swap(vertex_a, vertex_b);
4541 delta_x = x_b - x_a;
4542 delta_y = y_b - y_a;
4544 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4547 if(double_resolution)
4557 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4559 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4563 if(flags & RENDER_FLAGS_SHADE)
4564 control_mask |= 0x1;
4566 if(flags & RENDER_FLAGS_BLEND)
4568 control_mask |= 0x2;
4569 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4572 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4573 control_mask |= 0x4;
4575 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4576 control_mask |= 0x8;
4578 switch(control_mask)
4581 render_line_body(unshaded, unblended, undithered, no, none);
4585 render_line_body(shaded, unblended, undithered, no, none);
4589 render_line_body(unshaded, blended, undithered, no, average);
4593 render_line_body(shaded, blended, undithered, no, average);
4597 render_line_body(unshaded, unblended, dithered, no, none);
4601 render_line_body(shaded, unblended, dithered, no, none);
4605 render_line_body(unshaded, blended, dithered, no, average);
4609 render_line_body(shaded, blended, dithered, no, average);
4613 render_line_body(unshaded, unblended, undithered, yes, none);
4617 render_line_body(shaded, unblended, undithered, yes, none);
4621 render_line_body(unshaded, blended, undithered, yes, average);
4625 render_line_body(shaded, blended, undithered, yes, average);
4629 render_line_body(unshaded, unblended, dithered, yes, none);
4633 render_line_body(shaded, unblended, dithered, yes, none);
4637 render_line_body(unshaded, blended, dithered, yes, average);
4641 render_line_body(shaded, blended, dithered, yes, average);
4645 render_line_body(unshaded, blended, undithered, no, add);
4649 render_line_body(shaded, blended, undithered, no, add);
4653 render_line_body(unshaded, blended, dithered, no, add);
4657 render_line_body(shaded, blended, dithered, no, add);
4661 render_line_body(unshaded, blended, undithered, yes, add);
4665 render_line_body(shaded, blended, undithered, yes, add);
4669 render_line_body(unshaded, blended, dithered, yes, add);
4673 render_line_body(shaded, blended, dithered, yes, add);
4677 render_line_body(unshaded, blended, undithered, no, subtract);
4681 render_line_body(shaded, blended, undithered, no, subtract);
4685 render_line_body(unshaded, blended, dithered, no, subtract);
4689 render_line_body(shaded, blended, dithered, no, subtract);
4693 render_line_body(unshaded, blended, undithered, yes, subtract);
4697 render_line_body(shaded, blended, undithered, yes, subtract);
4701 render_line_body(unshaded, blended, dithered, yes, subtract);
4705 render_line_body(shaded, blended, dithered, yes, subtract);
4709 render_line_body(unshaded, blended, undithered, no, add_fourth);
4713 render_line_body(shaded, blended, undithered, no, add_fourth);
4717 render_line_body(unshaded, blended, dithered, no, add_fourth);
4721 render_line_body(shaded, blended, dithered, no, add_fourth);
4725 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4729 render_line_body(shaded, blended, undithered, yes, add_fourth);
4733 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4737 render_line_body(shaded, blended, dithered, yes, add_fourth);
4743 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4744 u32 width, u32 height)
4746 if((width == 0) || (height == 0))
4749 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4751 u32 r = color & 0xFF;
4752 u32 g = (color >> 8) & 0xFF;
4753 u32 b = (color >> 16) & 0xFF;
4754 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4756 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4758 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4760 u32 pitch = 512 - (width / 2);
4763 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4768 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4777 vram_ptr[0] = color_32bpp;
4778 vram_ptr[1] = color_32bpp;
4779 vram_ptr[2] = color_32bpp;
4780 vram_ptr[3] = color_32bpp;
4781 vram_ptr[4] = color_32bpp;
4782 vram_ptr[5] = color_32bpp;
4783 vram_ptr[6] = color_32bpp;
4784 vram_ptr[7] = color_32bpp;
4795 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4796 u32 width, u32 height)
4798 if((width == 0) || (height == 0))
4804 u32 r = color & 0xFF;
4805 u32 g = (color >> 8) & 0xFF;
4806 u32 b = (color >> 16) & 0xFF;
4807 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4809 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4811 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4813 u32 pitch = 1024 / 2 - (width / 2);
4821 vram_ptr[0] = color_32bpp;
4822 vram_ptr[1] = color_32bpp;
4823 vram_ptr[2] = color_32bpp;
4824 vram_ptr[3] = color_32bpp;
4825 vram_ptr[4] = color_32bpp;
4826 vram_ptr[5] = color_32bpp;
4827 vram_ptr[6] = color_32bpp;
4828 vram_ptr[7] = color_32bpp;
4839 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4840 u32 width, u32 height, u32 pitch)
4842 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4844 u32 mask_msb = psx_gpu->mask_msb;
4846 if((width == 0) || (height == 0))
4849 flush_render_block_buffer(psx_gpu);
4850 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4852 for(draw_y = 0; draw_y < height; draw_y++)
4854 for(draw_x = 0; draw_x < width; draw_x++)
4856 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4864 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4865 u32 dest_x, u32 dest_y, u32 width, u32 height)
4867 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4868 dest_x, dest_y, width, height, 1024);
4872 void initialize_reciprocal_table(void)
4875 u32 height_normalized;
4876 u32 height_reciprocal;
4879 for(height = 1; height < sizeof(reciprocal_table)
4880 / sizeof(reciprocal_table[0]); height++)
4882 shift = __builtin_clz(height);
4883 height_normalized = height << shift;
4884 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4887 shift = 32 - (51 - shift);
4889 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4894 #define dither_table_row(a, b, c, d) \
4895 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4897 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4899 vec_8x16u test_mask =
4900 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4902 psx_gpu->test_mask = test_mask;
4904 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4905 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4906 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4907 psx_gpu->viewport_mask = 0;
4908 psx_gpu->current_texture_page = 0;
4909 psx_gpu->current_texture_mask = 0;
4910 psx_gpu->last_8bpp_texture_page = 0;
4912 psx_gpu->clut_settings = 0;
4913 psx_gpu->texture_settings = 0;
4914 psx_gpu->render_state = 0;
4915 psx_gpu->render_state_base = 0;
4916 psx_gpu->num_blocks = 0;
4917 psx_gpu->uvrgb_phase = 0x8000;
4919 psx_gpu->vram_ptr = vram;
4920 psx_gpu->vram_out_ptr = vram;
4922 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4923 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4924 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4926 psx_gpu->mask_msb = 0;
4928 psx_gpu->texture_window_x = 0;
4929 psx_gpu->texture_window_y = 0;
4930 psx_gpu->texture_mask_width = 0xFF;
4931 psx_gpu->texture_mask_height = 0xFF;
4933 psx_gpu->render_mode = 0;
4935 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4937 initialize_reciprocal_table();
4938 psx_gpu->reciprocal_table_ptr = reciprocal_table;
4945 // (minus ones(4) * 4)
4947 // d0: (1 3 5 7): x1 ^ y1
4948 // d1: (2 3 6 7): y0
4949 // d2: (4 5 6 7): x0 ^ y0
4951 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
4952 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
4953 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
4954 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
4956 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
4958 psx_gpu->enhancement_x_threshold = 256;
4964 gettimeofday(&tv, NULL);
4966 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
4969 #if 0 //def NEON_BUILD
4974 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
4979 void init_counter(void)
4982 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
4983 value |= 5; // master enable, ccnt reset
4984 value &= ~8; // ccnt divider 0
4985 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
4986 // enable cycle counter
4987 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
4990 void triangle_benchmark(psx_gpu_struct *psx_gpu)
4997 const u32 iterations = 500000;
4999 psx_gpu->num_blocks = 64;
5000 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5002 for(i = 0; i < 64; i++)
5004 memset(&(psx_gpu->blocks[i].r), 0, 16);
5009 ticks = get_counter();
5011 for(i = 0; i < iterations; i++)
5013 texture_sprite_blocks_8bpp(psx_gpu);
5016 ticks_elapsed = get_counter() - ticks;
5018 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5023 #include "psx_gpu_4x.c"