2 * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of
7 * the License, or (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
21 #include "vector_ops.h"
23 #include "psx_gpu_simd.h"
26 void dump_r_d(const char *name, void *dump);
27 void dump_r_q(const char *name, void *dump);
28 #define dumprd(n) dump_r_d(#n, n.e)
29 #define dumprq(n) dump_r_q(#n, n.e)
33 u32 span_pixel_blocks = 0;
39 u32 sprites_16bpp = 0;
40 u32 sprite_blocks = 0;
41 u32 sprites_untextured = 0;
43 u32 trivial_rejects = 0;
47 u32 texel_blocks_4bpp = 0;
48 u32 texel_blocks_8bpp = 0;
49 u32 texel_blocks_16bpp = 0;
50 u32 texel_blocks_untextured = 0;
52 u32 render_buffer_flushes = 0;
53 u32 state_changes = 0;
54 u32 left_split_triangles = 0;
55 u32 flat_triangles = 0;
56 u32 clipped_triangles = 0;
57 u32 zero_block_spans = 0;
58 u32 texture_cache_loads = 0;
59 u32 false_modulated_blocks = 0;
61 #define stats_add(stat, count) // stat += count
63 /* double size for enhancement */
64 u32 reciprocal_table[512 * 2];
67 typedef s32 fixed_type;
69 #define EDGE_STEP_BITS 32
72 #define fixed_center(value) \
73 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
75 #define int_to_fixed(value) \
76 (((fixed_type)(value)) << FIXED_BITS) \
78 #define fixed_to_int(value) \
79 ((value) >> FIXED_BITS) \
81 #define fixed_to_double(value) \
82 ((value) / (double)(1 << FIXED_BITS)) \
84 #define double_to_fixed(value) \
85 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
87 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
88 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
89 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
90 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
92 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
93 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
95 struct render_block_handler_struct
98 texture_blocks_function_type *texture_blocks;
99 shade_blocks_function_type *shade_blocks;
100 blend_blocks_function_type *blend_blocks;
105 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
107 u32 shift = __builtin_clz(denominator);
108 u32 denominator_normalized = denominator << shift;
110 double numerator = (1ULL << 62) + denominator_normalized;
113 double denominator_normalized_dp_b;
114 u64 denominator_normalized_dp_u64;
117 double reciprocal_dp;
119 u64 numerator_u64 = (denominator_normalized >> 10) |
120 ((u64)(62 + 1023) << 52);
121 *((u64 *)(&numerator_b)) = numerator_u64;
123 denominator_normalized_dp_u64 =
124 (u64)(denominator_normalized << 21) |
125 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
126 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
128 // Implement with a DP divide
129 reciprocal_dp = numerator / denominator_normalized_dp_b;
130 reciprocal = reciprocal_dp;
132 if(reciprocal == 0x80000001)
133 reciprocal = 0x80000000;
135 *_shift = 62 - shift;
139 double reciprocal_estimate(double a)
144 q = (int)(a * 512.0);
145 /* a in units of 1/512 rounded down */
146 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
147 s = (int)(256.0 * r + 0.5);
149 /* r in units of 1/256 rounded to nearest */
151 return (double)s / 256.0;
154 u32 reciprocal_estimate_u32(u32 value)
157 volatile double dp_value;
158 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
160 if((value >> 31) == 0)
163 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
165 *dp_value_ptr = dp_value_u64;
167 dp_value = reciprocal_estimate(dp_value);
168 dp_value_u64 = *dp_value_ptr;
170 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
173 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
175 u32 shift = __builtin_clz(value);
176 u32 value_normalized = value << shift;
178 *_shift = 62 - shift;
180 value_normalized -= 2;
182 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
184 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
185 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
186 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
187 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
188 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
189 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
191 return reciprocal_normalized;
197 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
199 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
202 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
204 s32 coverage_x, coverage_y;
209 coverage_x = x2 >> 6;
210 coverage_y = y2 >> 8;
218 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
221 mask_down_right |= mask_down_right << 16;
223 coverage_x = x1 >> 6;
225 mask_up_left = 0xFFFF0000 << coverage_x;
227 mask_up_left = 0xFFFF0000;
229 coverage_y = y1 >> 8;
231 mask_up_left |= mask_up_left >> 16;
233 return mask_up_left & mask_down_right;
236 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
239 u32 mask = texture_region_mask(x1, y1, x2, y2);
241 psx_gpu->dirty_textures_4bpp_mask |= mask;
242 psx_gpu->dirty_textures_8bpp_mask |= mask;
243 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
248 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
249 u32 y1, u32 x2, u32 y2)
251 u32 mask = texture_region_mask(x1, y1, x2, y2) &
252 psx_gpu->viewport_mask;
254 psx_gpu->dirty_textures_4bpp_mask |= mask;
255 psx_gpu->dirty_textures_8bpp_mask |= mask;
256 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
261 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
264 u32 mask = texture_region_mask(x1, y1, x2, y2);
266 u8 *texture_page_ptr;
271 psx_gpu->dirty_textures_8bpp_mask |= mask;
272 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
274 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
275 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
277 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
278 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
279 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
280 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
288 texel_block = *vram_ptr;
290 texture_page_ptr[0] = texel_block & 0xF;
291 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
292 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
293 texture_page_ptr[3] = texel_block >> 12;
296 texture_page_ptr += 4;
310 psx_gpu->dirty_textures_4bpp_mask |= mask;
316 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
318 u32 current_texture_page = psx_gpu->current_texture_page;
319 u8 *texture_page_ptr = psx_gpu->texture_page_base;
320 u16 *vram_ptr = psx_gpu->vram_ptr;
326 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
327 vram_ptr += (current_texture_page & 0xF) * 64;
329 texture_cache_loads++;
336 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
346 texel_block = *vram_ptr;
348 texture_page_ptr[0] = texel_block & 0xF;
349 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
350 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
351 texture_page_ptr[3] = texel_block >> 12;
354 texture_page_ptr += 4;
368 vram_ptr -= (1024 * 16) - 4;
374 vram_ptr += (16 * 1024) - (4 * 16);
379 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
382 u16 *texture_page_ptr = psx_gpu->texture_page_base;
383 u16 *vram_ptr = psx_gpu->vram_ptr;
390 texture_cache_loads++;
392 vram_ptr += (texture_page >> 4) * 256 * 1024;
393 vram_ptr += (texture_page & 0xF) * 64;
395 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
396 texture_page_ptr += (8 * 16) * 8;
409 load_128b(texels, vram_ptr);
410 store_128b(texels, texture_page_ptr);
412 texture_page_ptr += 8;
420 vram_ptr -= (1024 * 16);
429 vram_ptr += (16 * 1024);
431 texture_page_ptr += (8 * 16) * 8;
439 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
441 u32 current_texture_page = psx_gpu->current_texture_page;
442 u32 update_textures =
443 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
445 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
447 if(update_textures & (1 << current_texture_page))
449 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
450 update_textures &= ~(1 << current_texture_page);
455 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
456 (current_texture_page & 0x10);
458 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
462 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
464 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
465 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
467 u32 num_blocks_dest = 0;
468 block_struct *block_src = psx_gpu->blocks;
469 block_struct *block_dest = psx_gpu->blocks;
471 u16 *vram_ptr = psx_gpu->vram_ptr;
474 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
476 for(i = 0; i < psx_gpu->num_blocks; i++)
478 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
479 if(fb_offset & (1 << 11))
481 *block_dest = *block_src;
490 for(i = 0; i < psx_gpu->num_blocks; i++)
492 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
493 if((fb_offset & (1 << 11)) == 0)
495 *block_dest = *block_src;
503 psx_gpu->num_blocks = num_blocks_dest;
506 if(psx_gpu->num_blocks)
508 render_block_handler_struct *render_block_handler =
509 psx_gpu->render_block_handler;
511 render_block_handler->texture_blocks(psx_gpu);
512 render_block_handler->shade_blocks(psx_gpu);
513 render_block_handler->blend_blocks(psx_gpu);
516 span_pixel_blocks += psx_gpu->num_blocks;
517 render_buffer_flushes++;
520 psx_gpu->num_blocks = 0;
527 #define setup_gradient_calculation_input(set, vertex) \
528 /* First type is: uvrg bxxx xxxx */\
529 /* Second type is: yyyy ybyy uvrg */\
530 /* Since x_a and y_c are the same the same variable is used for both. */\
531 x##set##_a_y##set##_c.e[0] = vertex->u; \
532 x##set##_a_y##set##_c.e[1] = vertex->v; \
533 x##set##_a_y##set##_c.e[2] = vertex->r; \
534 x##set##_a_y##set##_c.e[3] = vertex->g; \
535 dup_4x16b(x##set##_b, vertex->x); \
536 dup_4x16b(x##set##_c, vertex->x); \
537 dup_4x16b(y##set##_a, vertex->y); \
538 dup_4x16b(y##set##_b, vertex->y); \
539 x##set##_b.e[0] = vertex->b; \
540 y##set##_b.e[1] = vertex->b \
543 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
544 vertex_struct *b, vertex_struct *c)
546 u32 triangle_area = psx_gpu->triangle_area;
547 u32 winding_mask_scalar;
549 u32 triangle_area_shift;
550 u64 triangle_area_reciprocal =
551 fixed_reciprocal(triangle_area, &triangle_area_shift);
552 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
554 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
555 // ( d0 * d1 ) - ( d2 * d3 ) =
556 // ( m0 ) - ( m1 ) = gradient
558 // This is split to do 12 elements at a time over three sets: a, b, and c.
559 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
560 // two of the slots are unused.
562 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
565 vec_4x16s x0_a_y0_c, x0_b, x0_c;
566 vec_4x16s y0_a, y0_b;
567 vec_4x16s x1_a_y1_c, x1_b, x1_c;
568 vec_4x16s y1_a, y1_b;
569 vec_4x16s x2_a_y2_c, x2_b, x2_c;
570 vec_4x16s y2_a, y2_b;
574 vec_4x32u uvrgb_phase;
576 vec_4x16s d0_a_d3_c, d0_b, d0_c;
577 vec_4x16s d1_a, d1_b, d1_c_d2_a;
578 vec_4x16s d2_b, d2_c;
579 vec_4x16s d3_a, d3_b;
581 vec_4x32s m0_a, m0_b, m0_c;
582 vec_4x32s m1_a, m1_b, m1_c;
584 vec_4x32u gradient_area_a, gradient_area_c;
585 vec_2x32u gradient_area_b;
587 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
588 vec_2x32u gradient_area_sign_b;
589 vec_4x32u winding_mask;
591 vec_2x64u gradient_wide_a0, gradient_wide_a1;
592 vec_2x64u gradient_wide_c0, gradient_wide_c1;
593 vec_2x64u gradient_wide_b;
595 vec_4x32u gradient_a, gradient_c;
596 vec_2x32u gradient_b;
597 vec_16x8s gradient_shift;
599 setup_gradient_calculation_input(0, a);
600 setup_gradient_calculation_input(1, b);
601 setup_gradient_calculation_input(2, c);
603 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
604 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
605 shl_long_4x16b(b_base, x0_b, 16);
607 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
608 add_4x32b(b_base, b_base, uvrgb_phase);
610 // Can probably pair these, but it'll require careful register allocation
611 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
612 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
614 sub_4x16b(d0_b, x1_b, x0_b);
615 sub_4x16b(d0_c, x1_c, x0_c);
617 sub_4x16b(d1_a, y2_a, y1_a);
618 sub_4x16b(d1_b, y2_b, y1_b);
620 sub_4x16b(d2_b, x2_b, x1_b);
621 sub_4x16b(d2_c, x2_c, x1_c);
623 sub_4x16b(d3_a, y1_a, y0_a);
624 sub_4x16b(d3_b, y1_b, y0_b);
626 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
627 mul_long_4x16b(m0_b, d0_b, d1_b);
628 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
630 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
631 mul_long_4x16b(m1_b, d2_b, d3_b);
632 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
634 sub_4x32b(gradient_area_a, m0_a, m1_a);
635 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
636 sub_4x32b(gradient_area_c, m0_c, m1_c);
638 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
639 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
640 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
642 abs_4x32b(gradient_area_a, gradient_area_a);
643 abs_2x32b(gradient_area_b, gradient_area_b);
644 abs_4x32b(gradient_area_c, gradient_area_c);
646 winding_mask_scalar = -psx_gpu->triangle_winding;
648 dup_4x32b(winding_mask, winding_mask_scalar);
649 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
650 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
651 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
653 mul_scalar_long_2x32b(gradient_wide_a0,
654 vector_cast(vec_2x32s, gradient_area_a.low),
655 (s64)triangle_area_reciprocal);
656 mul_scalar_long_2x32b(gradient_wide_a1,
657 vector_cast(vec_2x32s, gradient_area_a.high),
658 (s64)triangle_area_reciprocal);
659 mul_scalar_long_2x32b(gradient_wide_b,
660 vector_cast(vec_2x32s, gradient_area_b),
661 (s64)triangle_area_reciprocal);
662 mul_scalar_long_2x32b(gradient_wide_c0,
663 vector_cast(vec_2x32s, gradient_area_c.low),
664 (s64)triangle_area_reciprocal);
665 mul_scalar_long_2x32b(gradient_wide_c1,
666 vector_cast(vec_2x32s, gradient_area_c.high),
667 (s64)triangle_area_reciprocal);
669 dup_16x8b(gradient_shift, triangle_area_shift);
670 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
671 vector_cast(vec_2x64u, gradient_shift));
672 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
673 vector_cast(vec_2x64u, gradient_shift));
674 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
675 vector_cast(vec_2x64u, gradient_shift));
676 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
677 vector_cast(vec_2x64u, gradient_shift));
678 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
679 vector_cast(vec_2x64u, gradient_shift));
681 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
682 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
683 mov_narrow_2x64b(gradient_b, gradient_wide_b);
684 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
685 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
687 shl_4x32b(gradient_a, gradient_a, 4);
688 shl_2x32b(gradient_b, gradient_b, 4);
689 shl_4x32b(gradient_c, gradient_c, 4);
691 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
692 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
693 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
695 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
696 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
697 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
699 u32 left_adjust = a->x;
700 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
701 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
711 eor_4x32b(zero, zero, zero);
712 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
713 add_2x32b(b_dx2, gradient_b, gradient_b);
714 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
715 add_2x32b(b_dx3, gradient_b, b_dx2);
717 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
719 psx_gpu->u_block_span.e[0] = zero.e[0];
720 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
721 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
722 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
724 psx_gpu->v_block_span.e[0] = zero.e[1];
725 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
726 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
727 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
729 psx_gpu->r_block_span.e[0] = zero.e[2];
730 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
731 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
732 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
734 psx_gpu->g_block_span.e[0] = zero.e[3];
735 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
736 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
737 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
739 psx_gpu->b_block_span.e[0] = zero.e[0];
740 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
741 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
742 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
744 psx_gpu->uvrg = uvrg_base;
745 psx_gpu->b = b_base.e[0];
747 psx_gpu->uvrg_dx = gradient_a;
748 psx_gpu->uvrg_dy = gradient_c;
749 psx_gpu->b_dy = gradient_b.e[1];
753 #define vector_check(_a, _b) \
754 if(memcmp(&_a, &_b, sizeof(_b))) \
756 if(sizeof(_b) == 8) \
758 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
759 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
763 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
764 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
769 #define scalar_check(_a, _b) \
771 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
775 #define setup_spans_debug_check(span_edge_data_element) \
777 u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data; \
778 if (_num_spans > MAX_SPANS) \
779 *(volatile int *)0 = 1; \
780 if (_num_spans < psx_gpu->num_spans) \
782 if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW) \
783 *(volatile int *)0 = 2; \
784 if(span_edge_data_element.y >= 2048) \
785 *(volatile int *)0 = 3; \
790 #define setup_spans_debug_check(span_edge_data_element) \
794 #define setup_spans_prologue_alternate_yes() \
795 vec_2x64s alternate_x; \
796 vec_2x64s alternate_dx_dy; \
797 vec_4x32s alternate_x_32; \
798 vec_4x16u alternate_x_16; \
800 vec_4x16u alternate_select; \
801 vec_4x16s y_mid_point; \
805 s32 edge_dx_dy_alt; \
808 #define setup_spans_prologue_alternate_no() \
810 #define setup_spans_prologue(alternate_active) \
811 edge_data_struct *span_edge_data; \
812 vec_4x32u *span_uvrg_offset; \
813 u32 *span_b_offset; \
817 vec_2x64s edges_xy; \
818 vec_2x32s edges_dx_dy; \
819 vec_2x32u edge_shifts; \
821 vec_2x64s left_x, right_x; \
822 vec_2x64s left_dx_dy, right_dx_dy; \
823 vec_4x32s left_x_32, right_x_32; \
824 vec_8x16s left_right_x_16; \
826 vec_8x16s left_edge; \
827 vec_8x16s right_edge; \
828 vec_4x16u span_shift; \
832 vec_4x16u c_0xFFFE; \
835 vec_2x32s x_starts; \
844 vec_4x32u uvrg = psx_gpu->uvrg; \
845 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
846 u32 b = psx_gpu->b; \
847 u32 b_dy = psx_gpu->b_dy; \
849 dup_2x32b(c_0x01, 0x01); \
850 setup_spans_prologue_alternate_##alternate_active() \
852 #define setup_spans_prologue_b() \
853 span_edge_data = psx_gpu->span_edge_data; \
854 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
855 span_b_offset = psx_gpu->span_b_offset; \
857 vec_8x16u c_0x0001; \
859 dup_8x16b(c_0x0001, 0x0001); \
860 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
861 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
862 add_8x16b(right_edge, right_edge, c_0x0001); \
863 dup_4x16b(c_0x04, 0x04); \
864 dup_4x16b(c_0x07, 0x07); \
865 dup_4x16b(c_0xFFFE, 0xFFFE); \
868 #define compute_edge_delta_x2() \
871 vec_2x32s height_reciprocals; \
872 vec_2x32s heights_b; \
875 u32 edge_shift = reciprocal_table[height]; \
877 dup_2x32b(heights, height); \
878 sub_2x32b(widths, x_ends, x_starts); \
880 dup_2x32b(edge_shifts, edge_shift); \
881 sub_2x32b(heights_b, heights, c_0x01); \
882 shr_2x32b(height_reciprocals, edge_shifts, 10); \
884 mla_2x32b(heights_b, x_starts, heights); \
885 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
886 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
887 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
890 #define compute_edge_delta_x3(start_c, height_a, height_b) \
893 vec_2x32s height_reciprocals; \
894 vec_2x32s heights_b; \
899 u32 height_reciprocal_alt; \
901 heights.e[0] = height_a; \
902 heights.e[1] = height_b; \
904 edge_shifts.e[0] = reciprocal_table[height_a]; \
905 edge_shifts.e[1] = reciprocal_table[height_b]; \
906 edge_shift_alt = reciprocal_table[height_minor_b]; \
908 sub_2x32b(widths, x_ends, x_starts); \
909 width_alt = x_c - start_c; \
911 shr_2x32b(height_reciprocals, edge_shifts, 10); \
912 height_reciprocal_alt = edge_shift_alt >> 10; \
914 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
915 edge_shift_alt &= 0x1F; \
917 sub_2x32b(heights_b, heights, c_0x01); \
918 height_b_alt = height_minor_b - 1; \
920 mla_2x32b(heights_b, x_starts, heights); \
921 height_b_alt += height_minor_b * start_c; \
923 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
924 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
926 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
927 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
931 #define setup_spans_adjust_y_up() \
932 sub_4x32b(y_x4, y_x4, c_0x04) \
934 #define setup_spans_adjust_y_down() \
935 add_4x32b(y_x4, y_x4, c_0x04) \
937 #define setup_spans_adjust_interpolants_up() \
938 sub_4x32b(uvrg, uvrg, uvrg_dy); \
941 #define setup_spans_adjust_interpolants_down() \
942 add_4x32b(uvrg, uvrg, uvrg_dy); \
946 #define setup_spans_clip_interpolants_increment() \
947 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
950 #define setup_spans_clip_interpolants_decrement() \
951 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
954 #define setup_spans_clip_alternate_yes() \
955 edge_alt += edge_dx_dy_alt * (s64)(clip) \
957 #define setup_spans_clip_alternate_no() \
959 #define setup_spans_clip(direction, alternate_active) \
961 clipped_triangles++; \
962 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
963 setup_spans_clip_alternate_##alternate_active(); \
964 setup_spans_clip_interpolants_##direction(); \
968 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
970 vec_2x64u edge_shifts_64; \
971 vec_2x64s edges_dx_dy_64; \
973 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
974 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
976 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
977 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
979 left_x.e[0] = edges_xy.e[left_index]; \
980 right_x.e[0] = edges_xy.e[right_index]; \
982 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
983 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
984 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
985 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
987 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
988 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
990 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
991 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
994 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
996 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
997 s64 edge_dx_dy_alt_64; \
999 dup_4x16b(y_mid_point, y_b); \
1001 edge_alt <<= edge_shift_alt; \
1002 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
1004 alternate_x.e[0] = edge_alt; \
1005 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
1006 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1008 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1009 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1013 #define setup_spans_y_select_up() \
1014 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1016 #define setup_spans_y_select_down() \
1017 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1019 #define setup_spans_y_select_alternate_yes(direction) \
1020 setup_spans_y_select_##direction() \
1022 #define setup_spans_y_select_alternate_no(direction) \
1024 #define setup_spans_alternate_select_left() \
1025 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1027 #define setup_spans_alternate_select_right() \
1028 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1030 #define setup_spans_alternate_select_none() \
1032 #define setup_spans_increment_alternate_yes() \
1033 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1034 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1035 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1036 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1037 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1039 #define setup_spans_increment_alternate_no() \
1041 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1043 span_uvrg_offset[0] = uvrg; \
1044 span_b_offset[0] = b; \
1045 setup_spans_adjust_interpolants_##direction(); \
1047 span_uvrg_offset[1] = uvrg; \
1048 span_b_offset[1] = b; \
1049 setup_spans_adjust_interpolants_##direction(); \
1051 span_uvrg_offset[2] = uvrg; \
1052 span_b_offset[2] = b; \
1053 setup_spans_adjust_interpolants_##direction(); \
1055 span_uvrg_offset[3] = uvrg; \
1056 span_b_offset[3] = b; \
1057 setup_spans_adjust_interpolants_##direction(); \
1059 span_uvrg_offset += 4; \
1060 span_b_offset += 4; \
1062 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1063 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1065 add_2x64b(left_x, left_x, left_dx_dy); \
1066 add_2x64b(right_x, right_x, right_dx_dy); \
1068 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1069 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1071 add_2x64b(left_x, left_x, left_dx_dy); \
1072 add_2x64b(right_x, right_x, right_dx_dy); \
1074 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1075 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1077 setup_spans_increment_alternate_##alternate_active(); \
1078 setup_spans_y_select_alternate_##alternate_active(direction); \
1079 setup_spans_alternate_select_##alternate(); \
1081 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1082 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1084 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1085 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1086 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1087 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1088 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1091 for(i = 0; i < 4; i++) \
1093 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1094 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1095 span_edge_data[i].right_mask = span_shift.e[i]; \
1096 span_edge_data[i].y = y_x4.e[i]; \
1097 setup_spans_debug_check(span_edge_data[i]); \
1100 span_edge_data += 4; \
1102 setup_spans_adjust_y_##direction(); \
1106 #define setup_spans_alternate_adjust_yes() \
1107 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1109 #define setup_spans_alternate_adjust_no() \
1112 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1113 setup_spans_alternate_adjust_##alternate_active(); \
1114 if(y_c > psx_gpu->viewport_end_y) \
1115 height -= y_c - psx_gpu->viewport_end_y - 1; \
1117 clip = psx_gpu->viewport_start_y - y_a; \
1122 setup_spans_clip(increment, alternate_active); \
1125 setup_spans_prologue_b(); \
1130 y_x4.e[1] = y_a + 1; \
1131 y_x4.e[2] = y_a + 2; \
1132 y_x4.e[3] = y_a + 3; \
1133 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1136 psx_gpu->num_spans = height; \
1139 setup_spans_set_x4(alternate, down, alternate_active); \
1141 } while(height > 0); \
1145 #define setup_spans_alternate_pre_increment_yes() \
1146 edge_alt += edge_dx_dy_alt \
1148 #define setup_spans_alternate_pre_increment_no() \
1150 #define setup_spans_up_decrement_height_yes() \
1153 #define setup_spans_up_decrement_height_no() \
1156 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1157 setup_spans_alternate_adjust_##alternate_active(); \
1160 if(y_c < psx_gpu->viewport_start_y) \
1161 height -= psx_gpu->viewport_start_y - y_c; \
1163 setup_spans_up_decrement_height_##alternate_active(); \
1165 clip = y_a - psx_gpu->viewport_end_y; \
1170 setup_spans_clip(decrement, alternate_active); \
1173 setup_spans_prologue_b(); \
1178 y_x4.e[1] = y_a - 1; \
1179 y_x4.e[2] = y_a - 2; \
1180 y_x4.e[3] = y_a - 3; \
1181 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1182 setup_spans_alternate_pre_increment_##alternate_active(); \
1183 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1185 setup_spans_adjust_interpolants_up(); \
1187 psx_gpu->num_spans = height; \
1190 setup_spans_set_x4(alternate, up, alternate_active); \
1195 #define index_left 0
1196 #define index_right 1
1198 #define setup_spans_up_up(minor, major) \
1199 setup_spans_prologue(yes); \
1200 s32 height_minor_a = y_a - y_b; \
1201 s32 height_minor_b = y_b - y_c; \
1202 s32 height = y_a - y_c; \
1204 dup_2x32b(x_starts, x_a); \
1205 x_ends.e[0] = x_c; \
1206 x_ends.e[1] = x_b; \
1208 compute_edge_delta_x3(x_b, height, height_minor_a); \
1209 setup_spans_up(index_##major, index_##minor, minor, yes) \
1214 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1215 vertex_struct *v_b, vertex_struct *v_c)
1217 setup_spans_up_up(left, right);
1220 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1221 vertex_struct *v_b, vertex_struct *v_c)
1223 setup_spans_up_up(right, left);
1226 #define setup_spans_down_down(minor, major) \
1227 setup_spans_prologue(yes); \
1228 s32 height_minor_a = y_b - y_a; \
1229 s32 height_minor_b = y_c - y_b; \
1230 s32 height = y_c - y_a; \
1232 dup_2x32b(x_starts, x_a); \
1233 x_ends.e[0] = x_c; \
1234 x_ends.e[1] = x_b; \
1236 compute_edge_delta_x3(x_b, height, height_minor_a); \
1237 setup_spans_down(index_##major, index_##minor, minor, yes) \
1239 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1240 vertex_struct *v_b, vertex_struct *v_c)
1242 setup_spans_down_down(left, right);
1245 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1246 vertex_struct *v_b, vertex_struct *v_c)
1248 setup_spans_down_down(right, left);
1251 #define setup_spans_up_flat() \
1252 s32 height = y_a - y_c; \
1255 compute_edge_delta_x2(); \
1256 setup_spans_up(index_left, index_right, none, no) \
1258 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1259 vertex_struct *v_b, vertex_struct *v_c)
1261 setup_spans_prologue(no);
1262 x_starts.e[0] = x_a;
1263 x_starts.e[1] = x_b;
1264 dup_2x32b(x_ends, x_c);
1266 setup_spans_up_flat();
1269 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1270 vertex_struct *v_b, vertex_struct *v_c)
1272 setup_spans_prologue(no);
1273 dup_2x32b(x_starts, x_a);
1277 setup_spans_up_flat();
1280 #define setup_spans_down_flat() \
1281 s32 height = y_c - y_a; \
1284 compute_edge_delta_x2(); \
1285 setup_spans_down(index_left, index_right, none, no) \
1287 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1288 vertex_struct *v_b, vertex_struct *v_c)
1290 setup_spans_prologue(no);
1291 x_starts.e[0] = x_a;
1292 x_starts.e[1] = x_b;
1293 dup_2x32b(x_ends, x_c);
1295 setup_spans_down_flat();
1298 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1299 vertex_struct *v_b, vertex_struct *v_c)
1301 setup_spans_prologue(no);
1302 dup_2x32b(x_starts, x_a);
1306 setup_spans_down_flat();
1309 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1310 vertex_struct *v_b, vertex_struct *v_c)
1312 setup_spans_prologue(no);
1320 s32 height_minor_a = y_a - y_b;
1321 s32 height_minor_b = y_c - y_a;
1322 s32 height_major = y_c - y_b;
1324 vec_2x64s edges_xy_b;
1325 vec_2x32s edges_dx_dy_b;
1326 vec_2x32u edge_shifts_b;
1328 vec_2x32s height_increment;
1330 x_starts.e[0] = x_a;
1331 x_starts.e[1] = x_c;
1332 dup_2x32b(x_ends, x_b);
1334 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1336 height_increment.e[0] = 0;
1337 height_increment.e[1] = height_minor_b;
1339 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1341 edges_xy_b.e[0] = edge_alt;
1342 edges_xy_b.e[1] = edges_xy.e[1];
1344 edge_shifts_b = edge_shifts;
1345 edge_shifts_b.e[0] = edge_shift_alt;
1347 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1348 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1352 if(y_b < psx_gpu->viewport_start_y)
1353 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1355 clip = y_a - psx_gpu->viewport_end_y;
1358 height_minor_a -= clip;
1360 setup_spans_clip(decrement, no);
1363 setup_spans_prologue_b();
1365 if(height_minor_a > 0)
1368 y_x4.e[1] = y_a - 1;
1369 y_x4.e[2] = y_a - 2;
1370 y_x4.e[3] = y_a - 3;
1371 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1372 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1373 setup_spans_adjust_interpolants_up();
1375 psx_gpu->num_spans = height_minor_a;
1376 while(height_minor_a > 0)
1378 setup_spans_set_x4(none, up, no);
1379 height_minor_a -= 4;
1382 span_edge_data += height_minor_a;
1383 span_uvrg_offset += height_minor_a;
1384 span_b_offset += height_minor_a;
1387 edges_xy = edges_xy_b;
1388 edges_dx_dy = edges_dx_dy_b;
1389 edge_shifts = edge_shifts_b;
1391 uvrg = psx_gpu->uvrg;
1396 if(y_c > psx_gpu->viewport_end_y)
1397 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1399 clip = psx_gpu->viewport_start_y - y_a;
1402 height_minor_b -= clip;
1404 setup_spans_clip(increment, no);
1407 if(height_minor_b > 0)
1410 y_x4.e[1] = y_a + 1;
1411 y_x4.e[2] = y_a + 2;
1412 y_x4.e[3] = y_a + 3;
1413 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1415 // FIXME: overflow corner case
1416 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1417 height_minor_b &= ~3;
1419 psx_gpu->num_spans += height_minor_b;
1420 while(height_minor_b > 0)
1422 setup_spans_set_x4(none, down, no);
1423 height_minor_b -= 4;
1427 left_split_triangles++;
1433 #define dither_table_entry_normal(value) \
1437 #define setup_blocks_load_msb_mask_indirect() \
1439 #define setup_blocks_load_msb_mask_direct() \
1440 vec_8x16u msb_mask; \
1441 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1444 #define setup_blocks_variables_shaded_textured(target) \
1445 vec_4x32u u_block; \
1446 vec_4x32u v_block; \
1447 vec_4x32u r_block; \
1448 vec_4x32u g_block; \
1449 vec_4x32u b_block; \
1450 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1451 vec_4x32u uvrg_dx4; \
1452 vec_4x32u uvrg_dx8; \
1454 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1455 u32 b_dx4 = b_dx << 2; \
1456 u32 b_dx8 = b_dx << 3; \
1459 vec_16x8u texture_mask; \
1460 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1461 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1462 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1463 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1465 #define setup_blocks_variables_shaded_untextured(target) \
1466 vec_4x32u r_block; \
1467 vec_4x32u g_block; \
1468 vec_4x32u b_block; \
1470 vec_4x32u rgb_dx4; \
1471 vec_4x32u rgb_dx8; \
1474 vec_8x8u d64_0x07; \
1479 dup_8x8b(d64_0x07, 0x07); \
1480 dup_8x8b(d64_1, 1); \
1481 dup_8x8b(d64_4, 4); \
1482 dup_8x8b(d64_128, 128); \
1484 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1485 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1486 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1487 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1489 #define setup_blocks_variables_unshaded_textured(target) \
1490 vec_4x32u u_block; \
1491 vec_4x32u v_block; \
1492 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1495 vec_2x32u uv = psx_gpu->uvrg.low; \
1497 vec_16x8u texture_mask; \
1498 shl_2x32b(uv_dx4, uv_dx, 2); \
1499 shl_2x32b(uv_dx8, uv_dx, 3); \
1500 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1501 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1504 #define setup_blocks_variables_unshaded_untextured_direct() \
1505 or_8x16b(colors, colors, msb_mask) \
1507 #define setup_blocks_variables_unshaded_untextured_indirect() \
1509 #define setup_blocks_variables_unshaded_untextured(target) \
1510 u32 color = psx_gpu->triangle_color; \
1513 u32 color_r = color & 0xFF; \
1514 u32 color_g = (color >> 8) & 0xFF; \
1515 u32 color_b = (color >> 16) & 0xFF; \
1517 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1518 ((color_b >> 3) << 10); \
1519 dup_8x16b(colors, color); \
1520 setup_blocks_variables_unshaded_untextured_##target() \
1522 #define setup_blocks_span_initialize_dithered_textured() \
1523 vec_8x16u dither_offsets; \
1524 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1526 #define setup_blocks_span_initialize_dithered_untextured() \
1527 vec_8x8u dither_offsets; \
1528 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1530 #define setup_blocks_span_initialize_dithered(texturing) \
1531 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1532 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1533 vec_8x8s dither_offsets_short; \
1536 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1537 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1538 setup_blocks_span_initialize_dithered_##texturing() \
1540 #define setup_blocks_span_initialize_undithered(texturing) \
1543 #define setup_blocks_span_initialize_shaded_textured() \
1545 vec_4x32u block_span; \
1546 u32 offset = span_edge_data->left_x; \
1548 uvrg = *span_uvrg_offset; \
1549 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1550 b = *span_b_offset; \
1551 b += b_dx * offset; \
1553 dup_4x32b(u_block, uvrg.e[0]); \
1554 dup_4x32b(v_block, uvrg.e[1]); \
1555 dup_4x32b(r_block, uvrg.e[2]); \
1556 dup_4x32b(g_block, uvrg.e[3]); \
1557 dup_4x32b(b_block, b); \
1559 block_span = psx_gpu->u_block_span; \
1560 add_4x32b(u_block, u_block, block_span); \
1561 block_span = psx_gpu->v_block_span; \
1562 add_4x32b(v_block, v_block, block_span); \
1563 block_span = psx_gpu->r_block_span; \
1564 add_4x32b(r_block, r_block, block_span); \
1565 block_span = psx_gpu->g_block_span; \
1566 add_4x32b(g_block, g_block, block_span); \
1567 block_span = psx_gpu->b_block_span; \
1568 add_4x32b(b_block, b_block, block_span); \
1571 #define setup_blocks_span_initialize_shaded_untextured() \
1573 vec_4x32u block_span; \
1574 u32 offset = span_edge_data->left_x; \
1576 rgb.low = span_uvrg_offset->high; \
1577 rgb.high.e[0] = *span_b_offset; \
1578 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1580 dup_4x32b(r_block, rgb.e[0]); \
1581 dup_4x32b(g_block, rgb.e[1]); \
1582 dup_4x32b(b_block, rgb.e[2]); \
1584 block_span = psx_gpu->r_block_span; \
1585 add_4x32b(r_block, r_block, block_span); \
1586 block_span = psx_gpu->g_block_span; \
1587 add_4x32b(g_block, g_block, block_span); \
1588 block_span = psx_gpu->b_block_span; \
1589 add_4x32b(b_block, b_block, block_span); \
1592 #define setup_blocks_span_initialize_unshaded_textured() \
1594 vec_4x32u block_span; \
1595 u32 offset = span_edge_data->left_x; \
1597 uv = span_uvrg_offset->low; \
1598 mla_scalar_2x32b(uv, uv_dx, offset); \
1600 dup_4x32b(u_block, uv.e[0]); \
1601 dup_4x32b(v_block, uv.e[1]); \
1603 block_span = psx_gpu->u_block_span; \
1604 add_4x32b(u_block, u_block, block_span); \
1605 block_span = psx_gpu->v_block_span; \
1606 add_4x32b(v_block, v_block, block_span); \
1609 #define setup_blocks_span_initialize_unshaded_untextured() \
1612 #define setup_blocks_texture_swizzled() \
1614 vec_8x8u u_saved = u; \
1615 sli_8x8b(u, v, 4); \
1616 sri_8x8b(v, u_saved, 4); \
1619 #define setup_blocks_texture_unswizzled() \
1621 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1624 vec_8x16u u_whole; \
1625 vec_8x16u v_whole; \
1626 vec_8x16u r_whole; \
1627 vec_8x16u g_whole; \
1628 vec_8x16u b_whole; \
1640 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1641 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1642 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1643 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1644 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1646 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1647 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1648 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1649 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1650 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1651 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1652 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1653 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1654 dup_4x32b(dx4, b_dx4); \
1655 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1657 mov_narrow_8x16b(u, u_whole); \
1658 mov_narrow_8x16b(v, v_whole); \
1659 mov_narrow_8x16b(r, r_whole); \
1660 mov_narrow_8x16b(g, g_whole); \
1661 mov_narrow_8x16b(b, b_whole); \
1663 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1664 add_4x32b(u_block, u_block, dx8); \
1665 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1666 add_4x32b(v_block, v_block, dx8); \
1667 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1668 add_4x32b(r_block, r_block, dx8); \
1669 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1670 add_4x32b(g_block, g_block, dx8); \
1671 dup_4x32b(dx8, b_dx8); \
1672 add_4x32b(b_block, b_block, dx8); \
1674 and_8x8b(u, u, texture_mask.low); \
1675 and_8x8b(v, v, texture_mask.high); \
1676 setup_blocks_texture_##swizzling(); \
1678 zip_8x16b(uv, u, v); \
1683 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1684 block->fb_ptr = fb_ptr; \
1687 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1690 vec_8x16u u_whole; \
1691 vec_8x16u v_whole; \
1700 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1701 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1703 dup_4x32b(dx4, uv_dx4.e[0]); \
1704 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1705 dup_4x32b(dx4, uv_dx4.e[1]); \
1706 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1708 mov_narrow_8x16b(u, u_whole); \
1709 mov_narrow_8x16b(v, v_whole); \
1711 dup_4x32b(dx8, uv_dx8.e[0]); \
1712 add_4x32b(u_block, u_block, dx8); \
1713 dup_4x32b(dx8, uv_dx8.e[1]); \
1714 add_4x32b(v_block, v_block, dx8); \
1716 and_8x8b(u, u, texture_mask.low); \
1717 and_8x8b(v, v, texture_mask.high); \
1718 setup_blocks_texture_##swizzling(); \
1720 zip_8x16b(uv, u, v); \
1722 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1723 block->fb_ptr = fb_ptr; \
1726 #define setup_blocks_store_shaded_untextured_dithered() \
1727 addq_8x8b(r, r, dither_offsets); \
1728 addq_8x8b(g, g, dither_offsets); \
1729 addq_8x8b(b, b, dither_offsets); \
1731 subq_8x8b(r, r, d64_4); \
1732 subq_8x8b(g, g, d64_4); \
1733 subq_8x8b(b, b, d64_4) \
1735 #define setup_blocks_store_shaded_untextured_undithered() \
1738 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1739 block->pixels = _pixels; \
1740 block->fb_ptr = fb_ptr \
1742 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1743 block->pixels = _pixels; \
1744 block->fb_ptr = fb_ptr \
1746 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1747 mul_long_8x8b(pixels, r, d64_1) \
1750 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1751 store_8x16b(_pixels, fb_ptr) \
1753 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1755 vec_8x16u fb_pixels; \
1756 vec_8x16u draw_mask; \
1757 vec_8x16u test_mask = psx_gpu->test_mask; \
1759 load_8x16b(fb_pixels, fb_ptr); \
1760 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1761 tst_8x16b(draw_mask, draw_mask, test_mask); \
1762 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1763 store_8x16b(fb_pixels, fb_ptr); \
1766 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1767 pixels = msb_mask; \
1768 mla_long_8x8b(pixels, r, d64_1) \
1771 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1774 vec_8x16u r_whole; \
1775 vec_8x16u g_whole; \
1776 vec_8x16u b_whole; \
1787 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1788 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1789 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1791 dup_4x32b(dx4, rgb_dx4.e[0]); \
1792 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1793 dup_4x32b(dx4, rgb_dx4.e[1]); \
1794 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1795 dup_4x32b(dx4, rgb_dx4.e[2]); \
1796 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1798 mov_narrow_8x16b(r, r_whole); \
1799 mov_narrow_8x16b(g, g_whole); \
1800 mov_narrow_8x16b(b, b_whole); \
1802 dup_4x32b(dx8, rgb_dx8.e[0]); \
1803 add_4x32b(r_block, r_block, dx8); \
1804 dup_4x32b(dx8, rgb_dx8.e[1]); \
1805 add_4x32b(g_block, g_block, dx8); \
1806 dup_4x32b(dx8, rgb_dx8.e[2]); \
1807 add_4x32b(b_block, b_block, dx8); \
1809 setup_blocks_store_shaded_untextured_##dithering(); \
1811 shr_8x8b(r, r, 3); \
1812 bic_8x8b(g, g, d64_0x07); \
1813 bic_8x8b(b, b, d64_0x07); \
1815 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1816 mla_long_8x8b(pixels, g, d64_4); \
1817 mla_long_8x8b(pixels, b, d64_128) \
1819 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1822 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1824 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1827 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1828 (_block)->draw_mask_bits = bits \
1830 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1832 vec_8x16u bits_mask; \
1833 vec_8x16u test_mask = psx_gpu->test_mask; \
1834 dup_8x16b(bits_mask, bits); \
1835 tst_8x16b(bits_mask, bits_mask, test_mask); \
1836 (_block)->draw_mask = bits_mask; \
1839 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1842 #define setup_blocks_add_blocks_indirect() \
1843 num_blocks += span_num_blocks; \
1845 if(num_blocks > MAX_BLOCKS) \
1847 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1848 flush_render_block_buffer(psx_gpu); \
1849 num_blocks = span_num_blocks; \
1850 block = psx_gpu->blocks; \
1853 #define setup_blocks_add_blocks_direct() \
1854 stats_add(texel_blocks_untextured, span_num_blocks); \
1855 span_pixel_blocks += span_num_blocks \
1858 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1859 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1860 psx_gpu_struct *psx_gpu) \
1862 setup_blocks_load_msb_mask_##target(); \
1863 setup_blocks_variables_##shading##_##texturing(target); \
1865 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1866 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1867 u32 *span_b_offset = psx_gpu->span_b_offset; \
1869 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1871 u32 num_spans = psx_gpu->num_spans; \
1876 u32 num_blocks = psx_gpu->num_blocks; \
1877 u32 span_num_blocks; \
1881 span_num_blocks = span_edge_data->num_blocks; \
1882 if(span_num_blocks) \
1884 y = span_edge_data->y; \
1885 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1887 setup_blocks_span_initialize_##shading##_##texturing(); \
1888 setup_blocks_span_initialize_##dithering(texturing); \
1890 setup_blocks_add_blocks_##target(); \
1892 s32 pixel_span = span_num_blocks * 8; \
1893 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1894 span_pixels += pixel_span; \
1896 span_num_blocks--; \
1897 while(span_num_blocks) \
1899 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1901 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1905 span_num_blocks--; \
1908 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1909 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1910 span_edge_data->right_mask); \
1916 zero_block_spans++; \
1921 span_uvrg_offset++; \
1925 psx_gpu->num_blocks = num_blocks; \
1929 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1933 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1934 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1936 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1937 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1939 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1940 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1941 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1942 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1944 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1945 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1947 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1949 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1950 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1953 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1955 block_struct *block = psx_gpu->blocks;
1956 u32 num_blocks = psx_gpu->num_blocks;
1957 stats_add(texel_blocks_4bpp, num_blocks);
1959 vec_8x8u texels_low;
1960 vec_8x8u texels_high;
1967 vec_16x8u clut_high;
1969 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1970 u16 *clut_ptr = psx_gpu->clut_ptr;
1972 // Can be done with one deinterleaving load on NEON
1973 load_8x16b(clut_a, clut_ptr);
1974 load_8x16b(clut_b, clut_ptr + 8);
1975 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
1977 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
1978 update_texture_4bpp_cache(psx_gpu);
1982 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
1983 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
1984 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
1985 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
1986 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
1987 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
1988 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
1989 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
1991 tbl_16(texels_low, texels, clut_low);
1992 tbl_16(texels_high, texels, clut_high);
1994 // Can be done with an interleaving store on NEON
1995 zip_8x16b(pixels, texels_low, texels_high);
1997 block->texels = pixels;
2004 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2006 block_struct *block = psx_gpu->blocks;
2007 u32 num_blocks = psx_gpu->num_blocks;
2009 stats_add(texel_blocks_8bpp, num_blocks);
2011 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2012 update_texture_8bpp_cache(psx_gpu);
2015 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2023 for(i = 0; i < 8; i++)
2025 offset = block->uv.e[i];
2027 texel = texture_ptr_8bpp[offset];
2028 texels.e[i] = psx_gpu->clut_ptr[texel];
2031 block->texels = texels;
2038 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2040 block_struct *block = psx_gpu->blocks;
2041 u32 num_blocks = psx_gpu->num_blocks;
2043 stats_add(texel_blocks_16bpp, num_blocks);
2047 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2053 for(i = 0; i < 8; i++)
2055 offset = block->uv.e[i];
2056 offset += ((offset & 0xFF00) * 3);
2058 texels.e[i] = texture_ptr_16bpp[offset];
2061 block->texels = texels;
2071 #define shade_blocks_load_msb_mask_indirect() \
2073 #define shade_blocks_load_msb_mask_direct() \
2074 vec_8x16u msb_mask; \
2075 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2077 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2078 block->draw_mask = _draw_mask; \
2079 block->pixels = _pixels \
2081 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2083 vec_8x16u fb_pixels; \
2084 or_8x16b(_pixels, _pixels, msb_mask); \
2085 load_8x16b(fb_pixels, block->fb_ptr); \
2086 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2087 store_8x16b(fb_pixels, block->fb_ptr); \
2091 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2092 if(psx_gpu->triangle_color == 0x808080) \
2094 false_modulated_blocks += num_blocks; \
2097 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2098 if(psx_gpu->triangle_color == 0x808080) \
2101 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2102 false_modulated_blocks += num_blocks; \
2107 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2110 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2113 u32 color = psx_gpu->triangle_color; \
2114 dup_8x8b(colors_r, color); \
2115 dup_8x8b(colors_g, color >> 8); \
2116 dup_8x8b(colors_b, color >> 16); \
2117 shade_blocks_textured_false_modulated_check_##dithering(target); \
2120 #define shade_blocks_textured_modulated_shaded_block_load() \
2121 colors_r = block->r; \
2122 colors_g = block->g; \
2123 colors_b = block->b \
2125 #define shade_blocks_textured_modulated_unshaded_block_load() \
2127 #define shade_blocks_textured_modulate_dithered(component) \
2128 pixels_##component = block->dither_offsets; \
2129 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2131 #define shade_blocks_textured_modulate_undithered(component) \
2132 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2134 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2135 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2136 psx_gpu_struct *psx_gpu) \
2138 block_struct *block = psx_gpu->blocks; \
2139 u32 num_blocks = psx_gpu->num_blocks; \
2142 vec_8x8u texels_r; \
2143 vec_8x8u texels_g; \
2144 vec_8x8u texels_b; \
2146 vec_8x8u colors_r; \
2147 vec_8x8u colors_g; \
2148 vec_8x8u colors_b; \
2150 vec_8x8u pixels_r_low; \
2151 vec_8x8u pixels_g_low; \
2152 vec_8x8u pixels_b_low; \
2155 vec_8x16u pixels_r; \
2156 vec_8x16u pixels_g; \
2157 vec_8x16u pixels_b; \
2159 vec_8x16u draw_mask; \
2160 vec_8x16u zero_mask; \
2162 vec_8x8u d64_0x07; \
2163 vec_8x8u d64_0x1F; \
2168 vec_8x16u d128_0x8000; \
2170 vec_8x16u test_mask = psx_gpu->test_mask; \
2171 u32 draw_mask_bits; \
2172 shade_blocks_load_msb_mask_##target(); \
2174 dup_8x8b(d64_0x07, 0x07); \
2175 dup_8x8b(d64_0x1F, 0x1F); \
2176 dup_8x8b(d64_1, 1); \
2177 dup_8x8b(d64_4, 4); \
2178 dup_8x8b(d64_128, 128); \
2180 dup_8x16b(d128_0x8000, 0x8000); \
2182 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2187 draw_mask_bits = block->draw_mask_bits; \
2188 dup_8x16b(draw_mask, draw_mask_bits); \
2189 tst_8x16b(draw_mask, draw_mask, test_mask); \
2191 shade_blocks_textured_modulated_##shading##_block_load(); \
2193 texels = block->texels; \
2195 mov_narrow_8x16b(texels_r, texels); \
2196 shr_narrow_8x16b(texels_g, texels, 5); \
2197 shr_narrow_8x16b(texels_b, texels, 7); \
2199 and_8x8b(texels_r, texels_r, d64_0x1F); \
2200 and_8x8b(texels_g, texels_g, d64_0x1F); \
2201 shr_8x8b(texels_b, texels_b, 3); \
2203 shade_blocks_textured_modulate_##dithering(r); \
2204 shade_blocks_textured_modulate_##dithering(g); \
2205 shade_blocks_textured_modulate_##dithering(b); \
2207 cmpeqz_8x16b(zero_mask, texels); \
2208 and_8x16b(pixels, texels, d128_0x8000); \
2210 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2211 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2212 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2214 or_8x16b(zero_mask, draw_mask, zero_mask); \
2216 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2217 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2218 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2220 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2221 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2222 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2224 shade_blocks_store_##target(zero_mask, pixels); \
2233 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2234 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2235 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2236 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2238 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2239 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2240 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2241 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2246 #define shade_blocks_textured_unmodulated_builder(target) \
2247 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2249 block_struct *block = psx_gpu->blocks; \
2250 u32 num_blocks = psx_gpu->num_blocks; \
2251 vec_8x16u draw_mask; \
2252 vec_8x16u test_mask = psx_gpu->test_mask; \
2253 u32 draw_mask_bits; \
2256 shade_blocks_load_msb_mask_##target(); \
2260 vec_8x16u zero_mask; \
2262 draw_mask_bits = block->draw_mask_bits; \
2263 dup_8x16b(draw_mask, draw_mask_bits); \
2264 tst_8x16b(draw_mask, draw_mask, test_mask); \
2266 pixels = block->texels; \
2268 cmpeqz_8x16b(zero_mask, pixels); \
2269 or_8x16b(zero_mask, draw_mask, zero_mask); \
2271 shade_blocks_store_##target(zero_mask, pixels); \
2278 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2279 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2282 block_struct *block = psx_gpu->blocks; \
2283 u32 num_blocks = psx_gpu->num_blocks; \
2284 vec_8x16u draw_mask; \
2285 vec_8x16u test_mask = psx_gpu->test_mask; \
2286 u32 draw_mask_bits; \
2289 shade_blocks_load_msb_mask_##target(); \
2293 vec_8x16u zero_mask; \
2295 draw_mask_bits = block->draw_mask_bits; \
2296 dup_8x16b(draw_mask, draw_mask_bits); \
2297 tst_8x16b(draw_mask, draw_mask, test_mask); \
2299 pixels = block->texels; \
2301 cmpeqz_8x16b(zero_mask, pixels); \
2302 or_8x16b(zero_mask, draw_mask, zero_mask); \
2304 shade_blocks_store_##target(zero_mask, pixels); \
2313 shade_blocks_textured_unmodulated_builder(indirect)
2314 shade_blocks_textured_unmodulated_builder(direct)
2316 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2320 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2322 block_struct *block = psx_gpu->blocks;
2323 u32 num_blocks = psx_gpu->num_blocks;
2325 vec_8x16u pixels = block->pixels;
2326 shade_blocks_load_msb_mask_direct();
2330 shade_blocks_store_direct(block->draw_mask, pixels);
2339 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2344 #define blend_blocks_mask_evaluate_on() \
2345 vec_8x16u mask_pixels; \
2346 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2347 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2349 #define blend_blocks_mask_evaluate_off() \
2351 #define blend_blocks_average() \
2353 vec_8x16u pixels_no_msb; \
2354 vec_8x16u fb_pixels_no_msb; \
2356 vec_8x16u d128_0x0421; \
2357 vec_8x16u d128_0x8000; \
2359 dup_8x16b(d128_0x0421, 0x0421); \
2360 dup_8x16b(d128_0x8000, 0x8000); \
2362 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2363 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2364 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2365 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2366 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2367 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2370 #define blend_blocks_add() \
2372 vec_8x16u pixels_rb, pixels_g; \
2373 vec_8x16u fb_rb, fb_g; \
2375 vec_8x16u d128_0x7C1F; \
2376 vec_8x16u d128_0x03E0; \
2378 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2379 dup_8x16b(d128_0x03E0, 0x03E0); \
2381 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2382 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2384 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2385 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2387 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2388 add_8x16b(fb_g, fb_g, pixels_g); \
2390 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2391 vector_cast(vec_16x8u, d128_0x7C1F)); \
2392 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2394 or_8x16b(blend_pixels, fb_rb, fb_g); \
2397 #define blend_blocks_subtract() \
2399 vec_8x16u pixels_rb, pixels_g; \
2400 vec_8x16u fb_rb, fb_g; \
2402 vec_8x16u d128_0x7C1F; \
2403 vec_8x16u d128_0x03E0; \
2405 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2406 dup_8x16b(d128_0x03E0, 0x03E0); \
2408 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2409 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2411 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2412 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2414 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2415 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2416 subs_8x16b(fb_g, fb_g, pixels_g); \
2418 or_8x16b(blend_pixels, fb_rb, fb_g); \
2421 #define blend_blocks_add_fourth() \
2423 vec_8x16u pixels_rb, pixels_g; \
2424 vec_8x16u pixels_fourth; \
2425 vec_8x16u fb_rb, fb_g; \
2427 vec_8x16u d128_0x7C1F; \
2428 vec_8x16u d128_0x1C07; \
2429 vec_8x16u d128_0x03E0; \
2430 vec_8x16u d128_0x00E0; \
2432 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2433 dup_8x16b(d128_0x1C07, 0x1C07); \
2434 dup_8x16b(d128_0x03E0, 0x03E0); \
2435 dup_8x16b(d128_0x00E0, 0x00E0); \
2437 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2439 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2440 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2442 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2443 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2445 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2446 add_8x16b(fb_g, fb_g, pixels_g); \
2448 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2449 vector_cast(vec_16x8u, d128_0x7C1F)); \
2450 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2452 or_8x16b(blend_pixels, fb_rb, fb_g); \
2455 #define blend_blocks_blended_combine_textured() \
2457 vec_8x16u blend_mask; \
2458 cmpltz_8x16b(blend_mask, pixels); \
2460 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2461 bif_8x16b(blend_pixels, pixels, blend_mask); \
2464 #define blend_blocks_blended_combine_untextured() \
2467 #define blend_blocks_body_blend(blend_mode, texturing) \
2469 blend_blocks_##blend_mode(); \
2470 blend_blocks_blended_combine_##texturing(); \
2473 #define blend_blocks_body_average(texturing) \
2474 blend_blocks_body_blend(average, texturing) \
2476 #define blend_blocks_body_add(texturing) \
2477 blend_blocks_body_blend(add, texturing) \
2479 #define blend_blocks_body_subtract(texturing) \
2480 blend_blocks_body_blend(subtract, texturing) \
2482 #define blend_blocks_body_add_fourth(texturing) \
2483 blend_blocks_body_blend(add_fourth, texturing) \
2485 #define blend_blocks_body_unblended(texturing) \
2486 blend_pixels = pixels \
2489 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2491 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2494 block_struct *block = psx_gpu->blocks; \
2495 u32 num_blocks = psx_gpu->num_blocks; \
2496 vec_8x16u draw_mask; \
2498 vec_8x16u blend_pixels; \
2499 vec_8x16u framebuffer_pixels; \
2500 vec_8x16u msb_mask; \
2504 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2508 pixels = block->pixels; \
2509 draw_mask = block->draw_mask; \
2510 fb_ptr = block->fb_ptr; \
2512 load_8x16b(framebuffer_pixels, fb_ptr); \
2514 blend_blocks_mask_evaluate_##mask_evaluate(); \
2515 blend_blocks_body_##blend_mode(texturing); \
2517 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2518 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2519 store_8x16b(framebuffer_pixels, fb_ptr); \
2529 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2533 blend_blocks_builder(textured, average, off);
2534 blend_blocks_builder(textured, average, on);
2535 blend_blocks_builder(textured, add, off);
2536 blend_blocks_builder(textured, add, on);
2537 blend_blocks_builder(textured, subtract, off);
2538 blend_blocks_builder(textured, subtract, on);
2539 blend_blocks_builder(textured, add_fourth, off);
2540 blend_blocks_builder(textured, add_fourth, on);
2542 blend_blocks_builder(untextured, average, off);
2543 blend_blocks_builder(untextured, average, on);
2544 blend_blocks_builder(untextured, add, off);
2545 blend_blocks_builder(untextured, add, on);
2546 blend_blocks_builder(untextured, subtract, off);
2547 blend_blocks_builder(untextured, subtract, on);
2548 blend_blocks_builder(untextured, add_fourth, off);
2549 blend_blocks_builder(untextured, add_fourth, on);
2551 blend_blocks_builder(textured, unblended, on);
2556 #define vertex_swap(_a, _b) \
2558 vertex_struct *temp_vertex = _a; \
2561 triangle_winding ^= 1; \
2565 // Setup blocks parametric-variables:
2566 // SHADE TEXTURE_MAP SWIZZLING
2573 // 8 inputs, 6 combinations
2575 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2576 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2578 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2579 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2581 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2583 setup_blocks_switch_untextured_##shading(dithering, target) \
2585 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2586 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2588 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2589 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2591 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2592 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2594 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2595 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2597 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2598 dithering, mask_evaluate) \
2599 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2601 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2603 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2605 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2607 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2609 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2610 dithering, mask_evaluate) \
2611 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2612 texture_mode, dithering) \
2614 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2615 blending, mask_evaluate) \
2616 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2617 dithering, mask_evaluate) \
2622 #define texture_blocks_switch_untextured(texture_mode) \
2623 texture_blocks_untextured \
2625 #define texture_blocks_switch_textured(texture_mode) \
2626 texture_blocks_##texture_mode \
2628 #define texture_blocks_switch(texturing, texture_mode) \
2629 texture_blocks_switch_##texturing(texture_mode) \
2632 // Shade blocks parametric-variables:
2633 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2642 // 16 inputs, 8 combinations
2644 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2645 shade_blocks_unshaded_untextured_##target \
2647 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2648 shade_blocks_textured_unmodulated_##target \
2650 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2651 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2653 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2654 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2656 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2657 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2659 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2660 shade_blocks_shaded_untextured \
2662 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2663 shade_blocks_textured_unmodulated_##target \
2665 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2666 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2668 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2669 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2671 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2672 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2674 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2676 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2678 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2680 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2682 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2684 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2686 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2687 dithering, mask_evaluate) \
2688 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2691 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2692 blending, mask_evaluate) \
2693 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2697 // Blend blocks parametric-variables:
2698 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2717 // 32 inputs, 18 combinations
2719 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2720 blend_blocks_textured_unblended_##mask_evaluate \
2722 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2723 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2725 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2726 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2729 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2730 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2732 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2734 texture_blocks_switch(texturing, texture_mode), \
2735 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2737 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2740 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2741 mask_evaluate, shading, dithering, texturing, blending) \
2742 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2743 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2744 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2745 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2747 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2748 mask_evaluate, shading, dithering, texturing) \
2749 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2750 mask_evaluate, shading, dithering, texturing, unblended), \
2751 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2752 mask_evaluate, shading, dithering, texturing, blended) \
2754 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2755 mask_evaluate, shading, dithering) \
2756 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2757 mask_evaluate, shading, dithering, untextured), \
2758 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2759 mask_evaluate, shading, dithering, textured) \
2761 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2762 mask_evaluate, shading) \
2763 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2764 mask_evaluate, shading, undithered), \
2765 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2766 mask_evaluate, shading, dithered) \
2768 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2770 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2772 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2775 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2776 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2777 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2779 #define render_blocks_switch_block_texture_mode(texture_mode) \
2780 render_blocks_switch_block_blend_mode(texture_mode, average), \
2781 render_blocks_switch_block_blend_mode(texture_mode, add), \
2782 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2783 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2785 #define render_blocks_switch_block() \
2786 render_blocks_switch_block_texture_mode(4bpp), \
2787 render_blocks_switch_block_texture_mode(8bpp), \
2788 render_blocks_switch_block_texture_mode(16bpp), \
2789 render_blocks_switch_block_texture_mode(16bpp) \
2792 render_block_handler_struct render_triangle_block_handlers[] =
2794 render_blocks_switch_block()
2797 #undef render_blocks_switch_block_modulation
2799 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2800 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2802 "texture mode: " #texture_mode "\n" \
2803 "blend mode: " #blend_mode "\n" \
2804 "mask evaluation: " #mask_evaluate "\n" \
2811 char *render_block_flag_strings[] =
2813 render_blocks_switch_block()
2817 #define triangle_y_direction_up 1
2818 #define triangle_y_direction_flat 2
2819 #define triangle_y_direction_down 0
2821 #define triangle_winding_positive 0
2822 #define triangle_winding_negative 1
2824 #define triangle_set_direction(direction_variable, value) \
2825 u32 direction_variable = (u32)(value) >> 31; \
2827 direction_variable = 2 \
2829 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2830 case (triangle_y_direction_##direction_a | \
2831 (triangle_y_direction_##direction_b << 2) | \
2832 (triangle_y_direction_##direction_c << 4) | \
2833 (triangle_winding_##winding << 6)) \
2835 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2836 vertex_struct *vertexes_out[3])
2838 s32 y_top, y_bottom;
2840 u32 triangle_winding = 0;
2842 vertex_struct *a = &(vertexes[0]);
2843 vertex_struct *b = &(vertexes[1]);
2844 vertex_struct *c = &(vertexes[2]);
2846 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2852 if(triangle_area == 0)
2874 if((y_bottom - y_top) >= 512)
2882 if(triangle_area < 0)
2884 triangle_area = -triangle_area;
2885 triangle_winding ^= 1;
2900 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2908 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2917 psx_gpu->triangle_area = triangle_area;
2918 psx_gpu->triangle_winding = triangle_winding;
2920 vertexes_out[0] = a;
2921 vertexes_out[1] = b;
2922 vertexes_out[2] = c;
2927 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2928 vertex_struct *vertex_ptrs[3], u32 flags)
2930 psx_gpu->num_spans = 0;
2932 vertex_struct *a = vertex_ptrs[0];
2933 vertex_struct *b = vertex_ptrs[1];
2934 vertex_struct *c = vertex_ptrs[2];
2936 s32 y_delta_a = b->y - a->y;
2937 s32 y_delta_b = c->y - b->y;
2938 s32 y_delta_c = c->y - a->y;
2940 triangle_set_direction(y_direction_a, y_delta_a);
2941 triangle_set_direction(y_direction_b, y_delta_b);
2942 triangle_set_direction(y_direction_c, y_delta_c);
2944 compute_all_gradients(psx_gpu, a, b, c);
2946 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2947 (psx_gpu->triangle_winding << 6))
2949 triangle_case(up, up, up, negative):
2950 triangle_case(up, up, flat, negative):
2951 triangle_case(up, up, down, negative):
2952 setup_spans_up_right(psx_gpu, a, b, c);
2955 triangle_case(flat, up, up, negative):
2956 triangle_case(flat, up, flat, negative):
2957 triangle_case(flat, up, down, negative):
2958 setup_spans_up_a(psx_gpu, a, b, c);
2961 triangle_case(down, up, up, negative):
2962 setup_spans_up_down(psx_gpu, a, c, b);
2965 triangle_case(down, up, flat, negative):
2966 setup_spans_down_a(psx_gpu, a, c, b);
2969 triangle_case(down, up, down, negative):
2970 setup_spans_down_right(psx_gpu, a, c, b);
2973 triangle_case(down, flat, up, negative):
2974 triangle_case(down, flat, flat, negative):
2975 triangle_case(down, flat, down, negative):
2976 setup_spans_down_b(psx_gpu, a, b, c);
2979 triangle_case(down, down, up, negative):
2980 triangle_case(down, down, flat, negative):
2981 triangle_case(down, down, down, negative):
2982 setup_spans_down_left(psx_gpu, a, b, c);
2985 triangle_case(up, up, up, positive):
2986 triangle_case(up, up, flat, positive):
2987 triangle_case(up, up, down, positive):
2988 setup_spans_up_left(psx_gpu, a, b, c);
2991 triangle_case(up, flat, up, positive):
2992 triangle_case(up, flat, flat, positive):
2993 triangle_case(up, flat, down, positive):
2994 setup_spans_up_b(psx_gpu, a, b, c);
2997 triangle_case(up, down, up, positive):
2998 setup_spans_up_right(psx_gpu, a, c, b);
3001 triangle_case(up, down, flat, positive):
3002 setup_spans_up_a(psx_gpu, a, c, b);
3005 triangle_case(up, down, down, positive):
3006 setup_spans_up_down(psx_gpu, a, b, c);
3009 triangle_case(flat, down, up, positive):
3010 triangle_case(flat, down, flat, positive):
3011 triangle_case(flat, down, down, positive):
3012 setup_spans_down_a(psx_gpu, a, b, c);
3015 triangle_case(down, down, up, positive):
3016 triangle_case(down, down, flat, positive):
3017 triangle_case(down, down, down, positive):
3018 setup_spans_down_right(psx_gpu, a, b, c);
3023 spans += psx_gpu->num_spans;
3026 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3030 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3032 for(i = 0; i < psx_gpu->num_spans; i++)
3034 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3035 psx_gpu->span_edge_data[i].num_blocks = 0;
3040 for(i = 0; i < psx_gpu->num_spans; i++)
3042 if(psx_gpu->span_edge_data[i].y & 1)
3043 psx_gpu->span_edge_data[i].num_blocks = 0;
3048 u32 render_state = flags &
3049 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3050 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3051 render_state |= psx_gpu->render_state_base;
3053 if((psx_gpu->render_state != render_state) ||
3054 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3056 psx_gpu->render_state = render_state;
3057 flush_render_block_buffer(psx_gpu);
3063 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3065 psx_gpu->render_block_handler =
3066 &(render_triangle_block_handlers[render_state]);
3067 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3071 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3074 vertex_struct *vertex_ptrs[3];
3075 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3076 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3079 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3081 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3083 block_struct *block = psx_gpu->blocks;
3084 u32 num_blocks = psx_gpu->num_blocks;
3087 vec_8x8u texel_indexes;
3089 u16 *clut_ptr = psx_gpu->clut_ptr;
3094 texel_indexes = block->r;
3096 for(i = 0; i < 8; i++)
3098 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3101 block->texels = texels;
3111 #define setup_sprite_tiled_initialize_4bpp_clut() \
3112 u16 *clut_ptr = psx_gpu->clut_ptr; \
3113 vec_8x16u clut_a, clut_b; \
3114 vec_16x8u clut_low, clut_high; \
3116 load_8x16b(clut_a, clut_ptr); \
3117 load_8x16b(clut_b, clut_ptr + 8); \
3118 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3120 #define setup_sprite_tiled_initialize_4bpp() \
3121 setup_sprite_tiled_initialize_4bpp_clut(); \
3123 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3124 update_texture_4bpp_cache(psx_gpu) \
3126 #define setup_sprite_tiled_initialize_8bpp() \
3127 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3128 update_texture_8bpp_cache(psx_gpu) \
3131 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3132 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3133 ((texture_offset + offset) & texture_mask); \
3135 load_64b(texels, texture_block_ptr) \
3138 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3139 num_blocks += tile_num_blocks; \
3140 sprite_blocks += tile_num_blocks; \
3142 if(num_blocks > MAX_BLOCKS) \
3144 flush_render_block_buffer(psx_gpu); \
3145 num_blocks = tile_num_blocks; \
3146 block = psx_gpu->blocks; \
3149 #define setup_sprite_tile_full_4bpp(edge) \
3151 vec_8x8u texels_low, texels_high; \
3153 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3155 while(sub_tile_height) \
3157 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3158 tbl_16(texels_low, texels, clut_low); \
3159 tbl_16(texels_high, texels, clut_high); \
3160 zip_8x16b(pixels, texels_low, texels_high); \
3162 block->texels = pixels; \
3163 block->draw_mask_bits = left_mask_bits; \
3164 block->fb_ptr = fb_ptr; \
3167 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3168 tbl_16(texels_low, texels, clut_low); \
3169 tbl_16(texels_high, texels, clut_high); \
3170 zip_8x16b(pixels, texels_low, texels_high); \
3172 block->texels = pixels; \
3173 block->draw_mask_bits = right_mask_bits; \
3174 block->fb_ptr = fb_ptr + 8; \
3178 texture_offset += 0x10; \
3179 sub_tile_height--; \
3181 texture_offset += 0xF00; \
3182 psx_gpu->num_blocks = num_blocks; \
3185 #define setup_sprite_tile_half_4bpp(edge) \
3187 vec_8x8u texels_low, texels_high; \
3189 setup_sprite_tile_add_blocks(sub_tile_height); \
3191 while(sub_tile_height) \
3193 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3194 tbl_16(texels_low, texels, clut_low); \
3195 tbl_16(texels_high, texels, clut_high); \
3196 zip_8x16b(pixels, texels_low, texels_high); \
3198 block->texels = pixels; \
3199 block->draw_mask_bits = edge##_mask_bits; \
3200 block->fb_ptr = fb_ptr; \
3204 texture_offset += 0x10; \
3205 sub_tile_height--; \
3207 texture_offset += 0xF00; \
3208 psx_gpu->num_blocks = num_blocks; \
3212 #define setup_sprite_tile_full_8bpp(edge) \
3214 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3216 while(sub_tile_height) \
3218 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3219 block->r = texels; \
3220 block->draw_mask_bits = left_mask_bits; \
3221 block->fb_ptr = fb_ptr; \
3224 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3225 block->r = texels; \
3226 block->draw_mask_bits = right_mask_bits; \
3227 block->fb_ptr = fb_ptr + 8; \
3231 texture_offset += 0x10; \
3232 sub_tile_height--; \
3234 texture_offset += 0xF00; \
3235 psx_gpu->num_blocks = num_blocks; \
3238 #define setup_sprite_tile_half_8bpp(edge) \
3240 setup_sprite_tile_add_blocks(sub_tile_height); \
3242 while(sub_tile_height) \
3244 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3245 block->r = texels; \
3246 block->draw_mask_bits = edge##_mask_bits; \
3247 block->fb_ptr = fb_ptr; \
3251 texture_offset += 0x10; \
3252 sub_tile_height--; \
3254 texture_offset += 0xF00; \
3255 psx_gpu->num_blocks = num_blocks; \
3259 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3260 texture_offset = texture_offset_base + 8; \
3263 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3264 texture_offset = texture_offset_base \
3266 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3267 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3269 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3270 texture_offset = texture_offset_base \
3272 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3275 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3277 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3278 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3280 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3283 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3287 sub_tile_height = column_data; \
3288 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3289 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3290 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3293 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3297 u32 tiles_remaining = column_data >> 16; \
3298 sub_tile_height = column_data & 0xFF; \
3299 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3300 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3301 tiles_remaining -= 1; \
3303 while(tiles_remaining) \
3305 sub_tile_height = 16; \
3306 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3307 tiles_remaining--; \
3310 sub_tile_height = (column_data >> 8) & 0xFF; \
3311 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3312 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3316 #define setup_sprite_column_data_single() \
3317 column_data = height \
3319 #define setup_sprite_column_data_multi() \
3320 column_data = 16 - offset_v; \
3321 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3322 column_data |= (tile_height - 1) << 16 \
3325 #define RIGHT_MASK_BIT_SHIFT 8
3326 #define RIGHT_MASK_BIT_SHIFT_4x 16
3328 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3329 edge_mode, edge, x4mode) \
3331 setup_sprite_column_data_##multi_height(); \
3332 left_mask_bits = left_block_mask | right_block_mask; \
3333 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3335 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3336 texture_mode, x4mode); \
3339 #define setup_sprite_tiled_advance_column() \
3340 texture_offset_base += 0x100; \
3341 if((texture_offset_base & 0xF00) == 0) \
3342 texture_offset_base -= (0x100 + 0xF00) \
3344 #define FB_PTR_MULTIPLIER 1
3345 #define FB_PTR_MULTIPLIER_4x 2
3347 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3348 left_mode, right_mode, x4mode) \
3350 setup_sprite_column_data_##multi_height(); \
3351 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3352 * FB_PTR_MULTIPLIER##x4mode; \
3355 left_mask_bits = left_block_mask; \
3356 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3358 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3359 texture_mode, x4mode); \
3360 fb_ptr += fb_ptr_advance_column; \
3362 left_mask_bits = 0x00; \
3363 right_mask_bits = 0x00; \
3367 setup_sprite_tiled_advance_column(); \
3368 setup_sprite_tile_column_height_##multi_height(full, none, \
3369 texture_mode, x4mode); \
3370 fb_ptr += fb_ptr_advance_column; \
3374 left_mask_bits = right_block_mask; \
3375 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3377 setup_sprite_tiled_advance_column(); \
3378 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3379 texture_mode, x4mode); \
3384 #define setup_sprite_tiled_initialize_4bpp_4x() \
3385 setup_sprite_tiled_initialize_4bpp_clut() \
3387 #define setup_sprite_tiled_initialize_8bpp_4x() \
3390 #define setup_sprite_tile_full_4bpp_4x(edge) \
3392 vec_8x8u texels_low, texels_high; \
3393 vec_8x16u pixels, pixels_wide; \
3394 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3395 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3396 u32 left_mask_bits_b = left_mask_bits >> 8; \
3397 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3398 u32 right_mask_bits_b = right_mask_bits >> 8; \
3400 while(sub_tile_height) \
3402 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3403 tbl_16(texels_low, texels, clut_low); \
3404 tbl_16(texels_high, texels, clut_high); \
3405 zip_8x16b(pixels, texels_low, texels_high); \
3407 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3408 block->texels = pixels_wide; \
3409 block->draw_mask_bits = left_mask_bits_a; \
3410 block->fb_ptr = fb_ptr; \
3413 block->texels = pixels_wide; \
3414 block->draw_mask_bits = left_mask_bits_a; \
3415 block->fb_ptr = fb_ptr + 1024; \
3418 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3419 block->texels = pixels_wide; \
3420 block->draw_mask_bits = left_mask_bits_b; \
3421 block->fb_ptr = fb_ptr + 8; \
3424 block->texels = pixels_wide; \
3425 block->draw_mask_bits = left_mask_bits_b; \
3426 block->fb_ptr = fb_ptr + 1024 + 8; \
3429 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3430 tbl_16(texels_low, texels, clut_low); \
3431 tbl_16(texels_high, texels, clut_high); \
3432 zip_8x16b(pixels, texels_low, texels_high); \
3434 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3435 block->texels = pixels_wide; \
3436 block->draw_mask_bits = right_mask_bits_a; \
3437 block->fb_ptr = fb_ptr + 16; \
3440 block->texels = pixels_wide; \
3441 block->draw_mask_bits = right_mask_bits_a; \
3442 block->fb_ptr = fb_ptr + 1024 + 16; \
3445 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3446 block->texels = pixels_wide; \
3447 block->draw_mask_bits = right_mask_bits_b; \
3448 block->fb_ptr = fb_ptr + 24; \
3451 block->texels = pixels_wide; \
3452 block->draw_mask_bits = right_mask_bits_b; \
3453 block->fb_ptr = fb_ptr + 1024 + 24; \
3457 texture_offset += 0x10; \
3458 sub_tile_height--; \
3460 texture_offset += 0xF00; \
3461 psx_gpu->num_blocks = num_blocks; \
3464 #define setup_sprite_tile_half_4bpp_4x(edge) \
3466 vec_8x8u texels_low, texels_high; \
3467 vec_8x16u pixels, pixels_wide; \
3468 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3469 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3470 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3472 while(sub_tile_height) \
3474 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3475 tbl_16(texels_low, texels, clut_low); \
3476 tbl_16(texels_high, texels, clut_high); \
3477 zip_8x16b(pixels, texels_low, texels_high); \
3479 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3480 block->texels = pixels_wide; \
3481 block->draw_mask_bits = edge##_mask_bits_a; \
3482 block->fb_ptr = fb_ptr; \
3485 block->texels = pixels_wide; \
3486 block->draw_mask_bits = edge##_mask_bits_a; \
3487 block->fb_ptr = fb_ptr + 1024; \
3490 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3491 block->texels = pixels_wide; \
3492 block->draw_mask_bits = edge##_mask_bits_b; \
3493 block->fb_ptr = fb_ptr + 8; \
3496 block->texels = pixels_wide; \
3497 block->draw_mask_bits = edge##_mask_bits_b; \
3498 block->fb_ptr = fb_ptr + 1024 + 8; \
3502 texture_offset += 0x10; \
3503 sub_tile_height--; \
3505 texture_offset += 0xF00; \
3506 psx_gpu->num_blocks = num_blocks; \
3510 #define setup_sprite_tile_full_8bpp_4x(edge) \
3512 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3513 vec_16x8u texels_wide; \
3514 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3515 u32 left_mask_bits_b = left_mask_bits >> 8; \
3516 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3517 u32 right_mask_bits_b = right_mask_bits >> 8; \
3519 while(sub_tile_height) \
3521 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3522 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3523 block->r = texels_wide.low; \
3524 block->draw_mask_bits = left_mask_bits_a; \
3525 block->fb_ptr = fb_ptr; \
3528 block->r = texels_wide.low; \
3529 block->draw_mask_bits = left_mask_bits_a; \
3530 block->fb_ptr = fb_ptr + 1024; \
3533 block->r = texels_wide.high; \
3534 block->draw_mask_bits = left_mask_bits_b; \
3535 block->fb_ptr = fb_ptr + 8; \
3538 block->r = texels_wide.high; \
3539 block->draw_mask_bits = left_mask_bits_b; \
3540 block->fb_ptr = fb_ptr + 1024 + 8; \
3543 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3544 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3545 block->r = texels_wide.low; \
3546 block->draw_mask_bits = right_mask_bits_a; \
3547 block->fb_ptr = fb_ptr + 16; \
3550 block->r = texels_wide.low; \
3551 block->draw_mask_bits = right_mask_bits_a; \
3552 block->fb_ptr = fb_ptr + 1024 + 16; \
3555 block->r = texels_wide.high; \
3556 block->draw_mask_bits = right_mask_bits_b; \
3557 block->fb_ptr = fb_ptr + 24; \
3560 block->r = texels_wide.high; \
3561 block->draw_mask_bits = right_mask_bits_b; \
3562 block->fb_ptr = fb_ptr + 24 + 1024; \
3566 texture_offset += 0x10; \
3567 sub_tile_height--; \
3569 texture_offset += 0xF00; \
3570 psx_gpu->num_blocks = num_blocks; \
3573 #define setup_sprite_tile_half_8bpp_4x(edge) \
3575 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3576 vec_16x8u texels_wide; \
3577 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3578 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3580 while(sub_tile_height) \
3582 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3583 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3584 block->r = texels_wide.low; \
3585 block->draw_mask_bits = edge##_mask_bits_a; \
3586 block->fb_ptr = fb_ptr; \
3589 block->r = texels_wide.low; \
3590 block->draw_mask_bits = edge##_mask_bits_a; \
3591 block->fb_ptr = fb_ptr + 1024; \
3594 block->r = texels_wide.high; \
3595 block->draw_mask_bits = edge##_mask_bits_b; \
3596 block->fb_ptr = fb_ptr + 8; \
3599 block->r = texels_wide.high; \
3600 block->draw_mask_bits = edge##_mask_bits_b; \
3601 block->fb_ptr = fb_ptr + 8 + 1024; \
3605 texture_offset += 0x10; \
3606 sub_tile_height--; \
3608 texture_offset += 0xF00; \
3609 psx_gpu->num_blocks = num_blocks; \
3613 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3614 texture_offset = texture_offset_base + 8; \
3617 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3618 texture_offset = texture_offset_base \
3620 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3621 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3623 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3624 texture_offset = texture_offset_base \
3626 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3629 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3631 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3632 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3634 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3637 #define setup_sprite_offset_u_adjust() \
3639 #define setup_sprite_comapre_left_block_mask() \
3640 ((left_block_mask & 0xFF) == 0xFF) \
3642 #define setup_sprite_comapre_right_block_mask() \
3643 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3646 #define setup_sprite_offset_u_adjust_4x() \
3648 offset_u_right = offset_u_right * 2 + 1 \
3650 #define setup_sprite_comapre_left_block_mask_4x() \
3651 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3653 #define setup_sprite_comapre_right_block_mask_4x() \
3654 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3657 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3658 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3659 s32 u, s32 v, s32 width, s32 height, u32 color) \
3661 s32 offset_u = u & 0xF; \
3662 s32 offset_v = v & 0xF; \
3664 s32 width_rounded = offset_u + width + 15; \
3665 s32 height_rounded = offset_v + height + 15; \
3666 s32 tile_height = height_rounded / 16; \
3667 s32 tile_width = width_rounded / 16; \
3668 u32 offset_u_right = width_rounded & 0xF; \
3670 setup_sprite_offset_u_adjust##x4mode(); \
3672 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3673 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3675 u32 left_mask_bits; \
3676 u32 right_mask_bits; \
3678 u32 sub_tile_height; \
3681 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3682 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3683 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3684 ((psx_gpu->texture_mask_height >> 4) << 12); \
3685 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3686 ((v & 0xF0) << 8); \
3687 u32 texture_offset_base = texture_offset; \
3690 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3691 u32 num_blocks = psx_gpu->num_blocks; \
3692 block_struct *block = psx_gpu->blocks + num_blocks; \
3694 u8 *texture_block_ptr; \
3697 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3699 control_mask = tile_width == 1; \
3700 control_mask |= (tile_height == 1) << 1; \
3701 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3702 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3704 sprites_##texture_mode++; \
3706 switch(control_mask) \
3710 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3715 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3720 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3725 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3730 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3735 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3740 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3745 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3750 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3755 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3760 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3765 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3770 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3775 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3782 setup_sprite_tiled_builder(4bpp,);
3783 setup_sprite_tiled_builder(8bpp,);
3785 setup_sprite_tiled_builder(4bpp,_4x);
3786 setup_sprite_tiled_builder(8bpp,_4x);
3789 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3791 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3792 s32 v, s32 width, s32 height, u32 color)
3794 u32 left_offset = u & 0x7;
3795 u32 width_rounded = width + left_offset + 7;
3797 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3798 u32 right_width = width_rounded & 0x7;
3799 u32 block_width = width_rounded / 8;
3800 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3802 u32 left_mask_bits = ~(0xFF << left_offset);
3803 u32 right_mask_bits = 0xFE << right_width;
3805 u32 texture_offset_base = u + (v * 1024);
3807 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3809 u32 blocks_remaining;
3810 u32 num_blocks = psx_gpu->num_blocks;
3811 block_struct *block = psx_gpu->blocks + num_blocks;
3813 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3814 u16 *texture_block_ptr;
3816 texture_offset_base &= ~0x7;
3818 stats_add(sprites_16bpp, 1);
3820 if(block_width == 1)
3822 u32 mask_bits = left_mask_bits | right_mask_bits;
3829 if(num_blocks > MAX_BLOCKS)
3831 flush_render_block_buffer(psx_gpu);
3833 block = psx_gpu->blocks;
3837 texture_page_ptr + (texture_offset_base & texture_mask);
3839 block->texels = *(vec_8x16u *)texture_block_ptr;
3840 block->draw_mask_bits = mask_bits;
3841 block->fb_ptr = fb_ptr;
3845 texture_offset_base += 1024;
3849 psx_gpu->num_blocks = num_blocks;
3858 blocks_remaining = block_width - 2;
3859 num_blocks += block_width;
3860 sprite_blocks += block_width;
3862 if(num_blocks > MAX_BLOCKS)
3864 flush_render_block_buffer(psx_gpu);
3865 num_blocks = block_width;
3866 block = psx_gpu->blocks;
3869 texture_offset = texture_offset_base;
3870 texture_offset_base += 1024;
3872 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3873 block->texels = *(vec_8x16u *)texture_block_ptr;
3875 block->draw_mask_bits = left_mask_bits;
3876 block->fb_ptr = fb_ptr;
3878 texture_offset += 8;
3882 while(blocks_remaining)
3884 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3885 block->texels = *(vec_8x16u *)texture_block_ptr;
3887 block->draw_mask_bits = 0;
3888 block->fb_ptr = fb_ptr;
3890 texture_offset += 8;
3897 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3898 block->texels = *(vec_8x16u *)texture_block_ptr;
3900 block->draw_mask_bits = right_mask_bits;
3901 block->fb_ptr = fb_ptr;
3903 fb_ptr += fb_ptr_pitch;
3907 psx_gpu->num_blocks = num_blocks;
3916 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3917 s32 v, s32 width, s32 height, u32 color)
3919 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
3920 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
3921 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
3923 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
3927 u32 right_width = ((width - 1) & 0x7) + 1;
3928 u32 right_mask_bits = (0xFF << right_width);
3929 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3930 u32 block_width = (width + 7) / 8;
3931 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3932 u32 blocks_remaining;
3933 u32 num_blocks = psx_gpu->num_blocks;
3934 block_struct *block = psx_gpu->blocks + num_blocks;
3936 u32 color_r = color & 0xFF;
3937 u32 color_g = (color >> 8) & 0xFF;
3938 u32 color_b = (color >> 16) & 0xFF;
3940 vec_8x16u right_mask;
3941 vec_8x16u test_mask = psx_gpu->test_mask;
3942 vec_8x16u zero_mask;
3944 sprites_untextured++;
3946 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3948 dup_8x16b(colors, color);
3949 dup_8x16b(zero_mask, 0x00);
3950 dup_8x16b(right_mask, right_mask_bits);
3951 tst_8x16b(right_mask, right_mask, test_mask);
3955 blocks_remaining = block_width - 1;
3956 num_blocks += block_width;
3959 sprite_blocks += block_width;
3962 if(num_blocks > MAX_BLOCKS)
3964 flush_render_block_buffer(psx_gpu);
3965 num_blocks = block_width;
3966 block = psx_gpu->blocks;
3969 while(blocks_remaining)
3971 block->pixels = colors;
3972 block->draw_mask = zero_mask;
3973 block->fb_ptr = fb_ptr;
3980 block->pixels = colors;
3981 block->draw_mask = right_mask;
3982 block->fb_ptr = fb_ptr;
3985 fb_ptr += fb_ptr_pitch;
3988 psx_gpu->num_blocks = num_blocks;
3994 void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y,
3995 s32 u, s32 v, s32 width, s32 height, u32 color)
3997 u32 r = color & 0xFF;
3998 u32 g = (color >> 8) & 0xFF;
3999 u32 b = (color >> 16) & 0xFF;
4000 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4002 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4004 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4009 if(psx_gpu->num_blocks > MAX_BLOCKS)
4011 flush_render_block_buffer(psx_gpu);
4018 vram_ptr = (void *)vram_ptr16;
4019 if((long)vram_ptr16 & 2)
4021 *vram_ptr16 = color_32bpp;
4022 vram_ptr = (void *)(vram_ptr16 + 1);
4026 while(num_width >= 4 * 2)
4028 vram_ptr[0] = color_32bpp;
4029 vram_ptr[1] = color_32bpp;
4030 vram_ptr[2] = color_32bpp;
4031 vram_ptr[3] = color_32bpp;
4037 while(num_width >= 2)
4039 *vram_ptr++ = color_32bpp;
4045 *(u16 *)vram_ptr = color_32bpp;
4054 #define setup_sprite_blocks_switch_textured(texture_mode) \
4055 setup_sprite_##texture_mode \
4057 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4058 setup_sprite_untextured \
4060 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4061 setup_sprite_blocks_switch_##texturing(texture_mode) \
4064 #define texture_sprite_blocks_switch_4bpp() \
4065 texture_blocks_untextured \
4067 #define texture_sprite_blocks_switch_8bpp() \
4068 texture_sprite_blocks_8bpp \
4070 #define texture_sprite_blocks_switch_16bpp() \
4071 texture_blocks_untextured \
4073 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4074 texture_blocks_untextured \
4076 #define texture_sprite_blocks_switch_textured(texture_mode) \
4077 texture_sprite_blocks_switch_##texture_mode() \
4079 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4080 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4082 setup_sprite_blocks_switch(texturing, texture_mode), \
4083 texture_sprite_blocks_switch_##texturing(texture_mode), \
4084 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4086 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4089 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4090 mask_evaluate, shading, dithering, texturing, blending) \
4091 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4092 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4093 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4094 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4096 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4097 mask_evaluate, shading, dithering, texturing) \
4098 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4099 mask_evaluate, shading, dithering, texturing, unblended), \
4100 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4101 mask_evaluate, shading, dithering, texturing, blended) \
4103 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4104 mask_evaluate, shading, dithering) \
4105 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4106 mask_evaluate, shading, dithering, untextured), \
4107 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4108 mask_evaluate, shading, dithering, textured) \
4110 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4111 mask_evaluate, shading) \
4112 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4113 mask_evaluate, shading, undithered), \
4114 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4115 mask_evaluate, shading, dithered) \
4117 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4118 blend_mode, mask_evaluate) \
4119 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4120 mask_evaluate, unshaded), \
4121 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4122 mask_evaluate, shaded) \
4124 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4125 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4127 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4130 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4131 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4132 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4133 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4134 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4136 #define render_sprite_blocks_switch_block() \
4137 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4138 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4139 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4140 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4143 render_block_handler_struct render_sprite_block_handlers[] =
4145 render_sprite_blocks_switch_block()
4149 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4150 s32 width, s32 height, u32 flags, u32 color)
4152 s32 x_right = x + width - 1;
4153 s32 y_bottom = y + height - 1;
4159 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4165 if(x < psx_gpu->viewport_start_x)
4167 u32 clip = psx_gpu->viewport_start_x - x;
4173 if(y < psx_gpu->viewport_start_y)
4175 s32 clip = psx_gpu->viewport_start_y - y;
4181 if(x_right > psx_gpu->viewport_end_x)
4182 width -= x_right - psx_gpu->viewport_end_x;
4184 if(y_bottom > psx_gpu->viewport_end_y)
4185 height -= y_bottom - psx_gpu->viewport_end_y;
4187 if((width <= 0) || (height <= 0))
4191 span_pixels += width * height;
4195 u32 render_state = flags &
4196 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4197 RENDER_FLAGS_TEXTURE_MAP);
4199 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4201 if((psx_gpu->render_state != render_state) ||
4202 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4204 psx_gpu->render_state = render_state;
4205 flush_render_block_buffer(psx_gpu);
4211 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4215 if(psx_gpu->triangle_color != color)
4217 flush_render_block_buffer(psx_gpu);
4218 psx_gpu->triangle_color = color;
4221 if(color == 0x808080)
4222 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4224 render_block_handler_struct *render_block_handler =
4225 &(render_sprite_block_handlers[render_state]);
4226 psx_gpu->render_block_handler = render_block_handler;
4228 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4229 (psx_gpu, x, y, u, v, width, height, color);
4232 #define draw_pixel_line_mask_evaluate_yes() \
4233 if((*vram_ptr & 0x8000) == 0) \
4235 #define draw_pixel_line_mask_evaluate_no() \
4238 #define draw_pixel_line_shaded() \
4240 color_r = fixed_to_int(current_r); \
4241 color_g = fixed_to_int(current_g); \
4242 color_b = fixed_to_int(current_b); \
4244 current_r += gradient_r; \
4245 current_g += gradient_g; \
4246 current_b += gradient_b; \
4249 #define draw_pixel_line_unshaded() \
4251 color_r = color & 0xFF; \
4252 color_g = (color >> 8) & 0xFF; \
4253 color_b = (color >> 16) & 0xFF; \
4257 #define draw_pixel_line_dithered(_x, _y) \
4259 u32 dither_xor = _x ^ _y; \
4260 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4261 dither_offset |= (_y & 0x1) << 1; \
4262 dither_offset |= (dither_xor & 0x1) << 2; \
4263 dither_offset -= 4; \
4265 color_r += dither_offset; \
4266 color_g += dither_offset; \
4267 color_b += dither_offset; \
4288 #define draw_pixel_line_undithered(_x, _y) \
4291 #define draw_pixel_line_average() \
4292 color_r = (color_r + fb_r) / 2; \
4293 color_g = (color_g + fb_g) / 2; \
4294 color_b = (color_b + fb_b) / 2 \
4296 #define draw_pixel_line_add() \
4311 #define draw_pixel_line_subtract() \
4312 color_r = fb_r - color_r; \
4313 color_g = fb_g - color_g; \
4314 color_b = fb_b - color_b; \
4325 #define draw_pixel_line_add_fourth() \
4326 color_r = fb_r + (color_r / 4); \
4327 color_g = fb_g + (color_g / 4); \
4328 color_b = fb_b + (color_b / 4); \
4340 #define draw_pixel_line_blended(blend_mode) \
4341 s32 fb_pixel = *vram_ptr; \
4342 s32 fb_r = fb_pixel & 0x1F; \
4343 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4344 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4346 draw_pixel_line_##blend_mode() \
4348 #define draw_pixel_line_unblended(blend_mode) \
4351 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4353 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4354 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4356 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4358 draw_pixel_line_##shading(); \
4359 draw_pixel_line_##dithering(_x, _y); \
4365 draw_pixel_line_##blending(blend_mode); \
4367 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4368 psx_gpu->mask_msb; \
4372 #define update_increment(value) \
4375 #define update_decrement(value) \
4378 #define update_vram_row_increment(value) \
4381 #define update_vram_row_decrement(value) \
4384 #define compare_increment(a, b) \
4387 #define compare_decrement(a, b) \
4390 #define set_line_gradients(minor) \
4392 s32 gradient_divisor = delta_##minor; \
4393 if(gradient_divisor != 0) \
4395 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4396 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4397 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4405 current_r = fixed_center(vertex_a->r); \
4406 current_g = fixed_center(vertex_a->g); \
4407 current_b = fixed_center(vertex_a->b); \
4410 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4411 mask_evaluate, blend_mode) \
4414 error_step = delta_y * 2; \
4415 error_wrap = delta_x * 2; \
4419 set_line_gradients(x); \
4421 for(current_x = x_a; current_x <= x_b; current_x++) \
4423 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4424 mask_evaluate, blend_mode); \
4425 error += error_step; \
4428 if(error >= error_wrap) \
4430 update_##direction(current_y); \
4431 update_vram_row_##direction(); \
4432 error -= error_wrap; \
4437 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4438 mask_evaluate, blend_mode) \
4441 error_step = delta_x * 2; \
4442 error_wrap = delta_y * 2; \
4446 set_line_gradients(y); \
4448 for(current_y = y_a; compare_##direction(current_y, y_b); \
4449 update_##direction(current_y)) \
4451 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4452 mask_evaluate, blend_mode); \
4453 error += error_step; \
4454 update_vram_row_##direction(); \
4456 if(error > error_wrap) \
4460 error -= error_wrap; \
4466 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4472 if(delta_x > delta_y) \
4474 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4475 mask_evaluate, blend_mode); \
4479 draw_line_span_vertical(decrement, shading, blending, dithering, \
4480 mask_evaluate, blend_mode); \
4485 if(delta_x > delta_y) \
4487 draw_line_span_horizontal(increment, shading, blending, dithering, \
4488 mask_evaluate, blend_mode); \
4492 draw_line_span_vertical(increment, shading, blending, dithering, \
4493 mask_evaluate, blend_mode); \
4498 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4499 u32 color, int double_resolution)
4501 s32 color_r, color_g, color_b;
4502 u32 triangle_winding = 0;
4504 fixed_type gradient_r = 0;
4505 fixed_type gradient_g = 0;
4506 fixed_type gradient_b = 0;
4507 fixed_type current_r = 0;
4508 fixed_type current_g = 0;
4509 fixed_type current_b = 0;
4514 s32 delta_x, delta_y;
4525 flush_render_block_buffer(psx_gpu);
4526 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4528 vertex_struct *vertex_a = &(vertexes[0]);
4529 vertex_struct *vertex_b = &(vertexes[1]);
4537 if(vertex_a->x >= vertex_b->x)
4539 vertex_swap(vertex_a, vertex_b);
4548 delta_x = x_b - x_a;
4549 delta_y = y_b - y_a;
4551 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4554 if(double_resolution)
4564 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4566 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4570 if(flags & RENDER_FLAGS_SHADE)
4571 control_mask |= 0x1;
4573 if(flags & RENDER_FLAGS_BLEND)
4575 control_mask |= 0x2;
4576 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4579 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4580 control_mask |= 0x4;
4582 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4583 control_mask |= 0x8;
4585 switch(control_mask)
4588 render_line_body(unshaded, unblended, undithered, no, none);
4592 render_line_body(shaded, unblended, undithered, no, none);
4596 render_line_body(unshaded, blended, undithered, no, average);
4600 render_line_body(shaded, blended, undithered, no, average);
4604 render_line_body(unshaded, unblended, dithered, no, none);
4608 render_line_body(shaded, unblended, dithered, no, none);
4612 render_line_body(unshaded, blended, dithered, no, average);
4616 render_line_body(shaded, blended, dithered, no, average);
4620 render_line_body(unshaded, unblended, undithered, yes, none);
4624 render_line_body(shaded, unblended, undithered, yes, none);
4628 render_line_body(unshaded, blended, undithered, yes, average);
4632 render_line_body(shaded, blended, undithered, yes, average);
4636 render_line_body(unshaded, unblended, dithered, yes, none);
4640 render_line_body(shaded, unblended, dithered, yes, none);
4644 render_line_body(unshaded, blended, dithered, yes, average);
4648 render_line_body(shaded, blended, dithered, yes, average);
4652 render_line_body(unshaded, blended, undithered, no, add);
4656 render_line_body(shaded, blended, undithered, no, add);
4660 render_line_body(unshaded, blended, dithered, no, add);
4664 render_line_body(shaded, blended, dithered, no, add);
4668 render_line_body(unshaded, blended, undithered, yes, add);
4672 render_line_body(shaded, blended, undithered, yes, add);
4676 render_line_body(unshaded, blended, dithered, yes, add);
4680 render_line_body(shaded, blended, dithered, yes, add);
4684 render_line_body(unshaded, blended, undithered, no, subtract);
4688 render_line_body(shaded, blended, undithered, no, subtract);
4692 render_line_body(unshaded, blended, dithered, no, subtract);
4696 render_line_body(shaded, blended, dithered, no, subtract);
4700 render_line_body(unshaded, blended, undithered, yes, subtract);
4704 render_line_body(shaded, blended, undithered, yes, subtract);
4708 render_line_body(unshaded, blended, dithered, yes, subtract);
4712 render_line_body(shaded, blended, dithered, yes, subtract);
4716 render_line_body(unshaded, blended, undithered, no, add_fourth);
4720 render_line_body(shaded, blended, undithered, no, add_fourth);
4724 render_line_body(unshaded, blended, dithered, no, add_fourth);
4728 render_line_body(shaded, blended, dithered, no, add_fourth);
4732 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4736 render_line_body(shaded, blended, undithered, yes, add_fourth);
4740 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4744 render_line_body(shaded, blended, dithered, yes, add_fourth);
4750 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4751 u32 width, u32 height)
4753 if((width == 0) || (height == 0))
4756 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4758 u32 r = color & 0xFF;
4759 u32 g = (color >> 8) & 0xFF;
4760 u32 b = (color >> 16) & 0xFF;
4761 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4763 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4765 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4767 u32 pitch = 512 - (width / 2);
4770 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4775 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4784 vram_ptr[0] = color_32bpp;
4785 vram_ptr[1] = color_32bpp;
4786 vram_ptr[2] = color_32bpp;
4787 vram_ptr[3] = color_32bpp;
4788 vram_ptr[4] = color_32bpp;
4789 vram_ptr[5] = color_32bpp;
4790 vram_ptr[6] = color_32bpp;
4791 vram_ptr[7] = color_32bpp;
4802 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4803 u32 width, u32 height)
4805 if((width == 0) || (height == 0))
4811 u32 r = color & 0xFF;
4812 u32 g = (color >> 8) & 0xFF;
4813 u32 b = (color >> 16) & 0xFF;
4814 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4816 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4818 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4820 u32 pitch = 1024 / 2 - (width / 2);
4828 vram_ptr[0] = color_32bpp;
4829 vram_ptr[1] = color_32bpp;
4830 vram_ptr[2] = color_32bpp;
4831 vram_ptr[3] = color_32bpp;
4832 vram_ptr[4] = color_32bpp;
4833 vram_ptr[5] = color_32bpp;
4834 vram_ptr[6] = color_32bpp;
4835 vram_ptr[7] = color_32bpp;
4846 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4847 u32 width, u32 height, u32 pitch)
4849 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4851 u32 mask_msb = psx_gpu->mask_msb;
4853 if((width == 0) || (height == 0))
4856 flush_render_block_buffer(psx_gpu);
4857 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4859 for(draw_y = 0; draw_y < height; draw_y++)
4861 for(draw_x = 0; draw_x < width; draw_x++)
4863 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4871 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4872 u32 dest_x, u32 dest_y, u32 width, u32 height)
4874 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4875 dest_x, dest_y, width, height, 1024);
4879 void initialize_reciprocal_table(void)
4882 u32 height_normalized;
4883 u32 height_reciprocal;
4886 for(height = 1; height < sizeof(reciprocal_table)
4887 / sizeof(reciprocal_table[0]); height++)
4889 shift = __builtin_clz(height);
4890 height_normalized = height << shift;
4891 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4894 shift = 32 - (51 - shift);
4896 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4901 #define dither_table_row(a, b, c, d) \
4902 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4904 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4906 vec_8x16u test_mask =
4907 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4909 psx_gpu->test_mask = test_mask;
4911 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4912 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4913 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4914 psx_gpu->viewport_mask = 0;
4915 psx_gpu->current_texture_page = 0;
4916 psx_gpu->current_texture_mask = 0;
4917 psx_gpu->last_8bpp_texture_page = 0;
4919 psx_gpu->clut_settings = 0;
4920 psx_gpu->texture_settings = 0;
4921 psx_gpu->render_state = 0;
4922 psx_gpu->render_state_base = 0;
4923 psx_gpu->num_blocks = 0;
4924 psx_gpu->uvrgb_phase = 0x8000;
4926 psx_gpu->vram_ptr = vram;
4927 psx_gpu->vram_out_ptr = vram;
4929 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4930 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4931 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4933 psx_gpu->mask_msb = 0;
4935 psx_gpu->texture_window_x = 0;
4936 psx_gpu->texture_window_y = 0;
4937 psx_gpu->texture_mask_width = 0xFF;
4938 psx_gpu->texture_mask_height = 0xFF;
4940 psx_gpu->render_mode = 0;
4942 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4944 initialize_reciprocal_table();
4945 psx_gpu->reciprocal_table_ptr = reciprocal_table;
4952 // (minus ones(4) * 4)
4954 // d0: (1 3 5 7): x1 ^ y1
4955 // d1: (2 3 6 7): y0
4956 // d2: (4 5 6 7): x0 ^ y0
4958 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
4959 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
4960 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
4961 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
4963 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
4965 psx_gpu->enhancement_x_threshold = 256;
4971 gettimeofday(&tv, NULL);
4973 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
4976 #if 0 //def NEON_BUILD
4981 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
4986 void init_counter(void)
4989 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
4990 value |= 5; // master enable, ccnt reset
4991 value &= ~8; // ccnt divider 0
4992 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
4993 // enable cycle counter
4994 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
4997 void triangle_benchmark(psx_gpu_struct *psx_gpu)
5004 const u32 iterations = 500000;
5006 psx_gpu->num_blocks = 64;
5007 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5009 for(i = 0; i < 64; i++)
5011 memset(&(psx_gpu->blocks[i].r), 0, 16);
5016 ticks = get_counter();
5018 for(i = 0; i < iterations; i++)
5020 texture_sprite_blocks_8bpp(psx_gpu);
5023 ticks_elapsed = get_counter() - ticks;
5025 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5030 #include "psx_gpu_4x.c"