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.
22 #include "vector_ops.h"
24 #include "psx_gpu_simd.h"
27 void dump_r_d(const char *name, void *dump);
28 void dump_r_q(const char *name, void *dump);
29 #define dumprd(n) dump_r_d(#n, n.e)
30 #define dumprq(n) dump_r_q(#n, n.e)
34 u32 span_pixel_blocks = 0;
40 u32 sprites_16bpp = 0;
41 u32 sprite_blocks = 0;
42 u32 sprites_untextured = 0;
44 u32 trivial_rejects = 0;
48 u32 texel_blocks_4bpp = 0;
49 u32 texel_blocks_8bpp = 0;
50 u32 texel_blocks_16bpp = 0;
51 u32 texel_blocks_untextured = 0;
53 u32 render_buffer_flushes = 0;
54 u32 state_changes = 0;
55 u32 left_split_triangles = 0;
56 u32 flat_triangles = 0;
57 u32 clipped_triangles = 0;
58 u32 zero_block_spans = 0;
59 u32 texture_cache_loads = 0;
60 u32 false_modulated_blocks = 0;
62 #define stats_add(stat, count) // stat += count
64 /* double size for enhancement */
65 u32 reciprocal_table[512 * 2];
68 typedef s32 fixed_type;
70 #define EDGE_STEP_BITS 32
73 #define fixed_center(value) \
74 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
76 #define int_to_fixed(value) \
77 (((fixed_type)(value)) << FIXED_BITS) \
79 #define fixed_to_int(value) \
80 ((value) >> FIXED_BITS) \
82 #define fixed_to_double(value) \
83 ((value) / (double)(1 << FIXED_BITS)) \
85 #define double_to_fixed(value) \
86 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
88 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
89 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
90 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
91 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
93 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
94 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
96 struct render_block_handler_struct
99 texture_blocks_function_type *texture_blocks;
100 shade_blocks_function_type *shade_blocks;
101 blend_blocks_function_type *blend_blocks;
106 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
108 u32 shift = __builtin_clz(denominator);
109 u32 denominator_normalized = denominator << shift;
111 double numerator = (1ULL << 62) + denominator_normalized;
114 double denominator_normalized_dp_b;
115 u64 denominator_normalized_dp_u64;
118 double reciprocal_dp;
120 u64 numerator_u64 = (denominator_normalized >> 10) |
121 ((u64)(62 + 1023) << 52);
122 *((u64 *)(&numerator_b)) = numerator_u64;
124 denominator_normalized_dp_u64 =
125 (u64)(denominator_normalized << 21) |
126 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
127 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
129 // Implement with a DP divide
130 reciprocal_dp = numerator / denominator_normalized_dp_b;
131 reciprocal = reciprocal_dp;
133 if(reciprocal == 0x80000001)
134 reciprocal = 0x80000000;
136 *_shift = 62 - shift;
140 double reciprocal_estimate(double a)
145 q = (int)(a * 512.0);
146 /* a in units of 1/512 rounded down */
147 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
148 s = (int)(256.0 * r + 0.5);
150 /* r in units of 1/256 rounded to nearest */
152 return (double)s / 256.0;
155 u32 reciprocal_estimate_u32(u32 value)
158 volatile double dp_value;
159 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
161 if((value >> 31) == 0)
164 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
166 *dp_value_ptr = dp_value_u64;
168 dp_value = reciprocal_estimate(dp_value);
169 dp_value_u64 = *dp_value_ptr;
171 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
174 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
176 u32 shift = __builtin_clz(value);
177 u32 value_normalized = value << shift;
179 *_shift = 62 - shift;
181 value_normalized -= 2;
183 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
185 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
186 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
187 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
188 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
189 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
190 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
192 return reciprocal_normalized;
198 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
200 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
203 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
205 s32 coverage_x, coverage_y;
210 coverage_x = x2 >> 6;
211 coverage_y = y2 >> 8;
219 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
222 mask_down_right |= mask_down_right << 16;
224 coverage_x = x1 >> 6;
226 mask_up_left = 0xFFFF0000 << coverage_x;
228 mask_up_left = 0xFFFF0000;
230 coverage_y = y1 >> 8;
232 mask_up_left |= mask_up_left >> 16;
234 return mask_up_left & mask_down_right;
237 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
240 u32 mask = texture_region_mask(x1, y1, x2, y2);
242 psx_gpu->dirty_textures_4bpp_mask |= mask;
243 psx_gpu->dirty_textures_8bpp_mask |= mask;
244 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
249 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
250 u32 y1, u32 x2, u32 y2)
252 u32 mask = texture_region_mask(x1, y1, x2, y2) &
253 psx_gpu->viewport_mask;
255 psx_gpu->dirty_textures_4bpp_mask |= mask;
256 psx_gpu->dirty_textures_8bpp_mask |= mask;
257 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
262 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
265 u32 mask = texture_region_mask(x1, y1, x2, y2);
267 u8 *texture_page_ptr;
272 psx_gpu->dirty_textures_8bpp_mask |= mask;
273 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
275 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
276 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
278 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
279 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
280 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
281 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
289 texel_block = *vram_ptr;
291 texture_page_ptr[0] = texel_block & 0xF;
292 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
293 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
294 texture_page_ptr[3] = texel_block >> 12;
297 texture_page_ptr += 4;
311 psx_gpu->dirty_textures_4bpp_mask |= mask;
317 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
319 u32 current_texture_page = psx_gpu->current_texture_page;
320 u8 *texture_page_ptr = psx_gpu->texture_page_base;
321 u16 *vram_ptr = psx_gpu->vram_ptr;
327 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
328 vram_ptr += (current_texture_page & 0xF) * 64;
330 texture_cache_loads++;
337 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
347 texel_block = *vram_ptr;
349 texture_page_ptr[0] = texel_block & 0xF;
350 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
351 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
352 texture_page_ptr[3] = texel_block >> 12;
355 texture_page_ptr += 4;
369 vram_ptr -= (1024 * 16) - 4;
375 vram_ptr += (16 * 1024) - (4 * 16);
380 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
383 u16 *texture_page_ptr = psx_gpu->texture_page_base;
384 u16 *vram_ptr = psx_gpu->vram_ptr;
391 texture_cache_loads++;
393 vram_ptr += (texture_page >> 4) * 256 * 1024;
394 vram_ptr += (texture_page & 0xF) * 64;
396 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
397 texture_page_ptr += (8 * 16) * 8;
410 load_128b(texels, vram_ptr);
411 store_128b(texels, texture_page_ptr);
413 texture_page_ptr += 8;
421 vram_ptr -= (1024 * 16);
430 vram_ptr += (16 * 1024);
432 texture_page_ptr += (8 * 16) * 8;
440 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
442 u32 current_texture_page = psx_gpu->current_texture_page;
443 u32 update_textures =
444 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
446 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
448 if(update_textures & (1 << current_texture_page))
450 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
451 update_textures &= ~(1 << current_texture_page);
456 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
457 (current_texture_page & 0x10);
459 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
463 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
465 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
466 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
468 u32 num_blocks_dest = 0;
469 block_struct *block_src = psx_gpu->blocks;
470 block_struct *block_dest = psx_gpu->blocks;
472 u16 *vram_ptr = psx_gpu->vram_ptr;
475 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
477 for(i = 0; i < psx_gpu->num_blocks; i++)
479 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
480 if(fb_offset & (1 << 11))
482 *block_dest = *block_src;
491 for(i = 0; i < psx_gpu->num_blocks; i++)
493 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
494 if((fb_offset & (1 << 11)) == 0)
496 *block_dest = *block_src;
504 psx_gpu->num_blocks = num_blocks_dest;
507 if(psx_gpu->num_blocks)
509 render_block_handler_struct *render_block_handler =
510 psx_gpu->render_block_handler;
512 render_block_handler->texture_blocks(psx_gpu);
513 render_block_handler->shade_blocks(psx_gpu);
514 render_block_handler->blend_blocks(psx_gpu);
517 span_pixel_blocks += psx_gpu->num_blocks;
518 render_buffer_flushes++;
521 psx_gpu->num_blocks = 0;
528 #define setup_gradient_calculation_input(set, vertex) \
529 /* First type is: uvrg bxxx xxxx */\
530 /* Second type is: yyyy ybyy uvrg */\
531 /* Since x_a and y_c are the same the same variable is used for both. */\
532 x##set##_a_y##set##_c.e[0] = vertex->u; \
533 x##set##_a_y##set##_c.e[1] = vertex->v; \
534 x##set##_a_y##set##_c.e[2] = vertex->r; \
535 x##set##_a_y##set##_c.e[3] = vertex->g; \
536 dup_4x16b(x##set##_b, vertex->x); \
537 dup_4x16b(x##set##_c, vertex->x); \
538 dup_4x16b(y##set##_a, vertex->y); \
539 dup_4x16b(y##set##_b, vertex->y); \
540 x##set##_b.e[0] = vertex->b; \
541 y##set##_b.e[1] = vertex->b \
544 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
545 vertex_struct *b, vertex_struct *c)
547 u32 triangle_area = psx_gpu->triangle_area;
548 u32 winding_mask_scalar;
550 u32 triangle_area_shift;
551 u64 triangle_area_reciprocal =
552 fixed_reciprocal(triangle_area, &triangle_area_shift);
553 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
555 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
556 // ( d0 * d1 ) - ( d2 * d3 ) =
557 // ( m0 ) - ( m1 ) = gradient
559 // This is split to do 12 elements at a time over three sets: a, b, and c.
560 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
561 // two of the slots are unused.
563 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
566 vec_4x16s x0_a_y0_c, x0_b, x0_c;
567 vec_4x16s y0_a, y0_b;
568 vec_4x16s x1_a_y1_c, x1_b, x1_c;
569 vec_4x16s y1_a, y1_b;
570 vec_4x16s x2_a_y2_c, x2_b, x2_c;
571 vec_4x16s y2_a, y2_b;
575 vec_4x32u uvrgb_phase;
577 vec_4x16s d0_a_d3_c, d0_b, d0_c;
578 vec_4x16s d1_a, d1_b, d1_c_d2_a;
579 vec_4x16s d2_b, d2_c;
580 vec_4x16s d3_a, d3_b;
582 vec_4x32s m0_a, m0_b, m0_c;
583 vec_4x32s m1_a, m1_b, m1_c;
585 vec_4x32u gradient_area_a, gradient_area_c;
586 vec_2x32u gradient_area_b;
588 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
589 vec_2x32u gradient_area_sign_b;
590 vec_4x32u winding_mask;
592 vec_2x64u gradient_wide_a0, gradient_wide_a1;
593 vec_2x64u gradient_wide_c0, gradient_wide_c1;
594 vec_2x64u gradient_wide_b;
596 vec_4x32u gradient_a, gradient_c;
597 vec_2x32u gradient_b;
598 vec_16x8s gradient_shift;
600 setup_gradient_calculation_input(0, a);
601 setup_gradient_calculation_input(1, b);
602 setup_gradient_calculation_input(2, c);
604 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
605 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
606 shl_long_4x16b(b_base, x0_b, 16);
608 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
609 add_4x32b(b_base, b_base, uvrgb_phase);
611 // Can probably pair these, but it'll require careful register allocation
612 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
613 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
615 sub_4x16b(d0_b, x1_b, x0_b);
616 sub_4x16b(d0_c, x1_c, x0_c);
618 sub_4x16b(d1_a, y2_a, y1_a);
619 sub_4x16b(d1_b, y2_b, y1_b);
621 sub_4x16b(d2_b, x2_b, x1_b);
622 sub_4x16b(d2_c, x2_c, x1_c);
624 sub_4x16b(d3_a, y1_a, y0_a);
625 sub_4x16b(d3_b, y1_b, y0_b);
627 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
628 mul_long_4x16b(m0_b, d0_b, d1_b);
629 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
631 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
632 mul_long_4x16b(m1_b, d2_b, d3_b);
633 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
635 sub_4x32b(gradient_area_a, m0_a, m1_a);
636 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
637 sub_4x32b(gradient_area_c, m0_c, m1_c);
639 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
640 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
641 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
643 abs_4x32b(gradient_area_a, gradient_area_a);
644 abs_2x32b(gradient_area_b, gradient_area_b);
645 abs_4x32b(gradient_area_c, gradient_area_c);
647 winding_mask_scalar = -psx_gpu->triangle_winding;
649 dup_4x32b(winding_mask, winding_mask_scalar);
650 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
651 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
652 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
654 mul_scalar_long_2x32b(gradient_wide_a0,
655 vector_cast(vec_2x32s, gradient_area_a.low),
656 (s64)triangle_area_reciprocal);
657 mul_scalar_long_2x32b(gradient_wide_a1,
658 vector_cast(vec_2x32s, gradient_area_a.high),
659 (s64)triangle_area_reciprocal);
660 mul_scalar_long_2x32b(gradient_wide_b,
661 vector_cast(vec_2x32s, gradient_area_b),
662 (s64)triangle_area_reciprocal);
663 mul_scalar_long_2x32b(gradient_wide_c0,
664 vector_cast(vec_2x32s, gradient_area_c.low),
665 (s64)triangle_area_reciprocal);
666 mul_scalar_long_2x32b(gradient_wide_c1,
667 vector_cast(vec_2x32s, gradient_area_c.high),
668 (s64)triangle_area_reciprocal);
670 dup_16x8b(gradient_shift, triangle_area_shift);
671 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
672 vector_cast(vec_2x64u, gradient_shift));
673 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
674 vector_cast(vec_2x64u, gradient_shift));
675 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
676 vector_cast(vec_2x64u, gradient_shift));
677 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
678 vector_cast(vec_2x64u, gradient_shift));
679 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
680 vector_cast(vec_2x64u, gradient_shift));
682 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
683 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
684 mov_narrow_2x64b(gradient_b, gradient_wide_b);
685 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
686 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
688 shl_4x32b(gradient_a, gradient_a, 4);
689 shl_2x32b(gradient_b, gradient_b, 4);
690 shl_4x32b(gradient_c, gradient_c, 4);
692 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
693 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
694 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
696 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
697 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
698 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
700 u32 left_adjust = a->x;
701 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
702 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
712 eor_4x32b(zero, zero, zero);
713 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
714 add_2x32b(b_dx2, gradient_b, gradient_b);
715 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
716 add_2x32b(b_dx3, gradient_b, b_dx2);
718 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
720 psx_gpu->u_block_span.e[0] = zero.e[0];
721 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
722 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
723 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
725 psx_gpu->v_block_span.e[0] = zero.e[1];
726 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
727 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
728 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
730 psx_gpu->r_block_span.e[0] = zero.e[2];
731 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
732 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
733 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
735 psx_gpu->g_block_span.e[0] = zero.e[3];
736 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
737 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
738 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
740 psx_gpu->b_block_span.e[0] = zero.e[0];
741 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
742 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
743 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
745 psx_gpu->uvrg = uvrg_base;
746 psx_gpu->b = b_base.e[0];
748 psx_gpu->uvrg_dx = gradient_a;
749 psx_gpu->uvrg_dy = gradient_c;
750 psx_gpu->b_dy = gradient_b.e[1];
754 #define vector_check(_a, _b) \
755 if(memcmp(&_a, &_b, sizeof(_b))) \
757 if(sizeof(_b) == 8) \
759 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
760 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
764 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
765 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
770 #define scalar_check(_a, _b) \
772 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
776 #define setup_spans_debug_check(span_edge_data_element) \
778 u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data; \
779 if (_num_spans > MAX_SPANS) \
780 *(volatile int *)0 = 1; \
781 if (_num_spans < psx_gpu->num_spans) \
783 if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW) \
784 *(volatile int *)0 = 2; \
785 if(span_edge_data_element.y >= 2048) \
786 *(volatile int *)0 = 3; \
791 #define setup_spans_debug_check(span_edge_data_element) \
795 #define setup_spans_prologue_alternate_yes() \
796 vec_2x64s alternate_x; \
797 vec_2x64s alternate_dx_dy; \
798 vec_4x32s alternate_x_32; \
799 vec_4x16u alternate_x_16; \
801 vec_4x16u alternate_select; \
802 vec_4x16s y_mid_point; \
806 s32 edge_dx_dy_alt; \
809 #define setup_spans_prologue_alternate_no() \
811 #define setup_spans_prologue(alternate_active) \
812 edge_data_struct *span_edge_data; \
813 vec_4x32u *span_uvrg_offset; \
814 u32 *span_b_offset; \
818 vec_2x64s edges_xy; \
819 vec_2x32s edges_dx_dy; \
820 vec_2x32u edge_shifts; \
822 vec_2x64s left_x, right_x; \
823 vec_2x64s left_dx_dy, right_dx_dy; \
824 vec_4x32s left_x_32, right_x_32; \
825 vec_8x16s left_right_x_16; \
827 vec_8x16s left_edge; \
828 vec_8x16s right_edge; \
829 vec_4x16u span_shift; \
833 vec_4x16u c_0xFFFE; \
836 vec_2x32s x_starts; \
845 vec_4x32u uvrg = psx_gpu->uvrg; \
846 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
847 u32 b = psx_gpu->b; \
848 u32 b_dy = psx_gpu->b_dy; \
850 dup_2x32b(c_0x01, 0x01); \
851 setup_spans_prologue_alternate_##alternate_active() \
853 #define setup_spans_prologue_b() \
854 span_edge_data = psx_gpu->span_edge_data; \
855 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
856 span_b_offset = psx_gpu->span_b_offset; \
858 vec_8x16u c_0x0001; \
860 dup_8x16b(c_0x0001, 0x0001); \
861 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
862 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
863 add_8x16b(right_edge, right_edge, c_0x0001); \
864 dup_4x16b(c_0x04, 0x04); \
865 dup_4x16b(c_0x07, 0x07); \
866 dup_4x16b(c_0xFFFE, 0xFFFE); \
869 #define compute_edge_delta_x2() \
872 vec_2x32s height_reciprocals; \
873 vec_2x32s heights_b; \
876 u32 edge_shift = reciprocal_table[height]; \
878 dup_2x32b(heights, height); \
879 sub_2x32b(widths, x_ends, x_starts); \
881 dup_2x32b(edge_shifts, edge_shift); \
882 sub_2x32b(heights_b, heights, c_0x01); \
883 shr_2x32b(height_reciprocals, edge_shifts, 10); \
885 mla_2x32b(heights_b, x_starts, heights); \
886 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
887 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
888 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
891 #define compute_edge_delta_x3(start_c, height_a, height_b) \
894 vec_2x32s height_reciprocals; \
895 vec_2x32s heights_b; \
900 u32 height_reciprocal_alt; \
902 heights.e[0] = height_a; \
903 heights.e[1] = height_b; \
905 edge_shifts.e[0] = reciprocal_table[height_a]; \
906 edge_shifts.e[1] = reciprocal_table[height_b]; \
907 edge_shift_alt = reciprocal_table[height_minor_b]; \
909 sub_2x32b(widths, x_ends, x_starts); \
910 width_alt = x_c - start_c; \
912 shr_2x32b(height_reciprocals, edge_shifts, 10); \
913 height_reciprocal_alt = edge_shift_alt >> 10; \
915 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
916 edge_shift_alt &= 0x1F; \
918 sub_2x32b(heights_b, heights, c_0x01); \
919 height_b_alt = height_minor_b - 1; \
921 mla_2x32b(heights_b, x_starts, heights); \
922 height_b_alt += height_minor_b * start_c; \
924 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
925 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
927 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
928 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
932 #define setup_spans_adjust_y_up() \
933 sub_4x32b(y_x4, y_x4, c_0x04) \
935 #define setup_spans_adjust_y_down() \
936 add_4x32b(y_x4, y_x4, c_0x04) \
938 #define setup_spans_adjust_interpolants_up() \
939 sub_4x32b(uvrg, uvrg, uvrg_dy); \
942 #define setup_spans_adjust_interpolants_down() \
943 add_4x32b(uvrg, uvrg, uvrg_dy); \
947 #define setup_spans_clip_interpolants_increment() \
948 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
951 #define setup_spans_clip_interpolants_decrement() \
952 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
955 #define setup_spans_clip_alternate_yes() \
956 edge_alt += edge_dx_dy_alt * (s64)(clip) \
958 #define setup_spans_clip_alternate_no() \
960 #define setup_spans_clip(direction, alternate_active) \
962 clipped_triangles++; \
963 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
964 setup_spans_clip_alternate_##alternate_active(); \
965 setup_spans_clip_interpolants_##direction(); \
969 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
971 vec_2x64u edge_shifts_64; \
972 vec_2x64s edges_dx_dy_64; \
974 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
975 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
977 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
978 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
980 left_x.e[0] = edges_xy.e[left_index]; \
981 right_x.e[0] = edges_xy.e[right_index]; \
983 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
984 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
985 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
986 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
988 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
989 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
991 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
992 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
995 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
997 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
998 s64 edge_dx_dy_alt_64; \
1000 dup_4x16b(y_mid_point, y_b); \
1002 edge_alt <<= edge_shift_alt; \
1003 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
1005 alternate_x.e[0] = edge_alt; \
1006 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
1007 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1009 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1010 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1014 #define setup_spans_y_select_up() \
1015 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1017 #define setup_spans_y_select_down() \
1018 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1020 #define setup_spans_y_select_alternate_yes(direction) \
1021 setup_spans_y_select_##direction() \
1023 #define setup_spans_y_select_alternate_no(direction) \
1025 #define setup_spans_alternate_select_left() \
1026 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1028 #define setup_spans_alternate_select_right() \
1029 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1031 #define setup_spans_alternate_select_none() \
1033 #define setup_spans_increment_alternate_yes() \
1034 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1035 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1036 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1037 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1038 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1040 #define setup_spans_increment_alternate_no() \
1042 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1044 span_uvrg_offset[0] = uvrg; \
1045 span_b_offset[0] = b; \
1046 setup_spans_adjust_interpolants_##direction(); \
1048 span_uvrg_offset[1] = uvrg; \
1049 span_b_offset[1] = b; \
1050 setup_spans_adjust_interpolants_##direction(); \
1052 span_uvrg_offset[2] = uvrg; \
1053 span_b_offset[2] = b; \
1054 setup_spans_adjust_interpolants_##direction(); \
1056 span_uvrg_offset[3] = uvrg; \
1057 span_b_offset[3] = b; \
1058 setup_spans_adjust_interpolants_##direction(); \
1060 span_uvrg_offset += 4; \
1061 span_b_offset += 4; \
1063 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1064 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1066 add_2x64b(left_x, left_x, left_dx_dy); \
1067 add_2x64b(right_x, right_x, right_dx_dy); \
1069 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1070 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1072 add_2x64b(left_x, left_x, left_dx_dy); \
1073 add_2x64b(right_x, right_x, right_dx_dy); \
1075 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1076 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1078 setup_spans_increment_alternate_##alternate_active(); \
1079 setup_spans_y_select_alternate_##alternate_active(direction); \
1080 setup_spans_alternate_select_##alternate(); \
1082 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1083 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1085 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1086 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1087 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1088 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1089 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1092 for(i = 0; i < 4; i++) \
1094 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1095 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1096 span_edge_data[i].right_mask = span_shift.e[i]; \
1097 span_edge_data[i].y = y_x4.e[i]; \
1098 setup_spans_debug_check(span_edge_data[i]); \
1101 span_edge_data += 4; \
1103 setup_spans_adjust_y_##direction(); \
1107 #define setup_spans_alternate_adjust_yes() \
1108 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1110 #define setup_spans_alternate_adjust_no() \
1113 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1114 setup_spans_alternate_adjust_##alternate_active(); \
1115 if(y_c > psx_gpu->viewport_end_y) \
1116 height -= y_c - psx_gpu->viewport_end_y - 1; \
1118 clip = psx_gpu->viewport_start_y - y_a; \
1123 setup_spans_clip(increment, alternate_active); \
1126 setup_spans_prologue_b(); \
1131 y_x4.e[1] = y_a + 1; \
1132 y_x4.e[2] = y_a + 2; \
1133 y_x4.e[3] = y_a + 3; \
1134 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1137 psx_gpu->num_spans = height; \
1140 setup_spans_set_x4(alternate, down, alternate_active); \
1142 } while(height > 0); \
1146 #define setup_spans_alternate_pre_increment_yes() \
1147 edge_alt += edge_dx_dy_alt \
1149 #define setup_spans_alternate_pre_increment_no() \
1151 #define setup_spans_up_decrement_height_yes() \
1154 #define setup_spans_up_decrement_height_no() \
1157 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1158 setup_spans_alternate_adjust_##alternate_active(); \
1161 if(y_c < psx_gpu->viewport_start_y) \
1162 height -= psx_gpu->viewport_start_y - y_c; \
1164 setup_spans_up_decrement_height_##alternate_active(); \
1166 clip = y_a - psx_gpu->viewport_end_y; \
1171 setup_spans_clip(decrement, alternate_active); \
1174 setup_spans_prologue_b(); \
1179 y_x4.e[1] = y_a - 1; \
1180 y_x4.e[2] = y_a - 2; \
1181 y_x4.e[3] = y_a - 3; \
1182 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1183 setup_spans_alternate_pre_increment_##alternate_active(); \
1184 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1186 setup_spans_adjust_interpolants_up(); \
1188 psx_gpu->num_spans = height; \
1191 setup_spans_set_x4(alternate, up, alternate_active); \
1196 #define index_left 0
1197 #define index_right 1
1199 #define setup_spans_up_up(minor, major) \
1200 setup_spans_prologue(yes); \
1201 s32 height_minor_a = y_a - y_b; \
1202 s32 height_minor_b = y_b - y_c; \
1203 s32 height = y_a - y_c; \
1205 dup_2x32b(x_starts, x_a); \
1206 x_ends.e[0] = x_c; \
1207 x_ends.e[1] = x_b; \
1209 compute_edge_delta_x3(x_b, height, height_minor_a); \
1210 setup_spans_up(index_##major, index_##minor, minor, yes) \
1215 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1216 vertex_struct *v_b, vertex_struct *v_c)
1218 setup_spans_up_up(left, right);
1221 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1222 vertex_struct *v_b, vertex_struct *v_c)
1224 setup_spans_up_up(right, left);
1227 #define setup_spans_down_down(minor, major) \
1228 setup_spans_prologue(yes); \
1229 s32 height_minor_a = y_b - y_a; \
1230 s32 height_minor_b = y_c - y_b; \
1231 s32 height = y_c - y_a; \
1233 dup_2x32b(x_starts, x_a); \
1234 x_ends.e[0] = x_c; \
1235 x_ends.e[1] = x_b; \
1237 compute_edge_delta_x3(x_b, height, height_minor_a); \
1238 setup_spans_down(index_##major, index_##minor, minor, yes) \
1240 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1241 vertex_struct *v_b, vertex_struct *v_c)
1243 setup_spans_down_down(left, right);
1246 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1247 vertex_struct *v_b, vertex_struct *v_c)
1249 setup_spans_down_down(right, left);
1252 #define setup_spans_up_flat() \
1253 s32 height = y_a - y_c; \
1256 compute_edge_delta_x2(); \
1257 setup_spans_up(index_left, index_right, none, no) \
1259 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1260 vertex_struct *v_b, vertex_struct *v_c)
1262 setup_spans_prologue(no);
1263 x_starts.e[0] = x_a;
1264 x_starts.e[1] = x_b;
1265 dup_2x32b(x_ends, x_c);
1267 setup_spans_up_flat();
1270 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1271 vertex_struct *v_b, vertex_struct *v_c)
1273 setup_spans_prologue(no);
1274 dup_2x32b(x_starts, x_a);
1278 setup_spans_up_flat();
1281 #define setup_spans_down_flat() \
1282 s32 height = y_c - y_a; \
1285 compute_edge_delta_x2(); \
1286 setup_spans_down(index_left, index_right, none, no) \
1288 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1289 vertex_struct *v_b, vertex_struct *v_c)
1291 setup_spans_prologue(no);
1292 x_starts.e[0] = x_a;
1293 x_starts.e[1] = x_b;
1294 dup_2x32b(x_ends, x_c);
1296 setup_spans_down_flat();
1299 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1300 vertex_struct *v_b, vertex_struct *v_c)
1302 setup_spans_prologue(no);
1303 dup_2x32b(x_starts, x_a);
1307 setup_spans_down_flat();
1310 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1311 vertex_struct *v_b, vertex_struct *v_c)
1313 setup_spans_prologue(no);
1321 s32 height_minor_a = y_a - y_b;
1322 s32 height_minor_b = y_c - y_a;
1323 s32 height_major = y_c - y_b;
1325 vec_2x64s edges_xy_b;
1326 vec_2x32s edges_dx_dy_b;
1327 vec_2x32u edge_shifts_b;
1329 vec_2x32s height_increment;
1331 x_starts.e[0] = x_a;
1332 x_starts.e[1] = x_c;
1333 dup_2x32b(x_ends, x_b);
1335 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1337 height_increment.e[0] = 0;
1338 height_increment.e[1] = height_minor_b;
1340 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1342 edges_xy_b.e[0] = edge_alt;
1343 edges_xy_b.e[1] = edges_xy.e[1];
1345 edge_shifts_b = edge_shifts;
1346 edge_shifts_b.e[0] = edge_shift_alt;
1348 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1349 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1353 if(y_b < psx_gpu->viewport_start_y)
1354 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1356 clip = y_a - psx_gpu->viewport_end_y;
1359 height_minor_a -= clip;
1361 setup_spans_clip(decrement, no);
1364 setup_spans_prologue_b();
1366 if(height_minor_a > 0)
1369 y_x4.e[1] = y_a - 1;
1370 y_x4.e[2] = y_a - 2;
1371 y_x4.e[3] = y_a - 3;
1372 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1373 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1374 setup_spans_adjust_interpolants_up();
1376 psx_gpu->num_spans = height_minor_a;
1377 while(height_minor_a > 0)
1379 setup_spans_set_x4(none, up, no);
1380 height_minor_a -= 4;
1383 span_edge_data += height_minor_a;
1384 span_uvrg_offset += height_minor_a;
1385 span_b_offset += height_minor_a;
1388 edges_xy = edges_xy_b;
1389 edges_dx_dy = edges_dx_dy_b;
1390 edge_shifts = edge_shifts_b;
1392 uvrg = psx_gpu->uvrg;
1397 if(y_c > psx_gpu->viewport_end_y)
1398 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1400 clip = psx_gpu->viewport_start_y - y_a;
1403 height_minor_b -= clip;
1405 setup_spans_clip(increment, no);
1408 if(height_minor_b > 0)
1411 y_x4.e[1] = y_a + 1;
1412 y_x4.e[2] = y_a + 2;
1413 y_x4.e[3] = y_a + 3;
1414 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1416 // FIXME: overflow corner case
1417 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1418 height_minor_b &= ~3;
1420 psx_gpu->num_spans += height_minor_b;
1421 while(height_minor_b > 0)
1423 setup_spans_set_x4(none, down, no);
1424 height_minor_b -= 4;
1428 left_split_triangles++;
1434 #define dither_table_entry_normal(value) \
1438 #define setup_blocks_load_msb_mask_indirect() \
1440 #define setup_blocks_load_msb_mask_direct() \
1441 vec_8x16u msb_mask; \
1442 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1445 #define setup_blocks_variables_shaded_textured(target) \
1446 vec_4x32u u_block; \
1447 vec_4x32u v_block; \
1448 vec_4x32u r_block; \
1449 vec_4x32u g_block; \
1450 vec_4x32u b_block; \
1451 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1452 vec_4x32u uvrg_dx4; \
1453 vec_4x32u uvrg_dx8; \
1455 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1456 u32 b_dx4 = b_dx << 2; \
1457 u32 b_dx8 = b_dx << 3; \
1460 vec_16x8u texture_mask; \
1461 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1462 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1463 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1464 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1466 #define setup_blocks_variables_shaded_untextured(target) \
1467 vec_4x32u r_block; \
1468 vec_4x32u g_block; \
1469 vec_4x32u b_block; \
1471 vec_4x32u rgb_dx4; \
1472 vec_4x32u rgb_dx8; \
1475 vec_8x8u d64_0x07; \
1480 dup_8x8b(d64_0x07, 0x07); \
1481 dup_8x8b(d64_1, 1); \
1482 dup_8x8b(d64_4, 4); \
1483 dup_8x8b(d64_128, 128); \
1485 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1486 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1487 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1488 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1490 #define setup_blocks_variables_unshaded_textured(target) \
1491 vec_4x32u u_block; \
1492 vec_4x32u v_block; \
1493 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1496 vec_2x32u uv = psx_gpu->uvrg.low; \
1498 vec_16x8u texture_mask; \
1499 shl_2x32b(uv_dx4, uv_dx, 2); \
1500 shl_2x32b(uv_dx8, uv_dx, 3); \
1501 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1502 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1505 #define setup_blocks_variables_unshaded_untextured_direct() \
1506 or_8x16b(colors, colors, msb_mask) \
1508 #define setup_blocks_variables_unshaded_untextured_indirect() \
1510 #define setup_blocks_variables_unshaded_untextured(target) \
1511 u32 color = psx_gpu->triangle_color; \
1514 u32 color_r = color & 0xFF; \
1515 u32 color_g = (color >> 8) & 0xFF; \
1516 u32 color_b = (color >> 16) & 0xFF; \
1518 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1519 ((color_b >> 3) << 10); \
1520 dup_8x16b(colors, color); \
1521 setup_blocks_variables_unshaded_untextured_##target() \
1523 #define setup_blocks_span_initialize_dithered_textured() \
1524 vec_8x16u dither_offsets; \
1525 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1527 #define setup_blocks_span_initialize_dithered_untextured() \
1528 vec_8x8u dither_offsets; \
1529 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1531 #define setup_blocks_span_initialize_dithered(texturing) \
1532 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1533 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1534 vec_8x8s dither_offsets_short; \
1537 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1538 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1539 setup_blocks_span_initialize_dithered_##texturing() \
1541 #define setup_blocks_span_initialize_undithered(texturing) \
1544 #define setup_blocks_span_initialize_shaded_textured() \
1546 vec_4x32u block_span; \
1547 u32 offset = span_edge_data->left_x; \
1549 uvrg = *span_uvrg_offset; \
1550 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1551 b = *span_b_offset; \
1552 b += b_dx * offset; \
1554 dup_4x32b(u_block, uvrg.e[0]); \
1555 dup_4x32b(v_block, uvrg.e[1]); \
1556 dup_4x32b(r_block, uvrg.e[2]); \
1557 dup_4x32b(g_block, uvrg.e[3]); \
1558 dup_4x32b(b_block, b); \
1560 block_span = psx_gpu->u_block_span; \
1561 add_4x32b(u_block, u_block, block_span); \
1562 block_span = psx_gpu->v_block_span; \
1563 add_4x32b(v_block, v_block, block_span); \
1564 block_span = psx_gpu->r_block_span; \
1565 add_4x32b(r_block, r_block, block_span); \
1566 block_span = psx_gpu->g_block_span; \
1567 add_4x32b(g_block, g_block, block_span); \
1568 block_span = psx_gpu->b_block_span; \
1569 add_4x32b(b_block, b_block, block_span); \
1572 #define setup_blocks_span_initialize_shaded_untextured() \
1574 vec_4x32u block_span; \
1575 u32 offset = span_edge_data->left_x; \
1577 rgb.low = span_uvrg_offset->high; \
1578 rgb.high.e[0] = *span_b_offset; \
1579 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1581 dup_4x32b(r_block, rgb.e[0]); \
1582 dup_4x32b(g_block, rgb.e[1]); \
1583 dup_4x32b(b_block, rgb.e[2]); \
1585 block_span = psx_gpu->r_block_span; \
1586 add_4x32b(r_block, r_block, block_span); \
1587 block_span = psx_gpu->g_block_span; \
1588 add_4x32b(g_block, g_block, block_span); \
1589 block_span = psx_gpu->b_block_span; \
1590 add_4x32b(b_block, b_block, block_span); \
1593 #define setup_blocks_span_initialize_unshaded_textured() \
1595 vec_4x32u block_span; \
1596 u32 offset = span_edge_data->left_x; \
1598 uv = span_uvrg_offset->low; \
1599 mla_scalar_2x32b(uv, uv_dx, offset); \
1601 dup_4x32b(u_block, uv.e[0]); \
1602 dup_4x32b(v_block, uv.e[1]); \
1604 block_span = psx_gpu->u_block_span; \
1605 add_4x32b(u_block, u_block, block_span); \
1606 block_span = psx_gpu->v_block_span; \
1607 add_4x32b(v_block, v_block, block_span); \
1610 #define setup_blocks_span_initialize_unshaded_untextured() \
1613 #define setup_blocks_texture_swizzled() \
1615 vec_8x8u u_saved = u; \
1616 sli_8x8b(u, v, 4); \
1617 sri_8x8b(v, u_saved, 4); \
1620 #define setup_blocks_texture_unswizzled() \
1622 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1625 vec_8x16u u_whole; \
1626 vec_8x16u v_whole; \
1627 vec_8x16u r_whole; \
1628 vec_8x16u g_whole; \
1629 vec_8x16u b_whole; \
1641 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1642 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1643 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1644 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1645 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1647 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1648 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1649 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1650 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1651 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1652 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1653 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1654 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1655 dup_4x32b(dx4, b_dx4); \
1656 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1658 mov_narrow_8x16b(u, u_whole); \
1659 mov_narrow_8x16b(v, v_whole); \
1660 mov_narrow_8x16b(r, r_whole); \
1661 mov_narrow_8x16b(g, g_whole); \
1662 mov_narrow_8x16b(b, b_whole); \
1664 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1665 add_4x32b(u_block, u_block, dx8); \
1666 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1667 add_4x32b(v_block, v_block, dx8); \
1668 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1669 add_4x32b(r_block, r_block, dx8); \
1670 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1671 add_4x32b(g_block, g_block, dx8); \
1672 dup_4x32b(dx8, b_dx8); \
1673 add_4x32b(b_block, b_block, dx8); \
1675 and_8x8b(u, u, texture_mask.low); \
1676 and_8x8b(v, v, texture_mask.high); \
1677 setup_blocks_texture_##swizzling(); \
1679 zip_8x16b(uv, u, v); \
1684 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1685 block->fb_ptr = fb_ptr; \
1688 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1691 vec_8x16u u_whole; \
1692 vec_8x16u v_whole; \
1701 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1702 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1704 dup_4x32b(dx4, uv_dx4.e[0]); \
1705 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1706 dup_4x32b(dx4, uv_dx4.e[1]); \
1707 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1709 mov_narrow_8x16b(u, u_whole); \
1710 mov_narrow_8x16b(v, v_whole); \
1712 dup_4x32b(dx8, uv_dx8.e[0]); \
1713 add_4x32b(u_block, u_block, dx8); \
1714 dup_4x32b(dx8, uv_dx8.e[1]); \
1715 add_4x32b(v_block, v_block, dx8); \
1717 and_8x8b(u, u, texture_mask.low); \
1718 and_8x8b(v, v, texture_mask.high); \
1719 setup_blocks_texture_##swizzling(); \
1721 zip_8x16b(uv, u, v); \
1723 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1724 block->fb_ptr = fb_ptr; \
1727 #define setup_blocks_store_shaded_untextured_dithered() \
1728 addq_8x8b(r, r, dither_offsets); \
1729 addq_8x8b(g, g, dither_offsets); \
1730 addq_8x8b(b, b, dither_offsets); \
1732 subq_8x8b(r, r, d64_4); \
1733 subq_8x8b(g, g, d64_4); \
1734 subq_8x8b(b, b, d64_4) \
1736 #define setup_blocks_store_shaded_untextured_undithered() \
1739 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1740 block->pixels = _pixels; \
1741 block->fb_ptr = fb_ptr \
1743 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1744 block->pixels = _pixels; \
1745 block->fb_ptr = fb_ptr \
1747 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1748 mul_long_8x8b(pixels, r, d64_1) \
1751 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1752 store_8x16b(_pixels, fb_ptr) \
1754 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1756 vec_8x16u fb_pixels; \
1757 vec_8x16u draw_mask; \
1758 vec_8x16u test_mask = psx_gpu->test_mask; \
1760 load_8x16b(fb_pixels, fb_ptr); \
1761 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1762 tst_8x16b(draw_mask, draw_mask, test_mask); \
1763 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1764 store_8x16b(fb_pixels, fb_ptr); \
1767 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1768 pixels = msb_mask; \
1769 mla_long_8x8b(pixels, r, d64_1) \
1772 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1775 vec_8x16u r_whole; \
1776 vec_8x16u g_whole; \
1777 vec_8x16u b_whole; \
1788 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1789 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1790 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1792 dup_4x32b(dx4, rgb_dx4.e[0]); \
1793 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1794 dup_4x32b(dx4, rgb_dx4.e[1]); \
1795 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1796 dup_4x32b(dx4, rgb_dx4.e[2]); \
1797 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1799 mov_narrow_8x16b(r, r_whole); \
1800 mov_narrow_8x16b(g, g_whole); \
1801 mov_narrow_8x16b(b, b_whole); \
1803 dup_4x32b(dx8, rgb_dx8.e[0]); \
1804 add_4x32b(r_block, r_block, dx8); \
1805 dup_4x32b(dx8, rgb_dx8.e[1]); \
1806 add_4x32b(g_block, g_block, dx8); \
1807 dup_4x32b(dx8, rgb_dx8.e[2]); \
1808 add_4x32b(b_block, b_block, dx8); \
1810 setup_blocks_store_shaded_untextured_##dithering(); \
1812 shr_8x8b(r, r, 3); \
1813 bic_8x8b(g, g, d64_0x07); \
1814 bic_8x8b(b, b, d64_0x07); \
1816 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1817 mla_long_8x8b(pixels, g, d64_4); \
1818 mla_long_8x8b(pixels, b, d64_128) \
1820 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1823 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1825 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1828 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1829 (_block)->draw_mask_bits = bits \
1831 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1833 vec_8x16u bits_mask; \
1834 vec_8x16u test_mask = psx_gpu->test_mask; \
1835 dup_8x16b(bits_mask, bits); \
1836 tst_8x16b(bits_mask, bits_mask, test_mask); \
1837 (_block)->draw_mask = bits_mask; \
1840 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1843 #define setup_blocks_add_blocks_indirect() \
1844 num_blocks += span_num_blocks; \
1846 if(num_blocks > MAX_BLOCKS) \
1848 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1849 flush_render_block_buffer(psx_gpu); \
1850 num_blocks = span_num_blocks; \
1851 block = psx_gpu->blocks; \
1854 #define setup_blocks_add_blocks_direct() \
1855 stats_add(texel_blocks_untextured, span_num_blocks); \
1856 span_pixel_blocks += span_num_blocks \
1859 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1860 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1861 psx_gpu_struct *psx_gpu) \
1863 setup_blocks_load_msb_mask_##target(); \
1864 setup_blocks_variables_##shading##_##texturing(target); \
1866 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1867 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1868 u32 *span_b_offset = psx_gpu->span_b_offset; \
1870 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1872 u32 num_spans = psx_gpu->num_spans; \
1877 u32 num_blocks = psx_gpu->num_blocks; \
1878 u32 span_num_blocks; \
1882 span_num_blocks = span_edge_data->num_blocks; \
1883 if(span_num_blocks) \
1885 y = span_edge_data->y; \
1886 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1888 setup_blocks_span_initialize_##shading##_##texturing(); \
1889 setup_blocks_span_initialize_##dithering(texturing); \
1891 setup_blocks_add_blocks_##target(); \
1893 s32 pixel_span = span_num_blocks * 8; \
1894 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1895 span_pixels += pixel_span; \
1897 span_num_blocks--; \
1898 while(span_num_blocks) \
1900 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1902 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1906 span_num_blocks--; \
1909 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1910 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1911 span_edge_data->right_mask); \
1917 zero_block_spans++; \
1922 span_uvrg_offset++; \
1926 psx_gpu->num_blocks = num_blocks; \
1930 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1934 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1935 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1937 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1938 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1940 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1941 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1942 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1943 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1945 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1946 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1948 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1950 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1951 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1954 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1956 block_struct *block = psx_gpu->blocks;
1957 u32 num_blocks = psx_gpu->num_blocks;
1958 stats_add(texel_blocks_4bpp, num_blocks);
1960 vec_8x8u texels_low;
1961 vec_8x8u texels_high;
1968 vec_16x8u clut_high;
1970 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1971 u16 *clut_ptr = psx_gpu->clut_ptr;
1973 // Can be done with one deinterleaving load on NEON
1974 load_8x16b(clut_a, clut_ptr);
1975 load_8x16b(clut_b, clut_ptr + 8);
1976 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
1978 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
1979 update_texture_4bpp_cache(psx_gpu);
1983 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
1984 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
1985 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
1986 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
1987 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
1988 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
1989 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
1990 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
1992 tbl_16(texels_low, texels, clut_low);
1993 tbl_16(texels_high, texels, clut_high);
1995 // Can be done with an interleaving store on NEON
1996 zip_8x16b(pixels, texels_low, texels_high);
1998 block->texels = pixels;
2005 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2007 block_struct *block = psx_gpu->blocks;
2008 u32 num_blocks = psx_gpu->num_blocks;
2010 stats_add(texel_blocks_8bpp, num_blocks);
2012 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2013 update_texture_8bpp_cache(psx_gpu);
2016 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2024 for(i = 0; i < 8; i++)
2026 offset = block->uv.e[i];
2028 texel = texture_ptr_8bpp[offset];
2029 texels.e[i] = psx_gpu->clut_ptr[texel];
2032 block->texels = texels;
2039 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2041 block_struct *block = psx_gpu->blocks;
2042 u32 num_blocks = psx_gpu->num_blocks;
2044 stats_add(texel_blocks_16bpp, num_blocks);
2048 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2054 for(i = 0; i < 8; i++)
2056 offset = block->uv.e[i];
2057 offset += ((offset & 0xFF00) * 3);
2059 texels.e[i] = texture_ptr_16bpp[offset];
2062 block->texels = texels;
2072 #define shade_blocks_load_msb_mask_indirect() \
2074 #define shade_blocks_load_msb_mask_direct() \
2075 vec_8x16u msb_mask; \
2076 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2078 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2079 block->draw_mask = _draw_mask; \
2080 block->pixels = _pixels \
2082 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2084 vec_8x16u fb_pixels; \
2085 or_8x16b(_pixels, _pixels, msb_mask); \
2086 load_8x16b(fb_pixels, block->fb_ptr); \
2087 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2088 store_8x16b(fb_pixels, block->fb_ptr); \
2092 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2093 if(psx_gpu->triangle_color == 0x808080) \
2095 false_modulated_blocks += num_blocks; \
2098 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2099 if(psx_gpu->triangle_color == 0x808080) \
2102 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2103 false_modulated_blocks += num_blocks; \
2108 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2111 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2114 u32 color = psx_gpu->triangle_color; \
2115 dup_8x8b(colors_r, color); \
2116 dup_8x8b(colors_g, color >> 8); \
2117 dup_8x8b(colors_b, color >> 16); \
2118 shade_blocks_textured_false_modulated_check_##dithering(target); \
2121 #define shade_blocks_textured_modulated_shaded_block_load() \
2122 colors_r = block->r; \
2123 colors_g = block->g; \
2124 colors_b = block->b \
2126 #define shade_blocks_textured_modulated_unshaded_block_load() \
2128 #define shade_blocks_textured_modulate_dithered(component) \
2129 pixels_##component = block->dither_offsets; \
2130 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2132 #define shade_blocks_textured_modulate_undithered(component) \
2133 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2135 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2136 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2137 psx_gpu_struct *psx_gpu) \
2139 block_struct *block = psx_gpu->blocks; \
2140 u32 num_blocks = psx_gpu->num_blocks; \
2143 vec_8x8u texels_r; \
2144 vec_8x8u texels_g; \
2145 vec_8x8u texels_b; \
2147 vec_8x8u colors_r; \
2148 vec_8x8u colors_g; \
2149 vec_8x8u colors_b; \
2151 vec_8x8u pixels_r_low; \
2152 vec_8x8u pixels_g_low; \
2153 vec_8x8u pixels_b_low; \
2156 vec_8x16u pixels_r; \
2157 vec_8x16u pixels_g; \
2158 vec_8x16u pixels_b; \
2160 vec_8x16u draw_mask; \
2161 vec_8x16u zero_mask; \
2163 vec_8x8u d64_0x07; \
2164 vec_8x8u d64_0x1F; \
2169 vec_8x16u d128_0x8000; \
2171 vec_8x16u test_mask = psx_gpu->test_mask; \
2172 u32 draw_mask_bits; \
2173 shade_blocks_load_msb_mask_##target(); \
2175 dup_8x8b(d64_0x07, 0x07); \
2176 dup_8x8b(d64_0x1F, 0x1F); \
2177 dup_8x8b(d64_1, 1); \
2178 dup_8x8b(d64_4, 4); \
2179 dup_8x8b(d64_128, 128); \
2181 dup_8x16b(d128_0x8000, 0x8000); \
2183 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2188 draw_mask_bits = block->draw_mask_bits; \
2189 dup_8x16b(draw_mask, draw_mask_bits); \
2190 tst_8x16b(draw_mask, draw_mask, test_mask); \
2192 shade_blocks_textured_modulated_##shading##_block_load(); \
2194 texels = block->texels; \
2196 mov_narrow_8x16b(texels_r, texels); \
2197 shr_narrow_8x16b(texels_g, texels, 5); \
2198 shr_narrow_8x16b(texels_b, texels, 7); \
2200 and_8x8b(texels_r, texels_r, d64_0x1F); \
2201 and_8x8b(texels_g, texels_g, d64_0x1F); \
2202 shr_8x8b(texels_b, texels_b, 3); \
2204 shade_blocks_textured_modulate_##dithering(r); \
2205 shade_blocks_textured_modulate_##dithering(g); \
2206 shade_blocks_textured_modulate_##dithering(b); \
2208 cmpeqz_8x16b(zero_mask, texels); \
2209 and_8x16b(pixels, texels, d128_0x8000); \
2211 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2212 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2213 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2215 or_8x16b(zero_mask, draw_mask, zero_mask); \
2217 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2218 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2219 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2221 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2222 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2223 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2225 shade_blocks_store_##target(zero_mask, pixels); \
2234 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2235 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2236 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2237 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2239 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2240 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2241 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2242 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2247 #define shade_blocks_textured_unmodulated_builder(target) \
2248 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2250 block_struct *block = psx_gpu->blocks; \
2251 u32 num_blocks = psx_gpu->num_blocks; \
2252 vec_8x16u draw_mask; \
2253 vec_8x16u test_mask = psx_gpu->test_mask; \
2254 u32 draw_mask_bits; \
2257 shade_blocks_load_msb_mask_##target(); \
2261 vec_8x16u zero_mask; \
2263 draw_mask_bits = block->draw_mask_bits; \
2264 dup_8x16b(draw_mask, draw_mask_bits); \
2265 tst_8x16b(draw_mask, draw_mask, test_mask); \
2267 pixels = block->texels; \
2269 cmpeqz_8x16b(zero_mask, pixels); \
2270 or_8x16b(zero_mask, draw_mask, zero_mask); \
2272 shade_blocks_store_##target(zero_mask, pixels); \
2279 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2280 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2283 block_struct *block = psx_gpu->blocks; \
2284 u32 num_blocks = psx_gpu->num_blocks; \
2285 vec_8x16u draw_mask; \
2286 vec_8x16u test_mask = psx_gpu->test_mask; \
2287 u32 draw_mask_bits; \
2290 shade_blocks_load_msb_mask_##target(); \
2294 vec_8x16u zero_mask; \
2296 draw_mask_bits = block->draw_mask_bits; \
2297 dup_8x16b(draw_mask, draw_mask_bits); \
2298 tst_8x16b(draw_mask, draw_mask, test_mask); \
2300 pixels = block->texels; \
2302 cmpeqz_8x16b(zero_mask, pixels); \
2303 or_8x16b(zero_mask, draw_mask, zero_mask); \
2305 shade_blocks_store_##target(zero_mask, pixels); \
2314 shade_blocks_textured_unmodulated_builder(indirect)
2315 shade_blocks_textured_unmodulated_builder(direct)
2317 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2321 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2323 block_struct *block = psx_gpu->blocks;
2324 u32 num_blocks = psx_gpu->num_blocks;
2326 vec_8x16u pixels = block->pixels;
2327 shade_blocks_load_msb_mask_direct();
2331 shade_blocks_store_direct(block->draw_mask, pixels);
2340 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2345 #define blend_blocks_mask_evaluate_on() \
2346 vec_8x16u mask_pixels; \
2347 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2348 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2350 #define blend_blocks_mask_evaluate_off() \
2352 #define blend_blocks_average() \
2354 vec_8x16u pixels_no_msb; \
2355 vec_8x16u fb_pixels_no_msb; \
2357 vec_8x16u d128_0x0421; \
2358 vec_8x16u d128_0x8000; \
2360 dup_8x16b(d128_0x0421, 0x0421); \
2361 dup_8x16b(d128_0x8000, 0x8000); \
2363 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2364 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2365 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2366 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2367 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2368 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2371 #define blend_blocks_add() \
2373 vec_8x16u pixels_rb, pixels_g; \
2374 vec_8x16u fb_rb, fb_g; \
2376 vec_8x16u d128_0x7C1F; \
2377 vec_8x16u d128_0x03E0; \
2379 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2380 dup_8x16b(d128_0x03E0, 0x03E0); \
2382 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2383 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2385 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2386 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2388 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2389 add_8x16b(fb_g, fb_g, pixels_g); \
2391 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2392 vector_cast(vec_16x8u, d128_0x7C1F)); \
2393 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2395 or_8x16b(blend_pixels, fb_rb, fb_g); \
2398 #define blend_blocks_subtract() \
2400 vec_8x16u pixels_rb, pixels_g; \
2401 vec_8x16u fb_rb, fb_g; \
2403 vec_8x16u d128_0x7C1F; \
2404 vec_8x16u d128_0x03E0; \
2406 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2407 dup_8x16b(d128_0x03E0, 0x03E0); \
2409 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2410 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2412 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2413 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2415 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2416 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2417 subs_8x16b(fb_g, fb_g, pixels_g); \
2419 or_8x16b(blend_pixels, fb_rb, fb_g); \
2422 #define blend_blocks_add_fourth() \
2424 vec_8x16u pixels_rb, pixels_g; \
2425 vec_8x16u pixels_fourth; \
2426 vec_8x16u fb_rb, fb_g; \
2428 vec_8x16u d128_0x7C1F; \
2429 vec_8x16u d128_0x1C07; \
2430 vec_8x16u d128_0x03E0; \
2431 vec_8x16u d128_0x00E0; \
2433 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2434 dup_8x16b(d128_0x1C07, 0x1C07); \
2435 dup_8x16b(d128_0x03E0, 0x03E0); \
2436 dup_8x16b(d128_0x00E0, 0x00E0); \
2438 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2440 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2441 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2443 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2444 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2446 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2447 add_8x16b(fb_g, fb_g, pixels_g); \
2449 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2450 vector_cast(vec_16x8u, d128_0x7C1F)); \
2451 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2453 or_8x16b(blend_pixels, fb_rb, fb_g); \
2456 #define blend_blocks_blended_combine_textured() \
2458 vec_8x16u blend_mask; \
2459 cmpltz_8x16b(blend_mask, pixels); \
2461 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2462 bif_8x16b(blend_pixels, pixels, blend_mask); \
2465 #define blend_blocks_blended_combine_untextured() \
2468 #define blend_blocks_body_blend(blend_mode, texturing) \
2470 blend_blocks_##blend_mode(); \
2471 blend_blocks_blended_combine_##texturing(); \
2474 #define blend_blocks_body_average(texturing) \
2475 blend_blocks_body_blend(average, texturing) \
2477 #define blend_blocks_body_add(texturing) \
2478 blend_blocks_body_blend(add, texturing) \
2480 #define blend_blocks_body_subtract(texturing) \
2481 blend_blocks_body_blend(subtract, texturing) \
2483 #define blend_blocks_body_add_fourth(texturing) \
2484 blend_blocks_body_blend(add_fourth, texturing) \
2486 #define blend_blocks_body_unblended(texturing) \
2487 blend_pixels = pixels \
2490 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2492 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2495 block_struct *block = psx_gpu->blocks; \
2496 u32 num_blocks = psx_gpu->num_blocks; \
2497 vec_8x16u draw_mask; \
2499 vec_8x16u blend_pixels; \
2500 vec_8x16u framebuffer_pixels; \
2501 vec_8x16u msb_mask; \
2505 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2509 pixels = block->pixels; \
2510 draw_mask = block->draw_mask; \
2511 fb_ptr = block->fb_ptr; \
2513 load_8x16b(framebuffer_pixels, fb_ptr); \
2515 blend_blocks_mask_evaluate_##mask_evaluate(); \
2516 blend_blocks_body_##blend_mode(texturing); \
2518 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2519 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2520 store_8x16b(framebuffer_pixels, fb_ptr); \
2530 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2534 blend_blocks_builder(textured, average, off);
2535 blend_blocks_builder(textured, average, on);
2536 blend_blocks_builder(textured, add, off);
2537 blend_blocks_builder(textured, add, on);
2538 blend_blocks_builder(textured, subtract, off);
2539 blend_blocks_builder(textured, subtract, on);
2540 blend_blocks_builder(textured, add_fourth, off);
2541 blend_blocks_builder(textured, add_fourth, on);
2543 blend_blocks_builder(untextured, average, off);
2544 blend_blocks_builder(untextured, average, on);
2545 blend_blocks_builder(untextured, add, off);
2546 blend_blocks_builder(untextured, add, on);
2547 blend_blocks_builder(untextured, subtract, off);
2548 blend_blocks_builder(untextured, subtract, on);
2549 blend_blocks_builder(untextured, add_fourth, off);
2550 blend_blocks_builder(untextured, add_fourth, on);
2552 blend_blocks_builder(textured, unblended, on);
2557 #define vertex_swap(_a, _b) \
2559 vertex_struct *temp_vertex = _a; \
2562 triangle_winding ^= 1; \
2566 // Setup blocks parametric-variables:
2567 // SHADE TEXTURE_MAP SWIZZLING
2574 // 8 inputs, 6 combinations
2576 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2577 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2579 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2580 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2582 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2584 setup_blocks_switch_untextured_##shading(dithering, target) \
2586 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2587 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2589 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2590 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2592 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2593 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2595 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2596 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2598 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2599 dithering, mask_evaluate) \
2600 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2602 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2604 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2606 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2608 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2610 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2611 dithering, mask_evaluate) \
2612 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2613 texture_mode, dithering) \
2615 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2616 blending, mask_evaluate) \
2617 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2618 dithering, mask_evaluate) \
2623 #define texture_blocks_switch_untextured(texture_mode) \
2624 texture_blocks_untextured \
2626 #define texture_blocks_switch_textured(texture_mode) \
2627 texture_blocks_##texture_mode \
2629 #define texture_blocks_switch(texturing, texture_mode) \
2630 texture_blocks_switch_##texturing(texture_mode) \
2633 // Shade blocks parametric-variables:
2634 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2643 // 16 inputs, 8 combinations
2645 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2646 shade_blocks_unshaded_untextured_##target \
2648 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2649 shade_blocks_textured_unmodulated_##target \
2651 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2652 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2654 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2655 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2657 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2658 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2660 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2661 shade_blocks_shaded_untextured \
2663 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2664 shade_blocks_textured_unmodulated_##target \
2666 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2667 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2669 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2670 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2672 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2673 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2675 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2677 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2679 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2681 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2683 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2685 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2687 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2688 dithering, mask_evaluate) \
2689 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2692 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2693 blending, mask_evaluate) \
2694 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2698 // Blend blocks parametric-variables:
2699 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2718 // 32 inputs, 18 combinations
2720 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2721 blend_blocks_textured_unblended_##mask_evaluate \
2723 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2724 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2726 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2727 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2730 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2731 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2733 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2735 texture_blocks_switch(texturing, texture_mode), \
2736 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2738 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2741 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2742 mask_evaluate, shading, dithering, texturing, blending) \
2743 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2744 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2745 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2746 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2748 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2749 mask_evaluate, shading, dithering, texturing) \
2750 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2751 mask_evaluate, shading, dithering, texturing, unblended), \
2752 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2753 mask_evaluate, shading, dithering, texturing, blended) \
2755 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2756 mask_evaluate, shading, dithering) \
2757 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2758 mask_evaluate, shading, dithering, untextured), \
2759 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2760 mask_evaluate, shading, dithering, textured) \
2762 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2763 mask_evaluate, shading) \
2764 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2765 mask_evaluate, shading, undithered), \
2766 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2767 mask_evaluate, shading, dithered) \
2769 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2771 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2773 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2776 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2777 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2778 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2780 #define render_blocks_switch_block_texture_mode(texture_mode) \
2781 render_blocks_switch_block_blend_mode(texture_mode, average), \
2782 render_blocks_switch_block_blend_mode(texture_mode, add), \
2783 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2784 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2786 #define render_blocks_switch_block() \
2787 render_blocks_switch_block_texture_mode(4bpp), \
2788 render_blocks_switch_block_texture_mode(8bpp), \
2789 render_blocks_switch_block_texture_mode(16bpp), \
2790 render_blocks_switch_block_texture_mode(16bpp) \
2793 render_block_handler_struct render_triangle_block_handlers[] =
2795 render_blocks_switch_block()
2798 #undef render_blocks_switch_block_modulation
2800 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2801 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2803 "texture mode: " #texture_mode "\n" \
2804 "blend mode: " #blend_mode "\n" \
2805 "mask evaluation: " #mask_evaluate "\n" \
2812 char *render_block_flag_strings[] =
2814 render_blocks_switch_block()
2818 #define triangle_y_direction_up 1
2819 #define triangle_y_direction_flat 2
2820 #define triangle_y_direction_down 0
2822 #define triangle_winding_positive 0
2823 #define triangle_winding_negative 1
2825 #define triangle_set_direction(direction_variable, value) \
2826 u32 direction_variable = (u32)(value) >> 31; \
2828 direction_variable = 2 \
2830 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2831 case (triangle_y_direction_##direction_a | \
2832 (triangle_y_direction_##direction_b << 2) | \
2833 (triangle_y_direction_##direction_c << 4) | \
2834 (triangle_winding_##winding << 6)) \
2836 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2837 vertex_struct *vertexes_out[3])
2839 s32 y_top, y_bottom;
2841 u32 triangle_winding = 0;
2843 vertex_struct *a = &(vertexes[0]);
2844 vertex_struct *b = &(vertexes[1]);
2845 vertex_struct *c = &(vertexes[2]);
2847 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2853 if(triangle_area == 0)
2875 if((y_bottom - y_top) >= 512)
2883 if(triangle_area < 0)
2885 triangle_area = -triangle_area;
2886 triangle_winding ^= 1;
2901 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2909 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2918 psx_gpu->triangle_area = triangle_area;
2919 psx_gpu->triangle_winding = triangle_winding;
2921 vertexes_out[0] = a;
2922 vertexes_out[1] = b;
2923 vertexes_out[2] = c;
2928 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2929 vertex_struct *vertex_ptrs[3], u32 flags)
2931 psx_gpu->num_spans = 0;
2933 vertex_struct *a = vertex_ptrs[0];
2934 vertex_struct *b = vertex_ptrs[1];
2935 vertex_struct *c = vertex_ptrs[2];
2937 s32 y_delta_a = b->y - a->y;
2938 s32 y_delta_b = c->y - b->y;
2939 s32 y_delta_c = c->y - a->y;
2941 triangle_set_direction(y_direction_a, y_delta_a);
2942 triangle_set_direction(y_direction_b, y_delta_b);
2943 triangle_set_direction(y_direction_c, y_delta_c);
2945 compute_all_gradients(psx_gpu, a, b, c);
2947 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2948 (psx_gpu->triangle_winding << 6))
2950 triangle_case(up, up, up, negative):
2951 triangle_case(up, up, flat, negative):
2952 triangle_case(up, up, down, negative):
2953 setup_spans_up_right(psx_gpu, a, b, c);
2956 triangle_case(flat, up, up, negative):
2957 triangle_case(flat, up, flat, negative):
2958 triangle_case(flat, up, down, negative):
2959 setup_spans_up_a(psx_gpu, a, b, c);
2962 triangle_case(down, up, up, negative):
2963 setup_spans_up_down(psx_gpu, a, c, b);
2966 triangle_case(down, up, flat, negative):
2967 setup_spans_down_a(psx_gpu, a, c, b);
2970 triangle_case(down, up, down, negative):
2971 setup_spans_down_right(psx_gpu, a, c, b);
2974 triangle_case(down, flat, up, negative):
2975 triangle_case(down, flat, flat, negative):
2976 triangle_case(down, flat, down, negative):
2977 setup_spans_down_b(psx_gpu, a, b, c);
2980 triangle_case(down, down, up, negative):
2981 triangle_case(down, down, flat, negative):
2982 triangle_case(down, down, down, negative):
2983 setup_spans_down_left(psx_gpu, a, b, c);
2986 triangle_case(up, up, up, positive):
2987 triangle_case(up, up, flat, positive):
2988 triangle_case(up, up, down, positive):
2989 setup_spans_up_left(psx_gpu, a, b, c);
2992 triangle_case(up, flat, up, positive):
2993 triangle_case(up, flat, flat, positive):
2994 triangle_case(up, flat, down, positive):
2995 setup_spans_up_b(psx_gpu, a, b, c);
2998 triangle_case(up, down, up, positive):
2999 setup_spans_up_right(psx_gpu, a, c, b);
3002 triangle_case(up, down, flat, positive):
3003 setup_spans_up_a(psx_gpu, a, c, b);
3006 triangle_case(up, down, down, positive):
3007 setup_spans_up_down(psx_gpu, a, b, c);
3010 triangle_case(flat, down, up, positive):
3011 triangle_case(flat, down, flat, positive):
3012 triangle_case(flat, down, down, positive):
3013 setup_spans_down_a(psx_gpu, a, b, c);
3016 triangle_case(down, down, up, positive):
3017 triangle_case(down, down, flat, positive):
3018 triangle_case(down, down, down, positive):
3019 setup_spans_down_right(psx_gpu, a, b, c);
3024 spans += psx_gpu->num_spans;
3027 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3031 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3033 for(i = 0; i < psx_gpu->num_spans; i++)
3035 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3036 psx_gpu->span_edge_data[i].num_blocks = 0;
3041 for(i = 0; i < psx_gpu->num_spans; i++)
3043 if(psx_gpu->span_edge_data[i].y & 1)
3044 psx_gpu->span_edge_data[i].num_blocks = 0;
3049 u32 render_state = flags &
3050 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3051 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3052 render_state |= psx_gpu->render_state_base;
3054 if((psx_gpu->render_state != render_state) ||
3055 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3057 psx_gpu->render_state = render_state;
3058 flush_render_block_buffer(psx_gpu);
3064 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3066 psx_gpu->render_block_handler =
3067 &(render_triangle_block_handlers[render_state]);
3068 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3072 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3075 vertex_struct *vertex_ptrs[3];
3076 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3077 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3080 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3082 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3084 block_struct *block = psx_gpu->blocks;
3085 u32 num_blocks = psx_gpu->num_blocks;
3088 vec_8x8u texel_indexes;
3090 u16 *clut_ptr = psx_gpu->clut_ptr;
3095 texel_indexes = block->r;
3097 for(i = 0; i < 8; i++)
3099 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3102 block->texels = texels;
3112 #define setup_sprite_tiled_initialize_4bpp_clut() \
3113 u16 *clut_ptr = psx_gpu->clut_ptr; \
3114 vec_8x16u clut_a, clut_b; \
3115 vec_16x8u clut_low, clut_high; \
3117 load_8x16b(clut_a, clut_ptr); \
3118 load_8x16b(clut_b, clut_ptr + 8); \
3119 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3121 #define setup_sprite_tiled_initialize_4bpp() \
3122 setup_sprite_tiled_initialize_4bpp_clut(); \
3124 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3125 update_texture_4bpp_cache(psx_gpu) \
3127 #define setup_sprite_tiled_initialize_8bpp() \
3128 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3129 update_texture_8bpp_cache(psx_gpu) \
3132 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3133 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3134 ((texture_offset + offset) & texture_mask); \
3136 load_64b(texels, texture_block_ptr) \
3139 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3140 num_blocks += tile_num_blocks; \
3141 sprite_blocks += tile_num_blocks; \
3143 if(num_blocks > MAX_BLOCKS) \
3145 flush_render_block_buffer(psx_gpu); \
3146 num_blocks = tile_num_blocks; \
3147 block = psx_gpu->blocks; \
3150 #define setup_sprite_tile_full_4bpp(edge) \
3152 vec_8x8u texels_low, texels_high; \
3154 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3156 while(sub_tile_height) \
3158 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3159 tbl_16(texels_low, texels, clut_low); \
3160 tbl_16(texels_high, texels, clut_high); \
3161 zip_8x16b(pixels, texels_low, texels_high); \
3163 block->texels = pixels; \
3164 block->draw_mask_bits = left_mask_bits; \
3165 block->fb_ptr = fb_ptr; \
3168 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3169 tbl_16(texels_low, texels, clut_low); \
3170 tbl_16(texels_high, texels, clut_high); \
3171 zip_8x16b(pixels, texels_low, texels_high); \
3173 block->texels = pixels; \
3174 block->draw_mask_bits = right_mask_bits; \
3175 block->fb_ptr = fb_ptr + 8; \
3179 texture_offset += 0x10; \
3180 sub_tile_height--; \
3182 texture_offset += 0xF00; \
3183 psx_gpu->num_blocks = num_blocks; \
3186 #define setup_sprite_tile_half_4bpp(edge) \
3188 vec_8x8u texels_low, texels_high; \
3190 setup_sprite_tile_add_blocks(sub_tile_height); \
3192 while(sub_tile_height) \
3194 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3195 tbl_16(texels_low, texels, clut_low); \
3196 tbl_16(texels_high, texels, clut_high); \
3197 zip_8x16b(pixels, texels_low, texels_high); \
3199 block->texels = pixels; \
3200 block->draw_mask_bits = edge##_mask_bits; \
3201 block->fb_ptr = fb_ptr; \
3205 texture_offset += 0x10; \
3206 sub_tile_height--; \
3208 texture_offset += 0xF00; \
3209 psx_gpu->num_blocks = num_blocks; \
3213 #define setup_sprite_tile_full_8bpp(edge) \
3215 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3217 while(sub_tile_height) \
3219 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3220 block->r = texels; \
3221 block->draw_mask_bits = left_mask_bits; \
3222 block->fb_ptr = fb_ptr; \
3225 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3226 block->r = texels; \
3227 block->draw_mask_bits = right_mask_bits; \
3228 block->fb_ptr = fb_ptr + 8; \
3232 texture_offset += 0x10; \
3233 sub_tile_height--; \
3235 texture_offset += 0xF00; \
3236 psx_gpu->num_blocks = num_blocks; \
3239 #define setup_sprite_tile_half_8bpp(edge) \
3241 setup_sprite_tile_add_blocks(sub_tile_height); \
3243 while(sub_tile_height) \
3245 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3246 block->r = texels; \
3247 block->draw_mask_bits = edge##_mask_bits; \
3248 block->fb_ptr = fb_ptr; \
3252 texture_offset += 0x10; \
3253 sub_tile_height--; \
3255 texture_offset += 0xF00; \
3256 psx_gpu->num_blocks = num_blocks; \
3260 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3261 texture_offset = texture_offset_base + 8; \
3264 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3265 texture_offset = texture_offset_base \
3267 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3268 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3270 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3271 texture_offset = texture_offset_base \
3273 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3276 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3278 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3279 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3281 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3284 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3288 sub_tile_height = column_data; \
3289 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3290 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3291 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3294 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3298 u32 tiles_remaining = column_data >> 16; \
3299 sub_tile_height = column_data & 0xFF; \
3300 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3301 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3302 tiles_remaining -= 1; \
3304 while(tiles_remaining) \
3306 sub_tile_height = 16; \
3307 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3308 tiles_remaining--; \
3311 sub_tile_height = (column_data >> 8) & 0xFF; \
3312 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3313 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3317 #define setup_sprite_column_data_single() \
3318 column_data = height \
3320 #define setup_sprite_column_data_multi() \
3321 column_data = 16 - offset_v; \
3322 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3323 column_data |= (tile_height - 1) << 16 \
3326 #define RIGHT_MASK_BIT_SHIFT 8
3327 #define RIGHT_MASK_BIT_SHIFT_4x 16
3329 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3330 edge_mode, edge, x4mode) \
3332 setup_sprite_column_data_##multi_height(); \
3333 left_mask_bits = left_block_mask | right_block_mask; \
3334 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3336 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3337 texture_mode, x4mode); \
3340 #define setup_sprite_tiled_advance_column() \
3341 texture_offset_base += 0x100; \
3342 if((texture_offset_base & 0xF00) == 0) \
3343 texture_offset_base -= (0x100 + 0xF00) \
3345 #define FB_PTR_MULTIPLIER 1
3346 #define FB_PTR_MULTIPLIER_4x 2
3348 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3349 left_mode, right_mode, x4mode) \
3351 setup_sprite_column_data_##multi_height(); \
3352 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3353 * FB_PTR_MULTIPLIER##x4mode; \
3356 left_mask_bits = left_block_mask; \
3357 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3359 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3360 texture_mode, x4mode); \
3361 fb_ptr += fb_ptr_advance_column; \
3363 left_mask_bits = 0x00; \
3364 right_mask_bits = 0x00; \
3368 setup_sprite_tiled_advance_column(); \
3369 setup_sprite_tile_column_height_##multi_height(full, none, \
3370 texture_mode, x4mode); \
3371 fb_ptr += fb_ptr_advance_column; \
3375 left_mask_bits = right_block_mask; \
3376 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3378 setup_sprite_tiled_advance_column(); \
3379 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3380 texture_mode, x4mode); \
3385 #define setup_sprite_tiled_initialize_4bpp_4x() \
3386 setup_sprite_tiled_initialize_4bpp_clut() \
3388 #define setup_sprite_tiled_initialize_8bpp_4x() \
3391 #define setup_sprite_tile_full_4bpp_4x(edge) \
3393 vec_8x8u texels_low, texels_high; \
3394 vec_8x16u pixels, pixels_wide; \
3395 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3396 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3397 u32 left_mask_bits_b = left_mask_bits >> 8; \
3398 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3399 u32 right_mask_bits_b = right_mask_bits >> 8; \
3401 while(sub_tile_height) \
3403 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3404 tbl_16(texels_low, texels, clut_low); \
3405 tbl_16(texels_high, texels, clut_high); \
3406 zip_8x16b(pixels, texels_low, texels_high); \
3408 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3409 block->texels = pixels_wide; \
3410 block->draw_mask_bits = left_mask_bits_a; \
3411 block->fb_ptr = fb_ptr; \
3414 block->texels = pixels_wide; \
3415 block->draw_mask_bits = left_mask_bits_a; \
3416 block->fb_ptr = fb_ptr + 1024; \
3419 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3420 block->texels = pixels_wide; \
3421 block->draw_mask_bits = left_mask_bits_b; \
3422 block->fb_ptr = fb_ptr + 8; \
3425 block->texels = pixels_wide; \
3426 block->draw_mask_bits = left_mask_bits_b; \
3427 block->fb_ptr = fb_ptr + 1024 + 8; \
3430 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3431 tbl_16(texels_low, texels, clut_low); \
3432 tbl_16(texels_high, texels, clut_high); \
3433 zip_8x16b(pixels, texels_low, texels_high); \
3435 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3436 block->texels = pixels_wide; \
3437 block->draw_mask_bits = right_mask_bits_a; \
3438 block->fb_ptr = fb_ptr + 16; \
3441 block->texels = pixels_wide; \
3442 block->draw_mask_bits = right_mask_bits_a; \
3443 block->fb_ptr = fb_ptr + 1024 + 16; \
3446 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3447 block->texels = pixels_wide; \
3448 block->draw_mask_bits = right_mask_bits_b; \
3449 block->fb_ptr = fb_ptr + 24; \
3452 block->texels = pixels_wide; \
3453 block->draw_mask_bits = right_mask_bits_b; \
3454 block->fb_ptr = fb_ptr + 1024 + 24; \
3458 texture_offset += 0x10; \
3459 sub_tile_height--; \
3461 texture_offset += 0xF00; \
3462 psx_gpu->num_blocks = num_blocks; \
3465 #define setup_sprite_tile_half_4bpp_4x(edge) \
3467 vec_8x8u texels_low, texels_high; \
3468 vec_8x16u pixels, pixels_wide; \
3469 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3470 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3471 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3473 while(sub_tile_height) \
3475 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3476 tbl_16(texels_low, texels, clut_low); \
3477 tbl_16(texels_high, texels, clut_high); \
3478 zip_8x16b(pixels, texels_low, texels_high); \
3480 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3481 block->texels = pixels_wide; \
3482 block->draw_mask_bits = edge##_mask_bits_a; \
3483 block->fb_ptr = fb_ptr; \
3486 block->texels = pixels_wide; \
3487 block->draw_mask_bits = edge##_mask_bits_a; \
3488 block->fb_ptr = fb_ptr + 1024; \
3491 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3492 block->texels = pixels_wide; \
3493 block->draw_mask_bits = edge##_mask_bits_b; \
3494 block->fb_ptr = fb_ptr + 8; \
3497 block->texels = pixels_wide; \
3498 block->draw_mask_bits = edge##_mask_bits_b; \
3499 block->fb_ptr = fb_ptr + 1024 + 8; \
3503 texture_offset += 0x10; \
3504 sub_tile_height--; \
3506 texture_offset += 0xF00; \
3507 psx_gpu->num_blocks = num_blocks; \
3511 #define setup_sprite_tile_full_8bpp_4x(edge) \
3513 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3514 vec_16x8u texels_wide; \
3515 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3516 u32 left_mask_bits_b = left_mask_bits >> 8; \
3517 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3518 u32 right_mask_bits_b = right_mask_bits >> 8; \
3520 while(sub_tile_height) \
3522 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3523 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3524 block->r = texels_wide.low; \
3525 block->draw_mask_bits = left_mask_bits_a; \
3526 block->fb_ptr = fb_ptr; \
3529 block->r = texels_wide.low; \
3530 block->draw_mask_bits = left_mask_bits_a; \
3531 block->fb_ptr = fb_ptr + 1024; \
3534 block->r = texels_wide.high; \
3535 block->draw_mask_bits = left_mask_bits_b; \
3536 block->fb_ptr = fb_ptr + 8; \
3539 block->r = texels_wide.high; \
3540 block->draw_mask_bits = left_mask_bits_b; \
3541 block->fb_ptr = fb_ptr + 1024 + 8; \
3544 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3545 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3546 block->r = texels_wide.low; \
3547 block->draw_mask_bits = right_mask_bits_a; \
3548 block->fb_ptr = fb_ptr + 16; \
3551 block->r = texels_wide.low; \
3552 block->draw_mask_bits = right_mask_bits_a; \
3553 block->fb_ptr = fb_ptr + 1024 + 16; \
3556 block->r = texels_wide.high; \
3557 block->draw_mask_bits = right_mask_bits_b; \
3558 block->fb_ptr = fb_ptr + 24; \
3561 block->r = texels_wide.high; \
3562 block->draw_mask_bits = right_mask_bits_b; \
3563 block->fb_ptr = fb_ptr + 24 + 1024; \
3567 texture_offset += 0x10; \
3568 sub_tile_height--; \
3570 texture_offset += 0xF00; \
3571 psx_gpu->num_blocks = num_blocks; \
3574 #define setup_sprite_tile_half_8bpp_4x(edge) \
3576 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3577 vec_16x8u texels_wide; \
3578 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3579 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3581 while(sub_tile_height) \
3583 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3584 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3585 block->r = texels_wide.low; \
3586 block->draw_mask_bits = edge##_mask_bits_a; \
3587 block->fb_ptr = fb_ptr; \
3590 block->r = texels_wide.low; \
3591 block->draw_mask_bits = edge##_mask_bits_a; \
3592 block->fb_ptr = fb_ptr + 1024; \
3595 block->r = texels_wide.high; \
3596 block->draw_mask_bits = edge##_mask_bits_b; \
3597 block->fb_ptr = fb_ptr + 8; \
3600 block->r = texels_wide.high; \
3601 block->draw_mask_bits = edge##_mask_bits_b; \
3602 block->fb_ptr = fb_ptr + 8 + 1024; \
3606 texture_offset += 0x10; \
3607 sub_tile_height--; \
3609 texture_offset += 0xF00; \
3610 psx_gpu->num_blocks = num_blocks; \
3614 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3615 texture_offset = texture_offset_base + 8; \
3618 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3619 texture_offset = texture_offset_base \
3621 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3622 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3624 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3625 texture_offset = texture_offset_base \
3627 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3630 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3632 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3633 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3635 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3638 #define setup_sprite_offset_u_adjust() \
3640 #define setup_sprite_comapre_left_block_mask() \
3641 ((left_block_mask & 0xFF) == 0xFF) \
3643 #define setup_sprite_comapre_right_block_mask() \
3644 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3647 #define setup_sprite_offset_u_adjust_4x() \
3649 offset_u_right = offset_u_right * 2 + 1 \
3651 #define setup_sprite_comapre_left_block_mask_4x() \
3652 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3654 #define setup_sprite_comapre_right_block_mask_4x() \
3655 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3658 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3659 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3660 s32 u, s32 v, s32 width, s32 height, u32 color) \
3662 s32 offset_u = u & 0xF; \
3663 s32 offset_v = v & 0xF; \
3665 s32 width_rounded = offset_u + width + 15; \
3666 s32 height_rounded = offset_v + height + 15; \
3667 s32 tile_height = height_rounded / 16; \
3668 s32 tile_width = width_rounded / 16; \
3669 u32 offset_u_right = width_rounded & 0xF; \
3671 setup_sprite_offset_u_adjust##x4mode(); \
3673 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3674 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3676 u32 left_mask_bits; \
3677 u32 right_mask_bits; \
3679 u32 sub_tile_height; \
3682 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3683 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3684 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3685 ((psx_gpu->texture_mask_height >> 4) << 12); \
3686 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3687 ((v & 0xF0) << 8); \
3688 u32 texture_offset_base = texture_offset; \
3691 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3692 u32 num_blocks = psx_gpu->num_blocks; \
3693 block_struct *block = psx_gpu->blocks + num_blocks; \
3695 u8 *texture_block_ptr; \
3698 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3700 control_mask = tile_width == 1; \
3701 control_mask |= (tile_height == 1) << 1; \
3702 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3703 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3705 sprites_##texture_mode++; \
3707 switch(control_mask) \
3711 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3716 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3721 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3726 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3731 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3736 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3741 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3746 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3751 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3756 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3761 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3766 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3771 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3776 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3783 setup_sprite_tiled_builder(4bpp,);
3784 setup_sprite_tiled_builder(8bpp,);
3786 setup_sprite_tiled_builder(4bpp,_4x);
3787 setup_sprite_tiled_builder(8bpp,_4x);
3790 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3792 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3793 s32 v, s32 width, s32 height, u32 color)
3795 u32 left_offset = u & 0x7;
3796 u32 width_rounded = width + left_offset + 7;
3798 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3799 u32 right_width = width_rounded & 0x7;
3800 u32 block_width = width_rounded / 8;
3801 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3803 u32 left_mask_bits = ~(0xFF << left_offset);
3804 u32 right_mask_bits = 0xFE << right_width;
3806 u32 texture_offset_base = u + (v * 1024);
3808 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3810 u32 blocks_remaining;
3811 u32 num_blocks = psx_gpu->num_blocks;
3812 block_struct *block = psx_gpu->blocks + num_blocks;
3814 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3815 u16 *texture_block_ptr;
3817 texture_offset_base &= ~0x7;
3819 stats_add(sprites_16bpp, 1);
3821 if(block_width == 1)
3823 u32 mask_bits = left_mask_bits | right_mask_bits;
3830 if(num_blocks > MAX_BLOCKS)
3832 flush_render_block_buffer(psx_gpu);
3834 block = psx_gpu->blocks;
3838 texture_page_ptr + (texture_offset_base & texture_mask);
3840 block->texels = *(vec_8x16u *)texture_block_ptr;
3841 block->draw_mask_bits = mask_bits;
3842 block->fb_ptr = fb_ptr;
3846 texture_offset_base += 1024;
3850 psx_gpu->num_blocks = num_blocks;
3859 blocks_remaining = block_width - 2;
3860 num_blocks += block_width;
3861 sprite_blocks += block_width;
3863 if(num_blocks > MAX_BLOCKS)
3865 flush_render_block_buffer(psx_gpu);
3866 num_blocks = block_width;
3867 block = psx_gpu->blocks;
3870 texture_offset = texture_offset_base;
3871 texture_offset_base += 1024;
3873 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3874 block->texels = *(vec_8x16u *)texture_block_ptr;
3876 block->draw_mask_bits = left_mask_bits;
3877 block->fb_ptr = fb_ptr;
3879 texture_offset += 8;
3883 while(blocks_remaining)
3885 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3886 block->texels = *(vec_8x16u *)texture_block_ptr;
3888 block->draw_mask_bits = 0;
3889 block->fb_ptr = fb_ptr;
3891 texture_offset += 8;
3898 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3899 block->texels = *(vec_8x16u *)texture_block_ptr;
3901 block->draw_mask_bits = right_mask_bits;
3902 block->fb_ptr = fb_ptr;
3904 fb_ptr += fb_ptr_pitch;
3908 psx_gpu->num_blocks = num_blocks;
3917 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3918 s32 v, s32 width, s32 height, u32 color)
3920 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
3921 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
3922 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
3924 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
3928 u32 right_width = ((width - 1) & 0x7) + 1;
3929 u32 right_mask_bits = (0xFF << right_width);
3930 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3931 u32 block_width = (width + 7) / 8;
3932 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3933 u32 blocks_remaining;
3934 u32 num_blocks = psx_gpu->num_blocks;
3935 block_struct *block = psx_gpu->blocks + num_blocks;
3937 u32 color_r = color & 0xFF;
3938 u32 color_g = (color >> 8) & 0xFF;
3939 u32 color_b = (color >> 16) & 0xFF;
3941 vec_8x16u right_mask;
3942 vec_8x16u test_mask = psx_gpu->test_mask;
3943 vec_8x16u zero_mask;
3945 sprites_untextured++;
3947 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3949 dup_8x16b(colors, color);
3950 dup_8x16b(zero_mask, 0x00);
3951 dup_8x16b(right_mask, right_mask_bits);
3952 tst_8x16b(right_mask, right_mask, test_mask);
3956 blocks_remaining = block_width - 1;
3957 num_blocks += block_width;
3960 sprite_blocks += block_width;
3963 if(num_blocks > MAX_BLOCKS)
3965 flush_render_block_buffer(psx_gpu);
3966 num_blocks = block_width;
3967 block = psx_gpu->blocks;
3970 while(blocks_remaining)
3972 block->pixels = colors;
3973 block->draw_mask = zero_mask;
3974 block->fb_ptr = fb_ptr;
3981 block->pixels = colors;
3982 block->draw_mask = right_mask;
3983 block->fb_ptr = fb_ptr;
3986 fb_ptr += fb_ptr_pitch;
3989 psx_gpu->num_blocks = num_blocks;
3995 void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y,
3996 s32 u, s32 v, s32 width, s32 height, u32 color)
3998 u32 r = color & 0xFF;
3999 u32 g = (color >> 8) & 0xFF;
4000 u32 b = (color >> 16) & 0xFF;
4001 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4003 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4005 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4010 if(psx_gpu->num_blocks > MAX_BLOCKS)
4012 flush_render_block_buffer(psx_gpu);
4019 vram_ptr = (void *)vram_ptr16;
4020 if((uintptr_t)vram_ptr16 & 2)
4022 *vram_ptr16 = color_32bpp;
4023 vram_ptr = (void *)(vram_ptr16 + 1);
4027 while(num_width >= 4 * 2)
4029 vram_ptr[0] = color_32bpp;
4030 vram_ptr[1] = color_32bpp;
4031 vram_ptr[2] = color_32bpp;
4032 vram_ptr[3] = color_32bpp;
4038 while(num_width >= 2)
4040 *vram_ptr++ = color_32bpp;
4046 *(u16 *)vram_ptr = color_32bpp;
4055 #define setup_sprite_blocks_switch_textured(texture_mode) \
4056 setup_sprite_##texture_mode \
4058 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4059 setup_sprite_untextured \
4061 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4062 setup_sprite_blocks_switch_##texturing(texture_mode) \
4065 #define texture_sprite_blocks_switch_4bpp() \
4066 texture_blocks_untextured \
4068 #define texture_sprite_blocks_switch_8bpp() \
4069 texture_sprite_blocks_8bpp \
4071 #define texture_sprite_blocks_switch_16bpp() \
4072 texture_blocks_untextured \
4074 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4075 texture_blocks_untextured \
4077 #define texture_sprite_blocks_switch_textured(texture_mode) \
4078 texture_sprite_blocks_switch_##texture_mode() \
4080 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4081 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4083 setup_sprite_blocks_switch(texturing, texture_mode), \
4084 texture_sprite_blocks_switch_##texturing(texture_mode), \
4085 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4087 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4090 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4091 mask_evaluate, shading, dithering, texturing, blending) \
4092 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4093 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4094 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4095 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4097 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4098 mask_evaluate, shading, dithering, texturing) \
4099 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4100 mask_evaluate, shading, dithering, texturing, unblended), \
4101 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4102 mask_evaluate, shading, dithering, texturing, blended) \
4104 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4105 mask_evaluate, shading, dithering) \
4106 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4107 mask_evaluate, shading, dithering, untextured), \
4108 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4109 mask_evaluate, shading, dithering, textured) \
4111 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4112 mask_evaluate, shading) \
4113 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4114 mask_evaluate, shading, undithered), \
4115 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4116 mask_evaluate, shading, dithered) \
4118 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4119 blend_mode, mask_evaluate) \
4120 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4121 mask_evaluate, unshaded), \
4122 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4123 mask_evaluate, shaded) \
4125 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4126 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4128 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4131 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4132 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4133 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4134 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4135 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4137 #define render_sprite_blocks_switch_block() \
4138 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4139 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4140 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4141 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4144 render_block_handler_struct render_sprite_block_handlers[] =
4146 render_sprite_blocks_switch_block()
4150 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4151 s32 width, s32 height, u32 flags, u32 color)
4153 s32 x_right = x + width - 1;
4154 s32 y_bottom = y + height - 1;
4160 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4166 if(x < psx_gpu->viewport_start_x)
4168 u32 clip = psx_gpu->viewport_start_x - x;
4174 if(y < psx_gpu->viewport_start_y)
4176 s32 clip = psx_gpu->viewport_start_y - y;
4182 if(x_right > psx_gpu->viewport_end_x)
4183 width -= x_right - psx_gpu->viewport_end_x;
4185 if(y_bottom > psx_gpu->viewport_end_y)
4186 height -= y_bottom - psx_gpu->viewport_end_y;
4188 if((width <= 0) || (height <= 0))
4192 span_pixels += width * height;
4196 u32 render_state = flags &
4197 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4198 RENDER_FLAGS_TEXTURE_MAP);
4200 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4202 if((psx_gpu->render_state != render_state) ||
4203 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4205 psx_gpu->render_state = render_state;
4206 flush_render_block_buffer(psx_gpu);
4212 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4216 if(psx_gpu->triangle_color != color)
4218 flush_render_block_buffer(psx_gpu);
4219 psx_gpu->triangle_color = color;
4222 if(color == 0x808080)
4223 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4225 render_block_handler_struct *render_block_handler =
4226 &(render_sprite_block_handlers[render_state]);
4227 psx_gpu->render_block_handler = render_block_handler;
4229 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4230 (psx_gpu, x, y, u, v, width, height, color);
4233 #define draw_pixel_line_mask_evaluate_yes() \
4234 if((*vram_ptr & 0x8000) == 0) \
4236 #define draw_pixel_line_mask_evaluate_no() \
4239 #define draw_pixel_line_shaded() \
4241 color_r = fixed_to_int(current_r); \
4242 color_g = fixed_to_int(current_g); \
4243 color_b = fixed_to_int(current_b); \
4245 current_r += gradient_r; \
4246 current_g += gradient_g; \
4247 current_b += gradient_b; \
4250 #define draw_pixel_line_unshaded() \
4252 color_r = color & 0xFF; \
4253 color_g = (color >> 8) & 0xFF; \
4254 color_b = (color >> 16) & 0xFF; \
4258 #define draw_pixel_line_dithered(_x, _y) \
4260 u32 dither_xor = _x ^ _y; \
4261 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4262 dither_offset |= (_y & 0x1) << 1; \
4263 dither_offset |= (dither_xor & 0x1) << 2; \
4264 dither_offset -= 4; \
4266 color_r += dither_offset; \
4267 color_g += dither_offset; \
4268 color_b += dither_offset; \
4289 #define draw_pixel_line_undithered(_x, _y) \
4292 #define draw_pixel_line_average() \
4293 color_r = (color_r + fb_r) / 2; \
4294 color_g = (color_g + fb_g) / 2; \
4295 color_b = (color_b + fb_b) / 2 \
4297 #define draw_pixel_line_add() \
4312 #define draw_pixel_line_subtract() \
4313 color_r = fb_r - color_r; \
4314 color_g = fb_g - color_g; \
4315 color_b = fb_b - color_b; \
4326 #define draw_pixel_line_add_fourth() \
4327 color_r = fb_r + (color_r / 4); \
4328 color_g = fb_g + (color_g / 4); \
4329 color_b = fb_b + (color_b / 4); \
4341 #define draw_pixel_line_blended(blend_mode) \
4342 s32 fb_pixel = *vram_ptr; \
4343 s32 fb_r = fb_pixel & 0x1F; \
4344 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4345 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4347 draw_pixel_line_##blend_mode() \
4349 #define draw_pixel_line_unblended(blend_mode) \
4352 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4354 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4355 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4357 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4359 draw_pixel_line_##shading(); \
4360 draw_pixel_line_##dithering(_x, _y); \
4366 draw_pixel_line_##blending(blend_mode); \
4368 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4369 psx_gpu->mask_msb; \
4373 #define update_increment(value) \
4376 #define update_decrement(value) \
4379 #define update_vram_row_increment(value) \
4382 #define update_vram_row_decrement(value) \
4385 #define compare_increment(a, b) \
4388 #define compare_decrement(a, b) \
4391 #define set_line_gradients(minor) \
4393 s32 gradient_divisor = delta_##minor; \
4394 if(gradient_divisor != 0) \
4396 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4397 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4398 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4406 current_r = fixed_center(vertex_a->r); \
4407 current_g = fixed_center(vertex_a->g); \
4408 current_b = fixed_center(vertex_a->b); \
4411 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4412 mask_evaluate, blend_mode) \
4415 error_step = delta_y * 2; \
4416 error_wrap = delta_x * 2; \
4420 set_line_gradients(x); \
4422 for(current_x = x_a; current_x <= x_b; current_x++) \
4424 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4425 mask_evaluate, blend_mode); \
4426 error += error_step; \
4429 if(error >= error_wrap) \
4431 update_##direction(current_y); \
4432 update_vram_row_##direction(); \
4433 error -= error_wrap; \
4438 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4439 mask_evaluate, blend_mode) \
4442 error_step = delta_x * 2; \
4443 error_wrap = delta_y * 2; \
4447 set_line_gradients(y); \
4449 for(current_y = y_a; compare_##direction(current_y, y_b); \
4450 update_##direction(current_y)) \
4452 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4453 mask_evaluate, blend_mode); \
4454 error += error_step; \
4455 update_vram_row_##direction(); \
4457 if(error > error_wrap) \
4461 error -= error_wrap; \
4467 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4473 if(delta_x > delta_y) \
4475 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4476 mask_evaluate, blend_mode); \
4480 draw_line_span_vertical(decrement, shading, blending, dithering, \
4481 mask_evaluate, blend_mode); \
4486 if(delta_x > delta_y) \
4488 draw_line_span_horizontal(increment, shading, blending, dithering, \
4489 mask_evaluate, blend_mode); \
4493 draw_line_span_vertical(increment, shading, blending, dithering, \
4494 mask_evaluate, blend_mode); \
4499 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4500 u32 color, int double_resolution)
4502 s32 color_r, color_g, color_b;
4503 u32 triangle_winding = 0;
4505 fixed_type gradient_r = 0;
4506 fixed_type gradient_g = 0;
4507 fixed_type gradient_b = 0;
4508 fixed_type current_r = 0;
4509 fixed_type current_g = 0;
4510 fixed_type current_b = 0;
4515 s32 delta_x, delta_y;
4526 flush_render_block_buffer(psx_gpu);
4527 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4529 vertex_struct *vertex_a = &(vertexes[0]);
4530 vertex_struct *vertex_b = &(vertexes[1]);
4538 if(vertex_a->x >= vertex_b->x)
4540 vertex_swap(vertex_a, vertex_b);
4549 delta_x = x_b - x_a;
4550 delta_y = y_b - y_a;
4552 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4555 if(double_resolution)
4565 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4567 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4571 if(flags & RENDER_FLAGS_SHADE)
4572 control_mask |= 0x1;
4574 if(flags & RENDER_FLAGS_BLEND)
4576 control_mask |= 0x2;
4577 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4580 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4581 control_mask |= 0x4;
4583 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4584 control_mask |= 0x8;
4586 switch(control_mask)
4589 render_line_body(unshaded, unblended, undithered, no, none);
4593 render_line_body(shaded, unblended, undithered, no, none);
4597 render_line_body(unshaded, blended, undithered, no, average);
4601 render_line_body(shaded, blended, undithered, no, average);
4605 render_line_body(unshaded, unblended, dithered, no, none);
4609 render_line_body(shaded, unblended, dithered, no, none);
4613 render_line_body(unshaded, blended, dithered, no, average);
4617 render_line_body(shaded, blended, dithered, no, average);
4621 render_line_body(unshaded, unblended, undithered, yes, none);
4625 render_line_body(shaded, unblended, undithered, yes, none);
4629 render_line_body(unshaded, blended, undithered, yes, average);
4633 render_line_body(shaded, blended, undithered, yes, average);
4637 render_line_body(unshaded, unblended, dithered, yes, none);
4641 render_line_body(shaded, unblended, dithered, yes, none);
4645 render_line_body(unshaded, blended, dithered, yes, average);
4649 render_line_body(shaded, blended, dithered, yes, average);
4653 render_line_body(unshaded, blended, undithered, no, add);
4657 render_line_body(shaded, blended, undithered, no, add);
4661 render_line_body(unshaded, blended, dithered, no, add);
4665 render_line_body(shaded, blended, dithered, no, add);
4669 render_line_body(unshaded, blended, undithered, yes, add);
4673 render_line_body(shaded, blended, undithered, yes, add);
4677 render_line_body(unshaded, blended, dithered, yes, add);
4681 render_line_body(shaded, blended, dithered, yes, add);
4685 render_line_body(unshaded, blended, undithered, no, subtract);
4689 render_line_body(shaded, blended, undithered, no, subtract);
4693 render_line_body(unshaded, blended, dithered, no, subtract);
4697 render_line_body(shaded, blended, dithered, no, subtract);
4701 render_line_body(unshaded, blended, undithered, yes, subtract);
4705 render_line_body(shaded, blended, undithered, yes, subtract);
4709 render_line_body(unshaded, blended, dithered, yes, subtract);
4713 render_line_body(shaded, blended, dithered, yes, subtract);
4717 render_line_body(unshaded, blended, undithered, no, add_fourth);
4721 render_line_body(shaded, blended, undithered, no, add_fourth);
4725 render_line_body(unshaded, blended, dithered, no, add_fourth);
4729 render_line_body(shaded, blended, dithered, no, add_fourth);
4733 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4737 render_line_body(shaded, blended, undithered, yes, add_fourth);
4741 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4745 render_line_body(shaded, blended, dithered, yes, add_fourth);
4751 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4752 u32 width, u32 height)
4754 if((width == 0) || (height == 0))
4757 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4759 u32 r = color & 0xFF;
4760 u32 g = (color >> 8) & 0xFF;
4761 u32 b = (color >> 16) & 0xFF;
4762 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4764 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4766 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4768 u32 pitch = 512 - (width / 2);
4771 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4776 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4785 vram_ptr[0] = color_32bpp;
4786 vram_ptr[1] = color_32bpp;
4787 vram_ptr[2] = color_32bpp;
4788 vram_ptr[3] = color_32bpp;
4789 vram_ptr[4] = color_32bpp;
4790 vram_ptr[5] = color_32bpp;
4791 vram_ptr[6] = color_32bpp;
4792 vram_ptr[7] = color_32bpp;
4803 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4804 u32 width, u32 height)
4806 if((width == 0) || (height == 0))
4812 u32 r = color & 0xFF;
4813 u32 g = (color >> 8) & 0xFF;
4814 u32 b = (color >> 16) & 0xFF;
4815 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4817 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4819 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4821 u32 pitch = 1024 / 2 - (width / 2);
4829 vram_ptr[0] = color_32bpp;
4830 vram_ptr[1] = color_32bpp;
4831 vram_ptr[2] = color_32bpp;
4832 vram_ptr[3] = color_32bpp;
4833 vram_ptr[4] = color_32bpp;
4834 vram_ptr[5] = color_32bpp;
4835 vram_ptr[6] = color_32bpp;
4836 vram_ptr[7] = color_32bpp;
4847 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4848 u32 width, u32 height, u32 pitch)
4850 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4852 u32 mask_msb = psx_gpu->mask_msb;
4854 if((width == 0) || (height == 0))
4857 flush_render_block_buffer(psx_gpu);
4858 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4860 for(draw_y = 0; draw_y < height; draw_y++)
4862 for(draw_x = 0; draw_x < width; draw_x++)
4864 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4872 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4873 u32 dest_x, u32 dest_y, u32 width, u32 height)
4875 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4876 dest_x, dest_y, width, height, 1024);
4880 void initialize_reciprocal_table(void)
4883 u32 height_normalized;
4884 u32 height_reciprocal;
4887 for(height = 1; height < sizeof(reciprocal_table)
4888 / sizeof(reciprocal_table[0]); height++)
4890 shift = __builtin_clz(height);
4891 height_normalized = height << shift;
4892 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4895 shift = 32 - (51 - shift);
4897 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4902 #define dither_table_row(a, b, c, d) \
4903 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4905 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4907 vec_8x16u test_mask =
4908 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4910 psx_gpu->test_mask = test_mask;
4912 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4913 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4914 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4915 psx_gpu->viewport_mask = 0;
4916 psx_gpu->current_texture_page = 0;
4917 psx_gpu->current_texture_mask = 0;
4918 psx_gpu->last_8bpp_texture_page = 0;
4920 psx_gpu->clut_settings = 0;
4921 psx_gpu->texture_settings = 0;
4922 psx_gpu->render_state = 0;
4923 psx_gpu->render_state_base = 0;
4924 psx_gpu->num_blocks = 0;
4925 psx_gpu->uvrgb_phase = 0x8000;
4927 psx_gpu->vram_ptr = vram;
4928 psx_gpu->vram_out_ptr = vram;
4930 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4931 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4932 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4934 psx_gpu->mask_msb = 0;
4936 psx_gpu->texture_window_x = 0;
4937 psx_gpu->texture_window_y = 0;
4938 psx_gpu->texture_mask_width = 0xFF;
4939 psx_gpu->texture_mask_height = 0xFF;
4941 psx_gpu->render_mode = 0;
4943 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4945 initialize_reciprocal_table();
4946 psx_gpu->reciprocal_table_ptr = reciprocal_table;
4953 // (minus ones(4) * 4)
4955 // d0: (1 3 5 7): x1 ^ y1
4956 // d1: (2 3 6 7): y0
4957 // d2: (4 5 6 7): x0 ^ y0
4959 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
4960 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
4961 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
4962 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
4964 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
4966 psx_gpu->enhancement_x_threshold = 256;
4972 gettimeofday(&tv, NULL);
4974 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
4977 #if 0 //def NEON_BUILD
4982 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
4987 void init_counter(void)
4990 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
4991 value |= 5; // master enable, ccnt reset
4992 value &= ~8; // ccnt divider 0
4993 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
4994 // enable cycle counter
4995 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
4998 void triangle_benchmark(psx_gpu_struct *psx_gpu)
5005 const u32 iterations = 500000;
5007 psx_gpu->num_blocks = 64;
5008 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5010 for(i = 0; i < 64; i++)
5012 memset(&(psx_gpu->blocks[i].r), 0, 16);
5017 ticks = get_counter();
5019 for(i = 0; i < iterations; i++)
5021 texture_sprite_blocks_8bpp(psx_gpu);
5024 ticks_elapsed = get_counter() - ticks;
5026 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5031 #include "psx_gpu_4x.c"