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.
24 #include "vector_ops.h"
26 #include "psx_gpu_simd.h"
27 #include "psx_gpu_offsets.h"
30 void dump_r_d(const char *name, void *dump);
31 void dump_r_q(const char *name, void *dump);
32 #define dumprd(n) dump_r_d(#n, n.e)
33 #define dumprq(n) dump_r_q(#n, n.e)
37 u32 span_pixel_blocks = 0;
43 u32 sprites_16bpp = 0;
44 u32 sprite_blocks = 0;
45 u32 sprites_untextured = 0;
47 u32 trivial_rejects = 0;
51 u32 texel_blocks_4bpp = 0;
52 u32 texel_blocks_8bpp = 0;
53 u32 texel_blocks_16bpp = 0;
54 u32 texel_blocks_untextured = 0;
56 u32 render_buffer_flushes = 0;
57 u32 state_changes = 0;
58 u32 left_split_triangles = 0;
59 u32 flat_triangles = 0;
60 u32 clipped_triangles = 0;
61 u32 zero_block_spans = 0;
62 u32 texture_cache_loads = 0;
63 u32 false_modulated_blocks = 0;
65 #define stats_add(stat, count) // stat += count
67 /* double size for enhancement */
68 u32 reciprocal_table[512 * 2];
71 typedef s32 fixed_type;
73 #define EDGE_STEP_BITS 32
76 #define fixed_center(value) \
77 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
79 #define int_to_fixed(value) \
80 (((fixed_type)(value)) << FIXED_BITS) \
82 #define fixed_to_int(value) \
83 ((value) >> FIXED_BITS) \
85 #define fixed_to_double(value) \
86 ((value) / (double)(1 << FIXED_BITS)) \
88 #define double_to_fixed(value) \
89 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
91 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
92 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
93 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
94 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
96 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
97 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
99 struct render_block_handler_struct
102 texture_blocks_function_type *texture_blocks;
103 shade_blocks_function_type *shade_blocks;
104 blend_blocks_function_type *blend_blocks;
109 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
111 u32 shift = __builtin_clz(denominator);
112 u32 denominator_normalized = denominator << shift;
114 double numerator = (1ULL << 62) + denominator_normalized;
117 double denominator_normalized_dp_b;
118 u64 denominator_normalized_dp_u64;
121 double reciprocal_dp;
123 u64 numerator_u64 = (denominator_normalized >> 10) |
124 ((u64)(62 + 1023) << 52);
125 *((u64 *)(&numerator_b)) = numerator_u64;
127 denominator_normalized_dp_u64 =
128 (u64)(denominator_normalized << 21) |
129 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
130 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
132 // Implement with a DP divide
133 reciprocal_dp = numerator / denominator_normalized_dp_b;
134 reciprocal = reciprocal_dp;
136 if(reciprocal == 0x80000001)
137 reciprocal = 0x80000000;
139 *_shift = 62 - shift;
143 double reciprocal_estimate(double a)
148 q = (int)(a * 512.0);
149 /* a in units of 1/512 rounded down */
150 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
151 s = (int)(256.0 * r + 0.5);
153 /* r in units of 1/256 rounded to nearest */
155 return (double)s / 256.0;
158 u32 reciprocal_estimate_u32(u32 value)
161 volatile double dp_value;
162 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
164 if((value >> 31) == 0)
167 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
169 *dp_value_ptr = dp_value_u64;
171 dp_value = reciprocal_estimate(dp_value);
172 dp_value_u64 = *dp_value_ptr;
174 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
177 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
179 u32 shift = __builtin_clz(value);
180 u32 value_normalized = value << shift;
182 *_shift = 62 - shift;
184 value_normalized -= 2;
186 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
188 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
189 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
190 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
191 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
192 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
193 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
195 return reciprocal_normalized;
201 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
203 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
206 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
208 s32 coverage_x, coverage_y;
213 coverage_x = x2 >> 6;
214 coverage_y = y2 >> 8;
222 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
225 mask_down_right |= mask_down_right << 16;
227 coverage_x = x1 >> 6;
229 mask_up_left = 0xFFFF0000 << coverage_x;
231 mask_up_left = 0xFFFF0000;
233 coverage_y = y1 >> 8;
235 mask_up_left |= mask_up_left >> 16;
237 return mask_up_left & mask_down_right;
240 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
243 u32 mask = texture_region_mask(x1, y1, x2, y2);
245 psx_gpu->dirty_textures_4bpp_mask |= mask;
246 psx_gpu->dirty_textures_8bpp_mask |= mask;
247 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
252 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
253 u32 y1, u32 x2, u32 y2)
255 u32 mask = texture_region_mask(x1, y1, x2, y2) &
256 psx_gpu->viewport_mask;
258 psx_gpu->dirty_textures_4bpp_mask |= mask;
259 psx_gpu->dirty_textures_8bpp_mask |= mask;
260 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
265 static void update_texture_cache_region_(psx_gpu_struct *psx_gpu,
266 u32 x1, u32 y1, u32 x2, u32 y2)
268 u32 mask = texture_region_mask(x1, y1, x2, y2);
270 u8 *texture_page_ptr;
275 psx_gpu->dirty_textures_8bpp_mask |= mask;
276 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
278 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
279 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
281 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
282 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
283 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
284 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
292 texel_block = *vram_ptr;
294 texture_page_ptr[0] = texel_block & 0xF;
295 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
296 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
297 texture_page_ptr[3] = texel_block >> 12;
300 texture_page_ptr += 4;
314 psx_gpu->dirty_textures_4bpp_mask |= mask;
318 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
327 update_texture_cache_region_(psx_gpu, x1, y1, x2, y2);
331 while (unlikely(w > 0));
336 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
338 u32 current_texture_page = psx_gpu->current_texture_page;
339 u8 *texture_page_ptr = psx_gpu->texture_page_base;
340 u16 *vram_ptr = psx_gpu->vram_ptr;
346 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
347 vram_ptr += (current_texture_page & 0xF) * 64;
349 texture_cache_loads++;
356 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
366 texel_block = *vram_ptr;
368 texture_page_ptr[0] = texel_block & 0xF;
369 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
370 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
371 texture_page_ptr[3] = texel_block >> 12;
374 texture_page_ptr += 4;
388 vram_ptr -= (1024 * 16) - 4;
394 vram_ptr += (16 * 1024) - (4 * 16);
399 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
402 u16 *texture_page_ptr = psx_gpu->texture_page_base;
403 u16 *vram_ptr = psx_gpu->vram_ptr;
410 texture_cache_loads++;
412 vram_ptr += (texture_page >> 4) * 256 * 1024;
413 vram_ptr += (texture_page & 0xF) * 64;
415 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
416 texture_page_ptr += (8 * 16) * 8;
429 load_128b(texels, vram_ptr);
430 store_128b(texels, texture_page_ptr);
432 texture_page_ptr += 8;
440 vram_ptr -= (1024 * 16);
449 vram_ptr += (16 * 1024);
451 texture_page_ptr += (8 * 16) * 8;
459 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
461 u32 current_texture_page = psx_gpu->current_texture_page;
462 u32 update_textures =
463 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
465 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
467 if(update_textures & (1 << current_texture_page))
469 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
470 update_textures &= ~(1 << current_texture_page);
475 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
476 (current_texture_page & 0x10);
478 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
482 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
484 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
485 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
487 u32 num_blocks_dest = 0;
488 block_struct *block_src = psx_gpu->blocks;
489 block_struct *block_dest = psx_gpu->blocks;
491 u16 *vram_ptr = psx_gpu->vram_ptr;
494 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
496 for(i = 0; i < psx_gpu->num_blocks; i++)
498 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
499 if(fb_offset & (1 << 11))
501 *block_dest = *block_src;
510 for(i = 0; i < psx_gpu->num_blocks; i++)
512 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
513 if((fb_offset & (1 << 11)) == 0)
515 *block_dest = *block_src;
523 psx_gpu->num_blocks = num_blocks_dest;
526 if(psx_gpu->num_blocks)
528 render_block_handler_struct *render_block_handler =
529 psx_gpu->render_block_handler;
531 render_block_handler->texture_blocks(psx_gpu);
532 render_block_handler->shade_blocks(psx_gpu);
533 render_block_handler->blend_blocks(psx_gpu);
536 span_pixel_blocks += psx_gpu->num_blocks;
537 render_buffer_flushes++;
540 psx_gpu->num_blocks = 0;
547 #define setup_gradient_calculation_input(set, vertex) \
548 /* First type is: uvrg bxxx xxxx */\
549 /* Second type is: yyyy ybyy uvrg */\
550 /* Since x_a and y_c are the same the same variable is used for both. */\
551 x##set##_a_y##set##_c.e[0] = vertex->u; \
552 x##set##_a_y##set##_c.e[1] = vertex->v; \
553 x##set##_a_y##set##_c.e[2] = vertex->r; \
554 x##set##_a_y##set##_c.e[3] = vertex->g; \
555 dup_4x16b(x##set##_b, vertex->x); \
556 dup_4x16b(x##set##_c, vertex->x); \
557 dup_4x16b(y##set##_a, vertex->y); \
558 dup_4x16b(y##set##_b, vertex->y); \
559 x##set##_b.e[0] = vertex->b; \
560 y##set##_b.e[1] = vertex->b \
563 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
564 vertex_struct *b, vertex_struct *c)
566 u32 triangle_area = psx_gpu->triangle_area;
567 u32 winding_mask_scalar;
569 u32 triangle_area_shift;
570 u64 triangle_area_reciprocal =
571 fixed_reciprocal(triangle_area, &triangle_area_shift);
572 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
574 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
575 // ( d0 * d1 ) - ( d2 * d3 ) =
576 // ( m0 ) - ( m1 ) = gradient
578 // This is split to do 12 elements at a time over three sets: a, b, and c.
579 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
580 // two of the slots are unused.
582 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
585 vec_4x16s x0_a_y0_c, x0_b, x0_c;
586 vec_4x16s y0_a, y0_b;
587 vec_4x16s x1_a_y1_c, x1_b, x1_c;
588 vec_4x16s y1_a, y1_b;
589 vec_4x16s x2_a_y2_c, x2_b, x2_c;
590 vec_4x16s y2_a, y2_b;
594 vec_4x32u uvrgb_phase;
596 vec_4x16s d0_a_d3_c, d0_b, d0_c;
597 vec_4x16s d1_a, d1_b, d1_c_d2_a;
598 vec_4x16s d2_b, d2_c;
599 vec_4x16s d3_a, d3_b;
601 vec_4x32s m0_a, m0_b, m0_c;
602 vec_4x32s m1_a, m1_b, m1_c;
604 vec_4x32u gradient_area_a, gradient_area_c;
605 vec_2x32u gradient_area_b;
607 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
608 vec_2x32u gradient_area_sign_b;
609 vec_4x32u winding_mask;
611 vec_2x64u gradient_wide_a0, gradient_wide_a1;
612 vec_2x64u gradient_wide_c0, gradient_wide_c1;
613 vec_2x64u gradient_wide_b;
615 vec_4x32u gradient_a, gradient_c;
616 vec_2x32u gradient_b;
617 vec_16x8s gradient_shift;
619 setup_gradient_calculation_input(0, a);
620 setup_gradient_calculation_input(1, b);
621 setup_gradient_calculation_input(2, c);
623 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
624 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
625 shl_long_4x16b(b_base, x0_b, 16);
627 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
628 add_4x32b(b_base, b_base, uvrgb_phase);
630 // Can probably pair these, but it'll require careful register allocation
631 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
632 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
634 sub_4x16b(d0_b, x1_b, x0_b);
635 sub_4x16b(d0_c, x1_c, x0_c);
637 sub_4x16b(d1_a, y2_a, y1_a);
638 sub_4x16b(d1_b, y2_b, y1_b);
640 sub_4x16b(d2_b, x2_b, x1_b);
641 sub_4x16b(d2_c, x2_c, x1_c);
643 sub_4x16b(d3_a, y1_a, y0_a);
644 sub_4x16b(d3_b, y1_b, y0_b);
646 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
647 mul_long_4x16b(m0_b, d0_b, d1_b);
648 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
650 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
651 mul_long_4x16b(m1_b, d2_b, d3_b);
652 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
654 sub_4x32b(gradient_area_a, m0_a, m1_a);
655 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
656 sub_4x32b(gradient_area_c, m0_c, m1_c);
658 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
659 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
660 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
662 abs_4x32b(gradient_area_a, gradient_area_a);
663 abs_2x32b(gradient_area_b, gradient_area_b);
664 abs_4x32b(gradient_area_c, gradient_area_c);
666 winding_mask_scalar = -psx_gpu->triangle_winding;
668 dup_4x32b(winding_mask, winding_mask_scalar);
669 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
670 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
671 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
673 mul_scalar_long_2x32b(gradient_wide_a0,
674 vector_cast(vec_2x32s, gradient_area_a.low),
675 (s64)triangle_area_reciprocal);
676 mul_scalar_long_2x32b(gradient_wide_a1,
677 vector_cast(vec_2x32s, gradient_area_a.high),
678 (s64)triangle_area_reciprocal);
679 mul_scalar_long_2x32b(gradient_wide_b,
680 vector_cast(vec_2x32s, gradient_area_b),
681 (s64)triangle_area_reciprocal);
682 mul_scalar_long_2x32b(gradient_wide_c0,
683 vector_cast(vec_2x32s, gradient_area_c.low),
684 (s64)triangle_area_reciprocal);
685 mul_scalar_long_2x32b(gradient_wide_c1,
686 vector_cast(vec_2x32s, gradient_area_c.high),
687 (s64)triangle_area_reciprocal);
689 dup_16x8b(gradient_shift, triangle_area_shift);
690 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
691 vector_cast(vec_2x64u, gradient_shift));
692 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
693 vector_cast(vec_2x64u, gradient_shift));
694 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
695 vector_cast(vec_2x64u, gradient_shift));
696 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
697 vector_cast(vec_2x64u, gradient_shift));
698 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
699 vector_cast(vec_2x64u, gradient_shift));
701 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
702 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
703 mov_narrow_2x64b(gradient_b, gradient_wide_b);
704 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
705 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
707 shl_4x32b(gradient_a, gradient_a, 4);
708 shl_2x32b(gradient_b, gradient_b, 4);
709 shl_4x32b(gradient_c, gradient_c, 4);
711 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
712 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
713 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
715 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
716 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
717 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
719 u32 left_adjust = a->x;
720 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
721 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
731 eor_4x32b(zero, zero, zero);
732 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
733 add_2x32b(b_dx2, gradient_b, gradient_b);
734 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
735 add_2x32b(b_dx3, gradient_b, b_dx2);
737 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
739 psx_gpu->u_block_span.e[0] = zero.e[0];
740 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
741 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
742 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
744 psx_gpu->v_block_span.e[0] = zero.e[1];
745 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
746 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
747 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
749 psx_gpu->r_block_span.e[0] = zero.e[2];
750 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
751 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
752 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
754 psx_gpu->g_block_span.e[0] = zero.e[3];
755 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
756 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
757 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
759 psx_gpu->b_block_span.e[0] = zero.e[0];
760 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
761 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
762 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
764 psx_gpu->uvrg = uvrg_base;
765 psx_gpu->b = b_base.e[0];
767 psx_gpu->uvrg_dx = gradient_a;
768 psx_gpu->uvrg_dy = gradient_c;
769 psx_gpu->b_dy = gradient_b.e[1];
773 #define vector_check(_a, _b) \
774 if(memcmp(&_a, &_b, sizeof(_b))) \
776 if(sizeof(_b) == 8) \
778 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
779 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
783 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
784 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
789 #define scalar_check(_a, _b) \
791 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
794 #if !defined(NEON_BUILD) && !defined(NDEBUG)
795 static void setup_spans_debug_check(psx_gpu_struct *psx_gpu,
796 edge_data_struct *span_edge_data_element)
798 u32 _num_spans = span_edge_data_element - psx_gpu->span_edge_data;
799 if (_num_spans > MAX_SPANS)
800 *(volatile int *)0 = 1;
801 if (_num_spans < psx_gpu->num_spans)
803 if(span_edge_data_element->num_blocks > MAX_BLOCKS_PER_ROW)
804 *(volatile int *)0 = 2;
805 if(span_edge_data_element->y >= 2048)
806 *(volatile int *)0 = 3;
810 #define setup_spans_debug_check(psx_gpu, span_edge_data_element)
813 #define setup_spans_prologue_alternate_yes() \
814 vec_2x64s alternate_x; \
815 vec_2x64s alternate_dx_dy; \
816 vec_4x32s alternate_x_32; \
817 vec_4x16u alternate_x_16; \
819 vec_4x16u alternate_select; \
820 vec_4x16s y_mid_point; \
824 s32 edge_dx_dy_alt; \
827 #define setup_spans_prologue_alternate_no() \
829 #define setup_spans_prologue(alternate_active) \
830 edge_data_struct *span_edge_data; \
831 vec_4x32u *span_uvrg_offset; \
832 u32 *span_b_offset; \
836 vec_2x64s edges_xy; \
837 vec_2x32s edges_dx_dy; \
838 vec_2x32u edge_shifts; \
840 vec_2x64s left_x, right_x; \
841 vec_2x64s left_dx_dy, right_dx_dy; \
842 vec_4x32s left_x_32, right_x_32; \
843 vec_8x16s left_right_x_16; \
845 vec_8x16s left_edge; \
846 vec_8x16s right_edge; \
847 vec_4x16u span_shift; \
851 vec_4x16u c_0xFFFE; \
854 vec_2x32s x_starts; \
863 vec_4x32u uvrg = psx_gpu->uvrg; \
864 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
865 u32 b = psx_gpu->b; \
866 u32 b_dy = psx_gpu->b_dy; \
868 dup_2x32b(c_0x01, 0x01); \
869 setup_spans_prologue_alternate_##alternate_active() \
871 #define setup_spans_prologue_b() \
872 span_edge_data = psx_gpu->span_edge_data; \
873 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
874 span_b_offset = psx_gpu->span_b_offset; \
876 vec_8x16u c_0x0001; \
877 vec_4x16u c_max_blocks_per_row; \
879 dup_8x16b(c_0x0001, 0x0001); \
880 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
881 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
882 add_8x16b(right_edge, right_edge, c_0x0001); \
883 dup_4x16b(c_0x04, 0x04); \
884 dup_4x16b(c_0x07, 0x07); \
885 dup_4x16b(c_0xFFFE, 0xFFFE); \
886 dup_4x16b(c_max_blocks_per_row, MAX_BLOCKS_PER_ROW); \
889 #define compute_edge_delta_x2() \
892 vec_2x32s height_reciprocals; \
893 vec_2x32s heights_b; \
896 u32 edge_shift = reciprocal_table[height]; \
898 dup_2x32b(heights, height); \
899 sub_2x32b(widths, x_ends, x_starts); \
901 dup_2x32b(edge_shifts, edge_shift); \
902 sub_2x32b(heights_b, heights, c_0x01); \
903 shr_2x32b(height_reciprocals, edge_shifts, 10); \
905 mla_2x32b(heights_b, x_starts, heights); \
906 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
907 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
908 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
911 #define compute_edge_delta_x3(start_c, height_a, height_b) \
914 vec_2x32s height_reciprocals; \
915 vec_2x32s heights_b; \
920 u32 height_reciprocal_alt; \
922 heights.e[0] = height_a; \
923 heights.e[1] = height_b; \
925 edge_shifts.e[0] = reciprocal_table[height_a]; \
926 edge_shifts.e[1] = reciprocal_table[height_b]; \
927 edge_shift_alt = reciprocal_table[height_minor_b]; \
929 sub_2x32b(widths, x_ends, x_starts); \
930 width_alt = x_c - start_c; \
932 shr_2x32b(height_reciprocals, edge_shifts, 10); \
933 height_reciprocal_alt = edge_shift_alt >> 10; \
935 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
936 edge_shift_alt &= 0x1F; \
938 sub_2x32b(heights_b, heights, c_0x01); \
939 height_b_alt = height_minor_b - 1; \
941 mla_2x32b(heights_b, x_starts, heights); \
942 height_b_alt += height_minor_b * start_c; \
944 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
945 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
947 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
948 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
952 #define setup_spans_adjust_y_up() \
953 sub_4x32b(y_x4, y_x4, c_0x04) \
955 #define setup_spans_adjust_y_down() \
956 add_4x32b(y_x4, y_x4, c_0x04) \
958 #define setup_spans_adjust_interpolants_up() \
959 sub_4x32b(uvrg, uvrg, uvrg_dy); \
962 #define setup_spans_adjust_interpolants_down() \
963 add_4x32b(uvrg, uvrg, uvrg_dy); \
967 #define setup_spans_clip_interpolants_increment() \
968 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
971 #define setup_spans_clip_interpolants_decrement() \
972 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
975 #define setup_spans_clip_alternate_yes() \
976 edge_alt += edge_dx_dy_alt * (s64)(clip) \
978 #define setup_spans_clip_alternate_no() \
980 #define setup_spans_clip(direction, alternate_active) \
982 clipped_triangles++; \
983 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
984 setup_spans_clip_alternate_##alternate_active(); \
985 setup_spans_clip_interpolants_##direction(); \
989 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
991 vec_2x64u edge_shifts_64; \
992 vec_2x64s edges_dx_dy_64; \
994 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
995 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
997 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
998 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
1000 left_x.e[0] = edges_xy.e[left_index]; \
1001 right_x.e[0] = edges_xy.e[right_index]; \
1003 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
1004 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
1005 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
1006 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
1008 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
1009 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
1011 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
1012 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
1015 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
1017 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
1018 s64 edge_dx_dy_alt_64; \
1020 dup_4x16b(y_mid_point, y_b); \
1022 edge_alt <<= edge_shift_alt; \
1023 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
1025 alternate_x.e[0] = edge_alt; \
1026 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
1027 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1029 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1030 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1034 #define setup_spans_y_select_up() \
1035 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1037 #define setup_spans_y_select_down() \
1038 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1040 #define setup_spans_y_select_alternate_yes(direction) \
1041 setup_spans_y_select_##direction() \
1043 #define setup_spans_y_select_alternate_no(direction) \
1045 #define setup_spans_alternate_select_left() \
1046 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1048 #define setup_spans_alternate_select_right() \
1049 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1051 #define setup_spans_alternate_select_none() \
1053 #define setup_spans_increment_alternate_yes() \
1054 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1055 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1056 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1057 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1058 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1060 #define setup_spans_increment_alternate_no() \
1062 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1064 span_uvrg_offset[0] = uvrg; \
1065 span_b_offset[0] = b; \
1066 setup_spans_adjust_interpolants_##direction(); \
1068 span_uvrg_offset[1] = uvrg; \
1069 span_b_offset[1] = b; \
1070 setup_spans_adjust_interpolants_##direction(); \
1072 span_uvrg_offset[2] = uvrg; \
1073 span_b_offset[2] = b; \
1074 setup_spans_adjust_interpolants_##direction(); \
1076 span_uvrg_offset[3] = uvrg; \
1077 span_b_offset[3] = b; \
1078 setup_spans_adjust_interpolants_##direction(); \
1080 span_uvrg_offset += 4; \
1081 span_b_offset += 4; \
1083 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1084 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1086 add_2x64b(left_x, left_x, left_dx_dy); \
1087 add_2x64b(right_x, right_x, right_dx_dy); \
1089 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1090 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1092 add_2x64b(left_x, left_x, left_dx_dy); \
1093 add_2x64b(right_x, right_x, right_dx_dy); \
1095 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1096 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1098 setup_spans_increment_alternate_##alternate_active(); \
1099 setup_spans_y_select_alternate_##alternate_active(direction); \
1100 setup_spans_alternate_select_##alternate(); \
1102 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1103 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1105 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1106 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1107 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1108 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1109 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1110 min_4x16b(left_right_x_16.high, left_right_x_16.high, c_max_blocks_per_row); \
1113 for(i = 0; i < 4; i++) \
1115 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1116 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1117 span_edge_data[i].right_mask = span_shift.e[i]; \
1118 span_edge_data[i].y = y_x4.e[i]; \
1119 setup_spans_debug_check(psx_gpu, &span_edge_data[i]); \
1122 span_edge_data += 4; \
1124 setup_spans_adjust_y_##direction(); \
1128 #define setup_spans_alternate_adjust_yes() \
1129 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1131 #define setup_spans_alternate_adjust_no() \
1134 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1135 setup_spans_alternate_adjust_##alternate_active(); \
1136 if(y_c > psx_gpu->viewport_end_y) \
1137 height -= y_c - psx_gpu->viewport_end_y - 1; \
1139 clip = psx_gpu->viewport_start_y - y_a; \
1144 setup_spans_clip(increment, alternate_active); \
1147 setup_spans_prologue_b(); \
1154 y_x4.e[1] = y_a + 1; \
1155 y_x4.e[2] = y_a + 2; \
1156 y_x4.e[3] = y_a + 3; \
1157 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1160 psx_gpu->num_spans = height; \
1163 setup_spans_set_x4(alternate, down, alternate_active); \
1165 } while(height > 0); \
1169 #define setup_spans_alternate_pre_increment_yes() \
1170 edge_alt += edge_dx_dy_alt \
1172 #define setup_spans_alternate_pre_increment_no() \
1174 #define setup_spans_up_decrement_height_yes() \
1177 #define setup_spans_up_decrement_height_no() \
1180 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1181 setup_spans_alternate_adjust_##alternate_active(); \
1184 if(y_c < psx_gpu->viewport_start_y) \
1185 height -= psx_gpu->viewport_start_y - y_c; \
1187 setup_spans_up_decrement_height_##alternate_active(); \
1189 clip = y_a - psx_gpu->viewport_end_y; \
1194 setup_spans_clip(decrement, alternate_active); \
1197 setup_spans_prologue_b(); \
1204 y_x4.e[1] = y_a - 1; \
1205 y_x4.e[2] = y_a - 2; \
1206 y_x4.e[3] = y_a - 3; \
1207 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1208 setup_spans_alternate_pre_increment_##alternate_active(); \
1209 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1211 setup_spans_adjust_interpolants_up(); \
1213 psx_gpu->num_spans = height; \
1216 setup_spans_set_x4(alternate, up, alternate_active); \
1221 #define index_left 0
1222 #define index_right 1
1224 #define setup_spans_up_up(minor, major) \
1225 setup_spans_prologue(yes); \
1226 s32 height_minor_a = y_a - y_b; \
1227 s32 height_minor_b = y_b - y_c; \
1228 s32 height = y_a - y_c; \
1230 dup_2x32b(x_starts, x_a); \
1231 x_ends.e[0] = x_c; \
1232 x_ends.e[1] = x_b; \
1234 compute_edge_delta_x3(x_b, height, height_minor_a); \
1235 setup_spans_up(index_##major, index_##minor, minor, yes) \
1240 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1241 vertex_struct *v_b, vertex_struct *v_c)
1243 setup_spans_up_up(left, right);
1246 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1247 vertex_struct *v_b, vertex_struct *v_c)
1249 setup_spans_up_up(right, left);
1252 #define setup_spans_down_down(minor, major) \
1253 setup_spans_prologue(yes); \
1254 s32 height_minor_a = y_b - y_a; \
1255 s32 height_minor_b = y_c - y_b; \
1256 s32 height = y_c - y_a; \
1258 dup_2x32b(x_starts, x_a); \
1259 x_ends.e[0] = x_c; \
1260 x_ends.e[1] = x_b; \
1262 compute_edge_delta_x3(x_b, height, height_minor_a); \
1263 setup_spans_down(index_##major, index_##minor, minor, yes) \
1265 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1266 vertex_struct *v_b, vertex_struct *v_c)
1268 setup_spans_down_down(left, right);
1271 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1272 vertex_struct *v_b, vertex_struct *v_c)
1274 setup_spans_down_down(right, left);
1277 #define setup_spans_up_flat() \
1278 s32 height = y_a - y_c; \
1281 compute_edge_delta_x2(); \
1282 setup_spans_up(index_left, index_right, none, no) \
1284 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1285 vertex_struct *v_b, vertex_struct *v_c)
1287 setup_spans_prologue(no);
1288 x_starts.e[0] = x_a;
1289 x_starts.e[1] = x_b;
1290 dup_2x32b(x_ends, x_c);
1292 setup_spans_up_flat();
1295 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1296 vertex_struct *v_b, vertex_struct *v_c)
1298 setup_spans_prologue(no);
1299 dup_2x32b(x_starts, x_a);
1303 setup_spans_up_flat();
1306 #define setup_spans_down_flat() \
1307 s32 height = y_c - y_a; \
1310 compute_edge_delta_x2(); \
1311 setup_spans_down(index_left, index_right, none, no) \
1313 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1314 vertex_struct *v_b, vertex_struct *v_c)
1316 setup_spans_prologue(no);
1317 x_starts.e[0] = x_a;
1318 x_starts.e[1] = x_b;
1319 dup_2x32b(x_ends, x_c);
1321 setup_spans_down_flat();
1324 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1325 vertex_struct *v_b, vertex_struct *v_c)
1327 setup_spans_prologue(no);
1328 dup_2x32b(x_starts, x_a);
1332 setup_spans_down_flat();
1335 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1336 vertex_struct *v_b, vertex_struct *v_c)
1338 setup_spans_prologue(no);
1346 s32 height_minor_a = y_a - y_b;
1347 s32 height_minor_b = y_c - y_a;
1348 s32 height_major = y_c - y_b;
1350 vec_2x64s edges_xy_b;
1351 vec_2x32s edges_dx_dy_b;
1352 vec_2x32u edge_shifts_b;
1354 vec_2x32s height_increment;
1356 x_starts.e[0] = x_a;
1357 x_starts.e[1] = x_c;
1358 dup_2x32b(x_ends, x_b);
1360 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1362 height_increment.e[0] = 0;
1363 height_increment.e[1] = height_minor_b;
1365 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1367 edges_xy_b.e[0] = edge_alt;
1368 edges_xy_b.e[1] = edges_xy.e[1];
1370 edge_shifts_b = edge_shifts;
1371 edge_shifts_b.e[0] = edge_shift_alt;
1373 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1374 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1378 if(y_b < psx_gpu->viewport_start_y)
1379 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1381 clip = y_a - psx_gpu->viewport_end_y;
1384 height_minor_a -= clip;
1386 setup_spans_clip(decrement, no);
1389 setup_spans_prologue_b();
1391 if (height_minor_a > 512)
1392 height_minor_a = 512;
1393 if (height_minor_a > 0)
1396 y_x4.e[1] = y_a - 1;
1397 y_x4.e[2] = y_a - 2;
1398 y_x4.e[3] = y_a - 3;
1399 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1400 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1401 setup_spans_adjust_interpolants_up();
1403 psx_gpu->num_spans = height_minor_a;
1404 while(height_minor_a > 0)
1406 setup_spans_set_x4(none, up, no);
1407 height_minor_a -= 4;
1410 span_edge_data += height_minor_a;
1411 span_uvrg_offset += height_minor_a;
1412 span_b_offset += height_minor_a;
1415 edges_xy = edges_xy_b;
1416 edges_dx_dy = edges_dx_dy_b;
1417 edge_shifts = edge_shifts_b;
1419 uvrg = psx_gpu->uvrg;
1424 if(y_c > psx_gpu->viewport_end_y)
1425 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1427 clip = psx_gpu->viewport_start_y - y_a;
1430 height_minor_b -= clip;
1432 setup_spans_clip(increment, no);
1435 if (height_minor_b > 512)
1436 height_minor_b = 512;
1437 if (height_minor_b > 0)
1440 y_x4.e[1] = y_a + 1;
1441 y_x4.e[2] = y_a + 2;
1442 y_x4.e[3] = y_a + 3;
1443 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1445 // FIXME: overflow corner case
1446 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1447 height_minor_b &= ~3;
1449 psx_gpu->num_spans += height_minor_b;
1450 while(height_minor_b > 0)
1452 setup_spans_set_x4(none, down, no);
1453 height_minor_b -= 4;
1457 left_split_triangles++;
1463 #define dither_table_entry_normal(value) \
1467 #define setup_blocks_load_msb_mask_indirect() \
1469 #define setup_blocks_load_msb_mask_direct() \
1470 vec_8x16u msb_mask; \
1471 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1474 #define setup_blocks_variables_shaded_textured(target) \
1475 vec_4x32u u_block; \
1476 vec_4x32u v_block; \
1477 vec_4x32u r_block; \
1478 vec_4x32u g_block; \
1479 vec_4x32u b_block; \
1480 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1481 vec_4x32u uvrg_dx4; \
1482 vec_4x32u uvrg_dx8; \
1484 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1485 u32 b_dx4 = b_dx << 2; \
1486 u32 b_dx8 = b_dx << 3; \
1489 vec_16x8u texture_mask; \
1490 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1491 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1492 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1493 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1495 #define setup_blocks_variables_shaded_untextured(target) \
1496 vec_4x32u r_block; \
1497 vec_4x32u g_block; \
1498 vec_4x32u b_block; \
1500 vec_4x32u rgb_dx4; \
1501 vec_4x32u rgb_dx8; \
1504 vec_8x8u d64_0x07; \
1509 dup_8x8b(d64_0x07, 0x07); \
1510 dup_8x8b(d64_1, 1); \
1511 dup_8x8b(d64_4, 4); \
1512 dup_8x8b(d64_128, 128); \
1514 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1515 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1516 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1517 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1519 #define setup_blocks_variables_unshaded_textured(target) \
1520 vec_4x32u u_block; \
1521 vec_4x32u v_block; \
1522 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1525 vec_2x32u uv = psx_gpu->uvrg.low; \
1527 vec_16x8u texture_mask; \
1528 shl_2x32b(uv_dx4, uv_dx, 2); \
1529 shl_2x32b(uv_dx8, uv_dx, 3); \
1530 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1531 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1534 #define setup_blocks_variables_unshaded_untextured_direct() \
1535 or_8x16b(colors, colors, msb_mask) \
1537 #define setup_blocks_variables_unshaded_untextured_indirect() \
1539 #define setup_blocks_variables_unshaded_untextured(target) \
1540 u32 color = psx_gpu->triangle_color; \
1543 u32 color_r = color & 0xFF; \
1544 u32 color_g = (color >> 8) & 0xFF; \
1545 u32 color_b = (color >> 16) & 0xFF; \
1547 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1548 ((color_b >> 3) << 10); \
1549 dup_8x16b(colors, color); \
1550 setup_blocks_variables_unshaded_untextured_##target() \
1552 #define setup_blocks_span_initialize_dithered_textured() \
1553 vec_8x16u dither_offsets; \
1554 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1556 #define setup_blocks_span_initialize_dithered_untextured() \
1557 vec_8x8u dither_offsets; \
1558 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1560 #define setup_blocks_span_initialize_dithered(texturing) \
1561 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1562 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1563 vec_8x8s dither_offsets_short; \
1566 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1567 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1568 setup_blocks_span_initialize_dithered_##texturing() \
1570 #define setup_blocks_span_initialize_undithered(texturing) \
1573 #define setup_blocks_span_initialize_shaded_textured() \
1575 vec_4x32u block_span; \
1576 u32 offset = span_edge_data->left_x; \
1578 uvrg = *span_uvrg_offset; \
1579 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1580 b = *span_b_offset; \
1581 b += b_dx * offset; \
1583 dup_4x32b(u_block, uvrg.e[0]); \
1584 dup_4x32b(v_block, uvrg.e[1]); \
1585 dup_4x32b(r_block, uvrg.e[2]); \
1586 dup_4x32b(g_block, uvrg.e[3]); \
1587 dup_4x32b(b_block, b); \
1589 block_span = psx_gpu->u_block_span; \
1590 add_4x32b(u_block, u_block, block_span); \
1591 block_span = psx_gpu->v_block_span; \
1592 add_4x32b(v_block, v_block, block_span); \
1593 block_span = psx_gpu->r_block_span; \
1594 add_4x32b(r_block, r_block, block_span); \
1595 block_span = psx_gpu->g_block_span; \
1596 add_4x32b(g_block, g_block, block_span); \
1597 block_span = psx_gpu->b_block_span; \
1598 add_4x32b(b_block, b_block, block_span); \
1601 #define setup_blocks_span_initialize_shaded_untextured() \
1603 vec_4x32u block_span; \
1604 u32 offset = span_edge_data->left_x; \
1606 rgb.low = span_uvrg_offset->high; \
1607 rgb.high.e[0] = *span_b_offset; \
1608 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1610 dup_4x32b(r_block, rgb.e[0]); \
1611 dup_4x32b(g_block, rgb.e[1]); \
1612 dup_4x32b(b_block, rgb.e[2]); \
1614 block_span = psx_gpu->r_block_span; \
1615 add_4x32b(r_block, r_block, block_span); \
1616 block_span = psx_gpu->g_block_span; \
1617 add_4x32b(g_block, g_block, block_span); \
1618 block_span = psx_gpu->b_block_span; \
1619 add_4x32b(b_block, b_block, block_span); \
1622 #define setup_blocks_span_initialize_unshaded_textured() \
1624 vec_4x32u block_span; \
1625 u32 offset = span_edge_data->left_x; \
1627 uv = span_uvrg_offset->low; \
1628 mla_scalar_2x32b(uv, uv_dx, offset); \
1630 dup_4x32b(u_block, uv.e[0]); \
1631 dup_4x32b(v_block, uv.e[1]); \
1633 block_span = psx_gpu->u_block_span; \
1634 add_4x32b(u_block, u_block, block_span); \
1635 block_span = psx_gpu->v_block_span; \
1636 add_4x32b(v_block, v_block, block_span); \
1639 #define setup_blocks_span_initialize_unshaded_untextured() \
1642 #define setup_blocks_texture_swizzled() \
1644 vec_8x8u u_saved = u; \
1645 sli_8x8b(u, v, 4); \
1646 sri_8x8b(v, u_saved, 4); \
1649 #define setup_blocks_texture_unswizzled() \
1651 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1654 vec_8x16u u_whole; \
1655 vec_8x16u v_whole; \
1656 vec_8x16u r_whole; \
1657 vec_8x16u g_whole; \
1658 vec_8x16u b_whole; \
1670 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1671 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1672 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1673 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1674 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1676 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1677 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1678 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1679 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1680 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1681 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1682 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1683 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1684 dup_4x32b(dx4, b_dx4); \
1685 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1687 mov_narrow_8x16b(u, u_whole); \
1688 mov_narrow_8x16b(v, v_whole); \
1689 mov_narrow_8x16b(r, r_whole); \
1690 mov_narrow_8x16b(g, g_whole); \
1691 mov_narrow_8x16b(b, b_whole); \
1693 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1694 add_4x32b(u_block, u_block, dx8); \
1695 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1696 add_4x32b(v_block, v_block, dx8); \
1697 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1698 add_4x32b(r_block, r_block, dx8); \
1699 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1700 add_4x32b(g_block, g_block, dx8); \
1701 dup_4x32b(dx8, b_dx8); \
1702 add_4x32b(b_block, b_block, dx8); \
1704 and_8x8b(u, u, texture_mask.low); \
1705 and_8x8b(v, v, texture_mask.high); \
1706 setup_blocks_texture_##swizzling(); \
1708 zip_8x16b(uv, u, v); \
1713 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1714 block->fb_ptr = fb_ptr; \
1717 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1720 vec_8x16u u_whole; \
1721 vec_8x16u v_whole; \
1730 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1731 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1733 dup_4x32b(dx4, uv_dx4.e[0]); \
1734 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1735 dup_4x32b(dx4, uv_dx4.e[1]); \
1736 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1738 mov_narrow_8x16b(u, u_whole); \
1739 mov_narrow_8x16b(v, v_whole); \
1741 dup_4x32b(dx8, uv_dx8.e[0]); \
1742 add_4x32b(u_block, u_block, dx8); \
1743 dup_4x32b(dx8, uv_dx8.e[1]); \
1744 add_4x32b(v_block, v_block, dx8); \
1746 and_8x8b(u, u, texture_mask.low); \
1747 and_8x8b(v, v, texture_mask.high); \
1748 setup_blocks_texture_##swizzling(); \
1750 zip_8x16b(uv, u, v); \
1752 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1753 block->fb_ptr = fb_ptr; \
1756 #define setup_blocks_store_shaded_untextured_dithered() \
1757 addq_8x8b(r, r, dither_offsets); \
1758 addq_8x8b(g, g, dither_offsets); \
1759 addq_8x8b(b, b, dither_offsets); \
1761 subq_8x8b(r, r, d64_4); \
1762 subq_8x8b(g, g, d64_4); \
1763 subq_8x8b(b, b, d64_4) \
1765 #define setup_blocks_store_shaded_untextured_undithered() \
1768 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1769 block->pixels = _pixels; \
1770 block->fb_ptr = fb_ptr \
1772 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1773 block->pixels = _pixels; \
1774 block->fb_ptr = fb_ptr \
1776 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1777 mul_long_8x8b(pixels, r, d64_1) \
1780 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1781 store_8x16b(_pixels, fb_ptr) \
1783 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1785 vec_8x16u fb_pixels; \
1786 vec_8x16u draw_mask; \
1787 vec_8x16u test_mask = psx_gpu->test_mask; \
1789 load_8x16b(fb_pixels, fb_ptr); \
1790 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1791 tst_8x16b(draw_mask, draw_mask, test_mask); \
1792 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1793 store_8x16b(fb_pixels, fb_ptr); \
1796 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1797 pixels = msb_mask; \
1798 mla_long_8x8b(pixels, r, d64_1) \
1801 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1804 vec_8x16u r_whole; \
1805 vec_8x16u g_whole; \
1806 vec_8x16u b_whole; \
1817 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1818 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1819 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1821 dup_4x32b(dx4, rgb_dx4.e[0]); \
1822 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1823 dup_4x32b(dx4, rgb_dx4.e[1]); \
1824 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1825 dup_4x32b(dx4, rgb_dx4.e[2]); \
1826 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1828 mov_narrow_8x16b(r, r_whole); \
1829 mov_narrow_8x16b(g, g_whole); \
1830 mov_narrow_8x16b(b, b_whole); \
1832 dup_4x32b(dx8, rgb_dx8.e[0]); \
1833 add_4x32b(r_block, r_block, dx8); \
1834 dup_4x32b(dx8, rgb_dx8.e[1]); \
1835 add_4x32b(g_block, g_block, dx8); \
1836 dup_4x32b(dx8, rgb_dx8.e[2]); \
1837 add_4x32b(b_block, b_block, dx8); \
1839 setup_blocks_store_shaded_untextured_##dithering(); \
1841 shr_8x8b(r, r, 3); \
1842 bic_8x8b(g, g, d64_0x07); \
1843 bic_8x8b(b, b, d64_0x07); \
1845 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1846 mla_long_8x8b(pixels, g, d64_4); \
1847 mla_long_8x8b(pixels, b, d64_128) \
1849 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1852 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1854 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1857 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1858 (_block)->draw_mask_bits = bits \
1860 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1862 vec_8x16u bits_mask; \
1863 vec_8x16u test_mask = psx_gpu->test_mask; \
1864 dup_8x16b(bits_mask, bits); \
1865 tst_8x16b(bits_mask, bits_mask, test_mask); \
1866 (_block)->draw_mask = bits_mask; \
1869 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1872 #define setup_blocks_add_blocks_indirect() \
1873 num_blocks += span_num_blocks; \
1875 if(num_blocks > MAX_BLOCKS) \
1877 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1878 flush_render_block_buffer(psx_gpu); \
1879 num_blocks = span_num_blocks; \
1880 block = psx_gpu->blocks; \
1883 #define setup_blocks_add_blocks_direct() \
1884 stats_add(texel_blocks_untextured, span_num_blocks); \
1885 span_pixel_blocks += span_num_blocks \
1888 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1889 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1890 psx_gpu_struct *psx_gpu) \
1892 setup_blocks_load_msb_mask_##target(); \
1893 setup_blocks_variables_##shading##_##texturing(target); \
1895 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1896 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1897 u32 *span_b_offset = psx_gpu->span_b_offset; \
1899 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1901 u32 num_spans = psx_gpu->num_spans; \
1906 u32 num_blocks = psx_gpu->num_blocks; \
1907 u32 span_num_blocks; \
1911 span_num_blocks = span_edge_data->num_blocks; \
1912 if(span_num_blocks) \
1914 y = span_edge_data->y; \
1915 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1917 setup_blocks_span_initialize_##shading##_##texturing(); \
1918 setup_blocks_span_initialize_##dithering(texturing); \
1920 setup_blocks_add_blocks_##target(); \
1922 s32 pixel_span = span_num_blocks * 8; \
1923 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1924 span_pixels += pixel_span; \
1926 span_num_blocks--; \
1927 while(span_num_blocks) \
1929 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1931 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1935 span_num_blocks--; \
1938 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1939 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1940 span_edge_data->right_mask); \
1946 zero_block_spans++; \
1951 span_uvrg_offset++; \
1955 psx_gpu->num_blocks = num_blocks; \
1959 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1963 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1964 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1966 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1967 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1969 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1970 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1971 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1972 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1974 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1975 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1977 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1979 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1980 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1983 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1985 block_struct *block = psx_gpu->blocks;
1986 u32 num_blocks = psx_gpu->num_blocks;
1987 stats_add(texel_blocks_4bpp, num_blocks);
1989 vec_8x8u texels_low;
1990 vec_8x8u texels_high;
1997 vec_16x8u clut_high;
1999 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2000 u16 *clut_ptr = psx_gpu->clut_ptr;
2002 // Can be done with one deinterleaving load on NEON
2003 load_8x16b(clut_a, clut_ptr);
2004 load_8x16b(clut_b, clut_ptr + 8);
2005 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
2007 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
2008 update_texture_4bpp_cache(psx_gpu);
2012 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
2013 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
2014 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
2015 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
2016 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
2017 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
2018 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
2019 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
2021 tbl_16(texels_low, texels, clut_low);
2022 tbl_16(texels_high, texels, clut_high);
2024 // Can be done with an interleaving store on NEON
2025 zip_8x16b(pixels, texels_low, texels_high);
2027 block->texels = pixels;
2034 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2036 block_struct *block = psx_gpu->blocks;
2037 u32 num_blocks = psx_gpu->num_blocks;
2039 stats_add(texel_blocks_8bpp, num_blocks);
2041 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2042 update_texture_8bpp_cache(psx_gpu);
2045 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2053 for(i = 0; i < 8; i++)
2055 offset = block->uv.e[i];
2057 texel = texture_ptr_8bpp[offset];
2058 texels.e[i] = psx_gpu->clut_ptr[texel];
2061 block->texels = texels;
2068 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2070 block_struct *block = psx_gpu->blocks;
2071 u32 num_blocks = psx_gpu->num_blocks;
2073 stats_add(texel_blocks_16bpp, num_blocks);
2077 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2083 for(i = 0; i < 8; i++)
2085 offset = block->uv.e[i];
2086 offset += ((offset & 0xFF00) * 3);
2088 texels.e[i] = texture_ptr_16bpp[offset];
2091 block->texels = texels;
2101 #define shade_blocks_load_msb_mask_indirect() \
2103 #define shade_blocks_load_msb_mask_direct() \
2104 vec_8x16u msb_mask; \
2105 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2107 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2108 block->draw_mask = _draw_mask; \
2109 block->pixels = _pixels \
2111 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2113 vec_8x16u fb_pixels; \
2114 or_8x16b(_pixels, _pixels, msb_mask); \
2115 load_8x16b(fb_pixels, block->fb_ptr); \
2116 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2117 store_8x16b(fb_pixels, block->fb_ptr); \
2121 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2122 if(psx_gpu->triangle_color == 0x808080) \
2124 false_modulated_blocks += num_blocks; \
2127 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2128 if(psx_gpu->triangle_color == 0x808080) \
2131 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2132 false_modulated_blocks += num_blocks; \
2137 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2140 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2143 u32 color = psx_gpu->triangle_color; \
2144 dup_8x8b(colors_r, color); \
2145 dup_8x8b(colors_g, color >> 8); \
2146 dup_8x8b(colors_b, color >> 16); \
2147 shade_blocks_textured_false_modulated_check_##dithering(target); \
2150 #define shade_blocks_textured_modulated_shaded_block_load() \
2151 colors_r = block->r; \
2152 colors_g = block->g; \
2153 colors_b = block->b \
2155 #define shade_blocks_textured_modulated_unshaded_block_load() \
2157 #define shade_blocks_textured_modulate_dithered(component) \
2158 pixels_##component = block->dither_offsets; \
2159 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2161 #define shade_blocks_textured_modulate_undithered(component) \
2162 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2164 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2165 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2166 psx_gpu_struct *psx_gpu) \
2168 block_struct *block = psx_gpu->blocks; \
2169 u32 num_blocks = psx_gpu->num_blocks; \
2172 vec_8x8u texels_r; \
2173 vec_8x8u texels_g; \
2174 vec_8x8u texels_b; \
2176 vec_8x8u colors_r; \
2177 vec_8x8u colors_g; \
2178 vec_8x8u colors_b; \
2180 vec_8x8u pixels_r_low; \
2181 vec_8x8u pixels_g_low; \
2182 vec_8x8u pixels_b_low; \
2185 vec_8x16u pixels_r; \
2186 vec_8x16u pixels_g; \
2187 vec_8x16u pixels_b; \
2189 vec_8x16u draw_mask; \
2190 vec_8x16u zero_mask; \
2192 vec_8x8u d64_0x07; \
2193 vec_8x8u d64_0x1F; \
2198 vec_8x16u d128_0x8000; \
2200 vec_8x16u test_mask = psx_gpu->test_mask; \
2201 u32 draw_mask_bits; \
2202 shade_blocks_load_msb_mask_##target(); \
2204 dup_8x8b(d64_0x07, 0x07); \
2205 dup_8x8b(d64_0x1F, 0x1F); \
2206 dup_8x8b(d64_1, 1); \
2207 dup_8x8b(d64_4, 4); \
2208 dup_8x8b(d64_128, 128); \
2210 dup_8x16b(d128_0x8000, 0x8000); \
2212 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2217 draw_mask_bits = block->draw_mask_bits; \
2218 dup_8x16b(draw_mask, draw_mask_bits); \
2219 tst_8x16b(draw_mask, draw_mask, test_mask); \
2221 shade_blocks_textured_modulated_##shading##_block_load(); \
2223 texels = block->texels; \
2225 mov_narrow_8x16b(texels_r, texels); \
2226 shr_narrow_8x16b(texels_g, texels, 5); \
2227 shr_narrow_8x16b(texels_b, texels, 7); \
2229 and_8x8b(texels_r, texels_r, d64_0x1F); \
2230 and_8x8b(texels_g, texels_g, d64_0x1F); \
2231 shr_8x8b(texels_b, texels_b, 3); \
2233 shade_blocks_textured_modulate_##dithering(r); \
2234 shade_blocks_textured_modulate_##dithering(g); \
2235 shade_blocks_textured_modulate_##dithering(b); \
2237 cmpeqz_8x16b(zero_mask, texels); \
2238 and_8x16b(pixels, texels, d128_0x8000); \
2240 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2241 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2242 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2244 or_8x16b(zero_mask, draw_mask, zero_mask); \
2246 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2247 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2248 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2250 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2251 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2252 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2254 shade_blocks_store_##target(zero_mask, pixels); \
2263 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2264 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2265 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2266 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2268 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2269 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2270 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2271 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2276 #define shade_blocks_textured_unmodulated_builder(target) \
2277 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2279 block_struct *block = psx_gpu->blocks; \
2280 u32 num_blocks = psx_gpu->num_blocks; \
2281 vec_8x16u draw_mask; \
2282 vec_8x16u test_mask = psx_gpu->test_mask; \
2283 u32 draw_mask_bits; \
2286 shade_blocks_load_msb_mask_##target(); \
2290 vec_8x16u zero_mask; \
2292 draw_mask_bits = block->draw_mask_bits; \
2293 dup_8x16b(draw_mask, draw_mask_bits); \
2294 tst_8x16b(draw_mask, draw_mask, test_mask); \
2296 pixels = block->texels; \
2298 cmpeqz_8x16b(zero_mask, pixels); \
2299 or_8x16b(zero_mask, draw_mask, zero_mask); \
2301 shade_blocks_store_##target(zero_mask, pixels); \
2308 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2309 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2312 block_struct *block = psx_gpu->blocks; \
2313 u32 num_blocks = psx_gpu->num_blocks; \
2314 vec_8x16u draw_mask; \
2315 vec_8x16u test_mask = psx_gpu->test_mask; \
2316 u32 draw_mask_bits; \
2319 shade_blocks_load_msb_mask_##target(); \
2323 vec_8x16u zero_mask; \
2325 draw_mask_bits = block->draw_mask_bits; \
2326 dup_8x16b(draw_mask, draw_mask_bits); \
2327 tst_8x16b(draw_mask, draw_mask, test_mask); \
2329 pixels = block->texels; \
2331 cmpeqz_8x16b(zero_mask, pixels); \
2332 or_8x16b(zero_mask, draw_mask, zero_mask); \
2334 shade_blocks_store_##target(zero_mask, pixels); \
2343 shade_blocks_textured_unmodulated_builder(indirect)
2344 shade_blocks_textured_unmodulated_builder(direct)
2346 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2350 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2352 block_struct *block = psx_gpu->blocks;
2353 u32 num_blocks = psx_gpu->num_blocks;
2355 vec_8x16u pixels = block->pixels;
2356 shade_blocks_load_msb_mask_direct();
2360 shade_blocks_store_direct(block->draw_mask, pixels);
2369 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2374 #define blend_blocks_mask_evaluate_on() \
2375 vec_8x16u mask_pixels; \
2376 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2377 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2379 #define blend_blocks_mask_evaluate_off() \
2381 #define blend_blocks_average() \
2383 vec_8x16u pixels_no_msb; \
2384 vec_8x16u fb_pixels_no_msb; \
2386 vec_8x16u d128_0x0421; \
2387 vec_8x16u d128_0x8000; \
2389 dup_8x16b(d128_0x0421, 0x0421); \
2390 dup_8x16b(d128_0x8000, 0x8000); \
2392 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2393 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2394 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2395 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2396 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2397 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2400 #define blend_blocks_add() \
2402 vec_8x16u pixels_rb, pixels_g; \
2403 vec_8x16u fb_rb, fb_g; \
2405 vec_8x16u d128_0x7C1F; \
2406 vec_8x16u d128_0x03E0; \
2408 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2409 dup_8x16b(d128_0x03E0, 0x03E0); \
2411 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2412 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2414 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2415 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2417 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2418 add_8x16b(fb_g, fb_g, pixels_g); \
2420 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2421 vector_cast(vec_16x8u, d128_0x7C1F)); \
2422 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2424 or_8x16b(blend_pixels, fb_rb, fb_g); \
2427 #define blend_blocks_subtract() \
2429 vec_8x16u pixels_rb, pixels_g; \
2430 vec_8x16u fb_rb, fb_g; \
2432 vec_8x16u d128_0x7C1F; \
2433 vec_8x16u d128_0x03E0; \
2435 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2436 dup_8x16b(d128_0x03E0, 0x03E0); \
2438 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2439 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2441 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2442 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2444 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2445 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2446 subs_8x16b(fb_g, fb_g, pixels_g); \
2448 or_8x16b(blend_pixels, fb_rb, fb_g); \
2451 #define blend_blocks_add_fourth() \
2453 vec_8x16u pixels_rb, pixels_g; \
2454 vec_8x16u pixels_fourth; \
2455 vec_8x16u fb_rb, fb_g; \
2457 vec_8x16u d128_0x7C1F; \
2458 vec_8x16u d128_0x1C07; \
2459 vec_8x16u d128_0x03E0; \
2460 vec_8x16u d128_0x00E0; \
2462 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2463 dup_8x16b(d128_0x1C07, 0x1C07); \
2464 dup_8x16b(d128_0x03E0, 0x03E0); \
2465 dup_8x16b(d128_0x00E0, 0x00E0); \
2467 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2469 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2470 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2472 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2473 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2475 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2476 add_8x16b(fb_g, fb_g, pixels_g); \
2478 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2479 vector_cast(vec_16x8u, d128_0x7C1F)); \
2480 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2482 or_8x16b(blend_pixels, fb_rb, fb_g); \
2485 #define blend_blocks_blended_combine_textured() \
2487 vec_8x16u blend_mask; \
2488 cmpltz_8x16b(blend_mask, pixels); \
2490 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2491 bif_8x16b(blend_pixels, pixels, blend_mask); \
2494 #define blend_blocks_blended_combine_untextured() \
2497 #define blend_blocks_body_blend(blend_mode, texturing) \
2499 blend_blocks_##blend_mode(); \
2500 blend_blocks_blended_combine_##texturing(); \
2503 #define blend_blocks_body_average(texturing) \
2504 blend_blocks_body_blend(average, texturing) \
2506 #define blend_blocks_body_add(texturing) \
2507 blend_blocks_body_blend(add, texturing) \
2509 #define blend_blocks_body_subtract(texturing) \
2510 blend_blocks_body_blend(subtract, texturing) \
2512 #define blend_blocks_body_add_fourth(texturing) \
2513 blend_blocks_body_blend(add_fourth, texturing) \
2515 #define blend_blocks_body_unblended(texturing) \
2516 blend_pixels = pixels \
2519 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2521 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2524 block_struct *block = psx_gpu->blocks; \
2525 u32 num_blocks = psx_gpu->num_blocks; \
2526 vec_8x16u draw_mask; \
2528 vec_8x16u blend_pixels; \
2529 vec_8x16u framebuffer_pixels; \
2530 vec_8x16u msb_mask; \
2534 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2538 pixels = block->pixels; \
2539 draw_mask = block->draw_mask; \
2540 fb_ptr = block->fb_ptr; \
2542 load_8x16b(framebuffer_pixels, fb_ptr); \
2544 blend_blocks_mask_evaluate_##mask_evaluate(); \
2545 blend_blocks_body_##blend_mode(texturing); \
2547 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2548 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2549 store_8x16b(framebuffer_pixels, fb_ptr); \
2559 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2563 blend_blocks_builder(textured, average, off);
2564 blend_blocks_builder(textured, average, on);
2565 blend_blocks_builder(textured, add, off);
2566 blend_blocks_builder(textured, add, on);
2567 blend_blocks_builder(textured, subtract, off);
2568 blend_blocks_builder(textured, subtract, on);
2569 blend_blocks_builder(textured, add_fourth, off);
2570 blend_blocks_builder(textured, add_fourth, on);
2572 blend_blocks_builder(untextured, average, off);
2573 blend_blocks_builder(untextured, average, on);
2574 blend_blocks_builder(untextured, add, off);
2575 blend_blocks_builder(untextured, add, on);
2576 blend_blocks_builder(untextured, subtract, off);
2577 blend_blocks_builder(untextured, subtract, on);
2578 blend_blocks_builder(untextured, add_fourth, off);
2579 blend_blocks_builder(untextured, add_fourth, on);
2581 blend_blocks_builder(textured, unblended, on);
2586 #define vertex_swap(_a, _b) \
2588 vertex_struct *temp_vertex = _a; \
2591 triangle_winding ^= 1; \
2595 // Setup blocks parametric-variables:
2596 // SHADE TEXTURE_MAP SWIZZLING
2603 // 8 inputs, 6 combinations
2605 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2606 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2608 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2609 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2611 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2613 setup_blocks_switch_untextured_##shading(dithering, target) \
2615 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2616 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2618 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2619 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2621 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2622 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2624 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2625 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2627 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2628 dithering, mask_evaluate) \
2629 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2631 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2633 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2635 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2637 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2639 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2640 dithering, mask_evaluate) \
2641 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2642 texture_mode, dithering) \
2644 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2645 blending, mask_evaluate) \
2646 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2647 dithering, mask_evaluate) \
2652 #define texture_blocks_switch_untextured(texture_mode) \
2653 texture_blocks_untextured \
2655 #define texture_blocks_switch_textured(texture_mode) \
2656 texture_blocks_##texture_mode \
2658 #define texture_blocks_switch(texturing, texture_mode) \
2659 texture_blocks_switch_##texturing(texture_mode) \
2662 // Shade blocks parametric-variables:
2663 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2672 // 16 inputs, 8 combinations
2674 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2675 shade_blocks_unshaded_untextured_##target \
2677 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2678 shade_blocks_textured_unmodulated_##target \
2680 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2681 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2683 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2684 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2686 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2687 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2689 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2690 shade_blocks_shaded_untextured \
2692 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2693 shade_blocks_textured_unmodulated_##target \
2695 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2696 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2698 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2699 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2701 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2702 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2704 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2706 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2708 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2710 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2712 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2714 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2716 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2717 dithering, mask_evaluate) \
2718 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2721 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2722 blending, mask_evaluate) \
2723 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2727 // Blend blocks parametric-variables:
2728 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2747 // 32 inputs, 18 combinations
2749 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2750 blend_blocks_textured_unblended_##mask_evaluate \
2752 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2753 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2755 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2756 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2759 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2760 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2762 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2764 texture_blocks_switch(texturing, texture_mode), \
2765 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2767 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2770 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2771 mask_evaluate, shading, dithering, texturing, blending) \
2772 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2773 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2774 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2775 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2777 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2778 mask_evaluate, shading, dithering, texturing) \
2779 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2780 mask_evaluate, shading, dithering, texturing, unblended), \
2781 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2782 mask_evaluate, shading, dithering, texturing, blended) \
2784 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2785 mask_evaluate, shading, dithering) \
2786 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2787 mask_evaluate, shading, dithering, untextured), \
2788 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2789 mask_evaluate, shading, dithering, textured) \
2791 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2792 mask_evaluate, shading) \
2793 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2794 mask_evaluate, shading, undithered), \
2795 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2796 mask_evaluate, shading, dithered) \
2798 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2800 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2802 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2805 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2806 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2807 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2809 #define render_blocks_switch_block_texture_mode(texture_mode) \
2810 render_blocks_switch_block_blend_mode(texture_mode, average), \
2811 render_blocks_switch_block_blend_mode(texture_mode, add), \
2812 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2813 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2815 #define render_blocks_switch_block() \
2816 render_blocks_switch_block_texture_mode(4bpp), \
2817 render_blocks_switch_block_texture_mode(8bpp), \
2818 render_blocks_switch_block_texture_mode(16bpp), \
2819 render_blocks_switch_block_texture_mode(16bpp) \
2822 render_block_handler_struct render_triangle_block_handlers[] =
2824 render_blocks_switch_block()
2827 #undef render_blocks_switch_block_modulation
2829 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2830 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2832 "texture mode: " #texture_mode "\n" \
2833 "blend mode: " #blend_mode "\n" \
2834 "mask evaluation: " #mask_evaluate "\n" \
2841 char *render_block_flag_strings[] =
2843 render_blocks_switch_block()
2847 #define triangle_y_direction_up 1
2848 #define triangle_y_direction_flat 2
2849 #define triangle_y_direction_down 0
2851 #define triangle_winding_positive 0
2852 #define triangle_winding_negative 1
2854 #define triangle_set_direction(direction_variable, value) \
2855 u32 direction_variable = (u32)(value) >> 31; \
2857 direction_variable = 2 \
2859 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2860 case (triangle_y_direction_##direction_a | \
2861 (triangle_y_direction_##direction_b << 2) | \
2862 (triangle_y_direction_##direction_c << 4) | \
2863 (triangle_winding_##winding << 6)) \
2865 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2866 vertex_struct *vertexes_out[3])
2868 s32 y_top, y_bottom;
2870 u32 triangle_winding = 0;
2872 vertex_struct *a = &(vertexes[0]);
2873 vertex_struct *b = &(vertexes[1]);
2874 vertex_struct *c = &(vertexes[2]);
2876 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2882 if(triangle_area == 0)
2904 if((y_bottom - y_top) >= 512)
2912 if(triangle_area < 0)
2914 triangle_area = -triangle_area;
2915 triangle_winding ^= 1;
2930 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2938 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2947 psx_gpu->triangle_area = triangle_area;
2948 psx_gpu->triangle_winding = triangle_winding;
2950 vertexes_out[0] = a;
2951 vertexes_out[1] = b;
2952 vertexes_out[2] = c;
2957 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2958 vertex_struct *vertex_ptrs[3], u32 flags)
2960 psx_gpu->num_spans = 0;
2962 vertex_struct *a = vertex_ptrs[0];
2963 vertex_struct *b = vertex_ptrs[1];
2964 vertex_struct *c = vertex_ptrs[2];
2966 s32 y_delta_a = b->y - a->y;
2967 s32 y_delta_b = c->y - b->y;
2968 s32 y_delta_c = c->y - a->y;
2970 triangle_set_direction(y_direction_a, y_delta_a);
2971 triangle_set_direction(y_direction_b, y_delta_b);
2972 triangle_set_direction(y_direction_c, y_delta_c);
2974 compute_all_gradients(psx_gpu, a, b, c);
2976 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2977 (psx_gpu->triangle_winding << 6))
2979 triangle_case(up, up, up, negative):
2980 triangle_case(up, up, flat, negative):
2981 triangle_case(up, up, down, negative):
2982 setup_spans_up_right(psx_gpu, a, b, c);
2985 triangle_case(flat, up, up, negative):
2986 triangle_case(flat, up, flat, negative):
2987 triangle_case(flat, up, down, negative):
2988 setup_spans_up_a(psx_gpu, a, b, c);
2991 triangle_case(down, up, up, negative):
2992 setup_spans_up_down(psx_gpu, a, c, b);
2995 triangle_case(down, up, flat, negative):
2996 setup_spans_down_a(psx_gpu, a, c, b);
2999 triangle_case(down, up, down, negative):
3000 setup_spans_down_right(psx_gpu, a, c, b);
3003 triangle_case(down, flat, up, negative):
3004 triangle_case(down, flat, flat, negative):
3005 triangle_case(down, flat, down, negative):
3006 setup_spans_down_b(psx_gpu, a, b, c);
3009 triangle_case(down, down, up, negative):
3010 triangle_case(down, down, flat, negative):
3011 triangle_case(down, down, down, negative):
3012 setup_spans_down_left(psx_gpu, a, b, c);
3015 triangle_case(up, up, up, positive):
3016 triangle_case(up, up, flat, positive):
3017 triangle_case(up, up, down, positive):
3018 setup_spans_up_left(psx_gpu, a, b, c);
3021 triangle_case(up, flat, up, positive):
3022 triangle_case(up, flat, flat, positive):
3023 triangle_case(up, flat, down, positive):
3024 setup_spans_up_b(psx_gpu, a, b, c);
3027 triangle_case(up, down, up, positive):
3028 setup_spans_up_right(psx_gpu, a, c, b);
3031 triangle_case(up, down, flat, positive):
3032 setup_spans_up_a(psx_gpu, a, c, b);
3035 triangle_case(up, down, down, positive):
3036 setup_spans_up_down(psx_gpu, a, b, c);
3039 triangle_case(flat, down, up, positive):
3040 triangle_case(flat, down, flat, positive):
3041 triangle_case(flat, down, down, positive):
3042 setup_spans_down_a(psx_gpu, a, b, c);
3045 triangle_case(down, down, up, positive):
3046 triangle_case(down, down, flat, positive):
3047 triangle_case(down, down, down, positive):
3048 setup_spans_down_right(psx_gpu, a, b, c);
3053 spans += psx_gpu->num_spans;
3056 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3060 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3062 for(i = 0; i < psx_gpu->num_spans; i++)
3064 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3065 psx_gpu->span_edge_data[i].num_blocks = 0;
3070 for(i = 0; i < psx_gpu->num_spans; i++)
3072 if(psx_gpu->span_edge_data[i].y & 1)
3073 psx_gpu->span_edge_data[i].num_blocks = 0;
3077 assert(psx_gpu->span_edge_data[0].y < 1024u);
3079 u32 render_state = flags &
3080 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3081 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3082 render_state |= psx_gpu->render_state_base;
3084 if((psx_gpu->render_state != render_state) ||
3085 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3087 psx_gpu->render_state = render_state;
3088 flush_render_block_buffer(psx_gpu);
3094 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3096 psx_gpu->render_block_handler =
3097 &(render_triangle_block_handlers[render_state]);
3098 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3102 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3105 vertex_struct *vertex_ptrs[3];
3106 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3107 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3110 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3112 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3114 block_struct *block = psx_gpu->blocks;
3115 u32 num_blocks = psx_gpu->num_blocks;
3118 vec_8x8u texel_indexes;
3120 u16 *clut_ptr = psx_gpu->clut_ptr;
3125 texel_indexes = block->r;
3127 for(i = 0; i < 8; i++)
3129 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3132 block->texels = texels;
3142 #define setup_sprite_tiled_initialize_4bpp_clut() \
3143 u16 *clut_ptr = psx_gpu->clut_ptr; \
3144 vec_8x16u clut_a, clut_b; \
3145 vec_16x8u clut_low, clut_high; \
3147 load_8x16b(clut_a, clut_ptr); \
3148 load_8x16b(clut_b, clut_ptr + 8); \
3149 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3151 #define setup_sprite_tiled_initialize_4bpp() \
3152 setup_sprite_tiled_initialize_4bpp_clut(); \
3154 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3155 update_texture_4bpp_cache(psx_gpu) \
3157 #define setup_sprite_tiled_initialize_8bpp() \
3158 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3159 update_texture_8bpp_cache(psx_gpu) \
3162 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3163 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3164 ((texture_offset + offset) & texture_mask); \
3166 load_64b(texels, texture_block_ptr) \
3169 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3170 num_blocks += tile_num_blocks; \
3171 sprite_blocks += tile_num_blocks; \
3173 if(num_blocks > MAX_BLOCKS) \
3175 flush_render_block_buffer(psx_gpu); \
3176 num_blocks = tile_num_blocks; \
3177 block = psx_gpu->blocks; \
3180 #define setup_sprite_tile_full_4bpp(edge) \
3182 vec_8x8u texels_low, texels_high; \
3184 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3186 while(sub_tile_height) \
3188 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3189 tbl_16(texels_low, texels, clut_low); \
3190 tbl_16(texels_high, texels, clut_high); \
3191 zip_8x16b(pixels, texels_low, texels_high); \
3193 block->texels = pixels; \
3194 block->draw_mask_bits = left_mask_bits; \
3195 block->fb_ptr = fb_ptr; \
3198 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3199 tbl_16(texels_low, texels, clut_low); \
3200 tbl_16(texels_high, texels, clut_high); \
3201 zip_8x16b(pixels, texels_low, texels_high); \
3203 block->texels = pixels; \
3204 block->draw_mask_bits = right_mask_bits; \
3205 block->fb_ptr = fb_ptr + 8; \
3209 texture_offset += 0x10; \
3210 sub_tile_height--; \
3212 texture_offset += 0xF00; \
3213 psx_gpu->num_blocks = num_blocks; \
3216 #define setup_sprite_tile_half_4bpp(edge) \
3218 vec_8x8u texels_low, texels_high; \
3220 setup_sprite_tile_add_blocks(sub_tile_height); \
3222 while(sub_tile_height) \
3224 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3225 tbl_16(texels_low, texels, clut_low); \
3226 tbl_16(texels_high, texels, clut_high); \
3227 zip_8x16b(pixels, texels_low, texels_high); \
3229 block->texels = pixels; \
3230 block->draw_mask_bits = edge##_mask_bits; \
3231 block->fb_ptr = fb_ptr; \
3235 texture_offset += 0x10; \
3236 sub_tile_height--; \
3238 texture_offset += 0xF00; \
3239 psx_gpu->num_blocks = num_blocks; \
3243 #define setup_sprite_tile_full_8bpp(edge) \
3245 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3247 while(sub_tile_height) \
3249 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3250 block->r = texels; \
3251 block->draw_mask_bits = left_mask_bits; \
3252 block->fb_ptr = fb_ptr; \
3255 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3256 block->r = texels; \
3257 block->draw_mask_bits = right_mask_bits; \
3258 block->fb_ptr = fb_ptr + 8; \
3262 texture_offset += 0x10; \
3263 sub_tile_height--; \
3265 texture_offset += 0xF00; \
3266 psx_gpu->num_blocks = num_blocks; \
3269 #define setup_sprite_tile_half_8bpp(edge) \
3271 setup_sprite_tile_add_blocks(sub_tile_height); \
3273 while(sub_tile_height) \
3275 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3276 block->r = texels; \
3277 block->draw_mask_bits = edge##_mask_bits; \
3278 block->fb_ptr = fb_ptr; \
3282 texture_offset += 0x10; \
3283 sub_tile_height--; \
3285 texture_offset += 0xF00; \
3286 psx_gpu->num_blocks = num_blocks; \
3290 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3291 texture_offset = texture_offset_base + 8; \
3294 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3295 texture_offset = texture_offset_base \
3297 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3298 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3300 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3301 texture_offset = texture_offset_base \
3303 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3306 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3308 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3309 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3311 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3314 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3318 sub_tile_height = column_data; \
3319 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3320 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3321 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3324 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3328 u32 tiles_remaining = column_data >> 16; \
3329 sub_tile_height = column_data & 0xFF; \
3330 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3331 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3332 tiles_remaining -= 1; \
3334 while(tiles_remaining) \
3336 sub_tile_height = 16; \
3337 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3338 tiles_remaining--; \
3341 sub_tile_height = (column_data >> 8) & 0xFF; \
3342 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3343 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3347 #define setup_sprite_column_data_single() \
3348 column_data = height \
3350 #define setup_sprite_column_data_multi() \
3351 column_data = 16 - offset_v; \
3352 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3353 column_data |= (tile_height - 1) << 16 \
3356 #define RIGHT_MASK_BIT_SHIFT 8
3357 #define RIGHT_MASK_BIT_SHIFT_4x 16
3359 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3360 edge_mode, edge, x4mode) \
3362 setup_sprite_column_data_##multi_height(); \
3363 left_mask_bits = left_block_mask | right_block_mask; \
3364 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3366 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3367 texture_mode, x4mode); \
3370 #define setup_sprite_tiled_advance_column() \
3371 texture_offset_base += 0x100; \
3372 if((texture_offset_base & 0xF00) == 0) \
3373 texture_offset_base -= (0x100 + 0xF00) \
3375 #define FB_PTR_MULTIPLIER 1
3376 #define FB_PTR_MULTIPLIER_4x 2
3378 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3379 left_mode, right_mode, x4mode) \
3381 setup_sprite_column_data_##multi_height(); \
3382 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3383 * FB_PTR_MULTIPLIER##x4mode; \
3386 left_mask_bits = left_block_mask; \
3387 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3389 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3390 texture_mode, x4mode); \
3391 fb_ptr += fb_ptr_advance_column; \
3393 left_mask_bits = 0x00; \
3394 right_mask_bits = 0x00; \
3398 setup_sprite_tiled_advance_column(); \
3399 setup_sprite_tile_column_height_##multi_height(full, none, \
3400 texture_mode, x4mode); \
3401 fb_ptr += fb_ptr_advance_column; \
3405 left_mask_bits = right_block_mask; \
3406 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3408 setup_sprite_tiled_advance_column(); \
3409 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3410 texture_mode, x4mode); \
3415 #define setup_sprite_tiled_initialize_4bpp_4x() \
3416 setup_sprite_tiled_initialize_4bpp_clut() \
3418 #define setup_sprite_tiled_initialize_8bpp_4x() \
3421 #define setup_sprite_tile_full_4bpp_4x(edge) \
3423 vec_8x8u texels_low, texels_high; \
3424 vec_8x16u pixels, pixels_wide; \
3425 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3426 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3427 u32 left_mask_bits_b = left_mask_bits >> 8; \
3428 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3429 u32 right_mask_bits_b = right_mask_bits >> 8; \
3431 while(sub_tile_height) \
3433 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3434 tbl_16(texels_low, texels, clut_low); \
3435 tbl_16(texels_high, texels, clut_high); \
3436 zip_8x16b(pixels, texels_low, texels_high); \
3438 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3439 block->texels = pixels_wide; \
3440 block->draw_mask_bits = left_mask_bits_a; \
3441 block->fb_ptr = fb_ptr; \
3444 block->texels = pixels_wide; \
3445 block->draw_mask_bits = left_mask_bits_a; \
3446 block->fb_ptr = fb_ptr + 1024; \
3449 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3450 block->texels = pixels_wide; \
3451 block->draw_mask_bits = left_mask_bits_b; \
3452 block->fb_ptr = fb_ptr + 8; \
3455 block->texels = pixels_wide; \
3456 block->draw_mask_bits = left_mask_bits_b; \
3457 block->fb_ptr = fb_ptr + 1024 + 8; \
3460 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3461 tbl_16(texels_low, texels, clut_low); \
3462 tbl_16(texels_high, texels, clut_high); \
3463 zip_8x16b(pixels, texels_low, texels_high); \
3465 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3466 block->texels = pixels_wide; \
3467 block->draw_mask_bits = right_mask_bits_a; \
3468 block->fb_ptr = fb_ptr + 16; \
3471 block->texels = pixels_wide; \
3472 block->draw_mask_bits = right_mask_bits_a; \
3473 block->fb_ptr = fb_ptr + 1024 + 16; \
3476 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3477 block->texels = pixels_wide; \
3478 block->draw_mask_bits = right_mask_bits_b; \
3479 block->fb_ptr = fb_ptr + 24; \
3482 block->texels = pixels_wide; \
3483 block->draw_mask_bits = right_mask_bits_b; \
3484 block->fb_ptr = fb_ptr + 1024 + 24; \
3488 texture_offset += 0x10; \
3489 sub_tile_height--; \
3491 texture_offset += 0xF00; \
3492 psx_gpu->num_blocks = num_blocks; \
3495 #define setup_sprite_tile_half_4bpp_4x(edge) \
3497 vec_8x8u texels_low, texels_high; \
3498 vec_8x16u pixels, pixels_wide; \
3499 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3500 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3501 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3503 while(sub_tile_height) \
3505 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3506 tbl_16(texels_low, texels, clut_low); \
3507 tbl_16(texels_high, texels, clut_high); \
3508 zip_8x16b(pixels, texels_low, texels_high); \
3510 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3511 block->texels = pixels_wide; \
3512 block->draw_mask_bits = edge##_mask_bits_a; \
3513 block->fb_ptr = fb_ptr; \
3516 block->texels = pixels_wide; \
3517 block->draw_mask_bits = edge##_mask_bits_a; \
3518 block->fb_ptr = fb_ptr + 1024; \
3521 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3522 block->texels = pixels_wide; \
3523 block->draw_mask_bits = edge##_mask_bits_b; \
3524 block->fb_ptr = fb_ptr + 8; \
3527 block->texels = pixels_wide; \
3528 block->draw_mask_bits = edge##_mask_bits_b; \
3529 block->fb_ptr = fb_ptr + 1024 + 8; \
3533 texture_offset += 0x10; \
3534 sub_tile_height--; \
3536 texture_offset += 0xF00; \
3537 psx_gpu->num_blocks = num_blocks; \
3541 #define setup_sprite_tile_full_8bpp_4x(edge) \
3543 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3544 vec_16x8u texels_wide; \
3545 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3546 u32 left_mask_bits_b = left_mask_bits >> 8; \
3547 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3548 u32 right_mask_bits_b = right_mask_bits >> 8; \
3550 while(sub_tile_height) \
3552 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3553 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3554 block->r = texels_wide.low; \
3555 block->draw_mask_bits = left_mask_bits_a; \
3556 block->fb_ptr = fb_ptr; \
3559 block->r = texels_wide.low; \
3560 block->draw_mask_bits = left_mask_bits_a; \
3561 block->fb_ptr = fb_ptr + 1024; \
3564 block->r = texels_wide.high; \
3565 block->draw_mask_bits = left_mask_bits_b; \
3566 block->fb_ptr = fb_ptr + 8; \
3569 block->r = texels_wide.high; \
3570 block->draw_mask_bits = left_mask_bits_b; \
3571 block->fb_ptr = fb_ptr + 1024 + 8; \
3574 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3575 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3576 block->r = texels_wide.low; \
3577 block->draw_mask_bits = right_mask_bits_a; \
3578 block->fb_ptr = fb_ptr + 16; \
3581 block->r = texels_wide.low; \
3582 block->draw_mask_bits = right_mask_bits_a; \
3583 block->fb_ptr = fb_ptr + 1024 + 16; \
3586 block->r = texels_wide.high; \
3587 block->draw_mask_bits = right_mask_bits_b; \
3588 block->fb_ptr = fb_ptr + 24; \
3591 block->r = texels_wide.high; \
3592 block->draw_mask_bits = right_mask_bits_b; \
3593 block->fb_ptr = fb_ptr + 24 + 1024; \
3597 texture_offset += 0x10; \
3598 sub_tile_height--; \
3600 texture_offset += 0xF00; \
3601 psx_gpu->num_blocks = num_blocks; \
3604 #define setup_sprite_tile_half_8bpp_4x(edge) \
3606 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3607 vec_16x8u texels_wide; \
3608 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3609 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3611 while(sub_tile_height) \
3613 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3614 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3615 block->r = texels_wide.low; \
3616 block->draw_mask_bits = edge##_mask_bits_a; \
3617 block->fb_ptr = fb_ptr; \
3620 block->r = texels_wide.low; \
3621 block->draw_mask_bits = edge##_mask_bits_a; \
3622 block->fb_ptr = fb_ptr + 1024; \
3625 block->r = texels_wide.high; \
3626 block->draw_mask_bits = edge##_mask_bits_b; \
3627 block->fb_ptr = fb_ptr + 8; \
3630 block->r = texels_wide.high; \
3631 block->draw_mask_bits = edge##_mask_bits_b; \
3632 block->fb_ptr = fb_ptr + 8 + 1024; \
3636 texture_offset += 0x10; \
3637 sub_tile_height--; \
3639 texture_offset += 0xF00; \
3640 psx_gpu->num_blocks = num_blocks; \
3644 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3645 texture_offset = texture_offset_base + 8; \
3648 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3649 texture_offset = texture_offset_base \
3651 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3652 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3654 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3655 texture_offset = texture_offset_base \
3657 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3660 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3662 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3663 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3665 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3668 #define setup_sprite_offset_u_adjust() \
3670 #define setup_sprite_comapre_left_block_mask() \
3671 ((left_block_mask & 0xFF) == 0xFF) \
3673 #define setup_sprite_comapre_right_block_mask() \
3674 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3677 #define setup_sprite_offset_u_adjust_4x() \
3679 offset_u_right = offset_u_right * 2 + 1 \
3681 #define setup_sprite_comapre_left_block_mask_4x() \
3682 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3684 #define setup_sprite_comapre_right_block_mask_4x() \
3685 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3688 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3689 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3690 s32 u, s32 v, s32 width, s32 height, u32 color) \
3692 s32 offset_u = u & 0xF; \
3693 s32 offset_v = v & 0xF; \
3695 s32 width_rounded = offset_u + width + 15; \
3696 s32 height_rounded = offset_v + height + 15; \
3697 s32 tile_height = height_rounded / 16; \
3698 s32 tile_width = width_rounded / 16; \
3699 u32 offset_u_right = width_rounded & 0xF; \
3701 setup_sprite_offset_u_adjust##x4mode(); \
3703 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3704 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3706 u32 left_mask_bits; \
3707 u32 right_mask_bits; \
3709 u32 sub_tile_height; \
3712 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3713 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3714 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3715 ((psx_gpu->texture_mask_height >> 4) << 12); \
3716 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3717 ((v & 0xF0) << 8); \
3718 u32 texture_offset_base = texture_offset; \
3721 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3722 u32 num_blocks = psx_gpu->num_blocks; \
3723 block_struct *block = psx_gpu->blocks + num_blocks; \
3725 u8 *texture_block_ptr; \
3728 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3730 control_mask = tile_width == 1; \
3731 control_mask |= (tile_height == 1) << 1; \
3732 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3733 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3735 sprites_##texture_mode++; \
3737 switch(control_mask) \
3741 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3746 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3751 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3756 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3761 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3766 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3771 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3776 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3781 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3786 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3791 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3796 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3801 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3806 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3813 setup_sprite_tiled_builder(4bpp,);
3814 setup_sprite_tiled_builder(8bpp,);
3816 setup_sprite_tiled_builder(4bpp,_4x);
3817 setup_sprite_tiled_builder(8bpp,_4x);
3820 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3822 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3823 s32 v, s32 width, s32 height, u32 color)
3825 u32 left_offset = u & 0x7;
3826 u32 width_rounded = width + left_offset + 7;
3828 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3829 u32 right_width = width_rounded & 0x7;
3830 u32 block_width = width_rounded / 8;
3831 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3833 u32 left_mask_bits = ~(0xFF << left_offset);
3834 u32 right_mask_bits = 0xFE << right_width;
3836 u32 texture_offset_base = u + (v * 1024);
3838 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3840 u32 blocks_remaining;
3841 u32 num_blocks = psx_gpu->num_blocks;
3842 block_struct *block = psx_gpu->blocks + num_blocks;
3844 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3845 u16 *texture_block_ptr;
3847 texture_offset_base &= ~0x7;
3849 stats_add(sprites_16bpp, 1);
3851 if(block_width == 1)
3853 u32 mask_bits = left_mask_bits | right_mask_bits;
3860 if(num_blocks > MAX_BLOCKS)
3862 flush_render_block_buffer(psx_gpu);
3864 block = psx_gpu->blocks;
3868 texture_page_ptr + (texture_offset_base & texture_mask);
3870 block->texels = *(vec_8x16u *)texture_block_ptr;
3871 block->draw_mask_bits = mask_bits;
3872 block->fb_ptr = fb_ptr;
3876 texture_offset_base += 1024;
3880 psx_gpu->num_blocks = num_blocks;
3889 blocks_remaining = block_width - 2;
3890 num_blocks += block_width;
3891 sprite_blocks += block_width;
3893 if(num_blocks > MAX_BLOCKS)
3895 flush_render_block_buffer(psx_gpu);
3896 num_blocks = block_width;
3897 block = psx_gpu->blocks;
3900 texture_offset = texture_offset_base;
3901 texture_offset_base += 1024;
3903 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3904 block->texels = *(vec_8x16u *)texture_block_ptr;
3906 block->draw_mask_bits = left_mask_bits;
3907 block->fb_ptr = fb_ptr;
3909 texture_offset += 8;
3913 while(blocks_remaining)
3915 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3916 block->texels = *(vec_8x16u *)texture_block_ptr;
3918 block->draw_mask_bits = 0;
3919 block->fb_ptr = fb_ptr;
3921 texture_offset += 8;
3928 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3929 block->texels = *(vec_8x16u *)texture_block_ptr;
3931 block->draw_mask_bits = right_mask_bits;
3932 block->fb_ptr = fb_ptr;
3934 fb_ptr += fb_ptr_pitch;
3938 psx_gpu->num_blocks = num_blocks;
3947 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3948 s32 v, s32 width, s32 height, u32 color)
3950 u32 right_width = ((width - 1) & 0x7) + 1;
3951 u32 right_mask_bits = (0xFF << right_width);
3952 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3953 u32 block_width = (width + 7) / 8;
3954 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3955 u32 blocks_remaining;
3956 u32 num_blocks = psx_gpu->num_blocks;
3957 block_struct *block = psx_gpu->blocks + num_blocks;
3959 u32 color_r = color & 0xFF;
3960 u32 color_g = (color >> 8) & 0xFF;
3961 u32 color_b = (color >> 16) & 0xFF;
3963 vec_8x16u right_mask;
3964 vec_8x16u test_mask = psx_gpu->test_mask;
3965 vec_8x16u zero_mask;
3967 sprites_untextured++;
3969 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3971 dup_8x16b(colors, color);
3972 dup_8x16b(zero_mask, 0x00);
3973 dup_8x16b(right_mask, right_mask_bits);
3974 tst_8x16b(right_mask, right_mask, test_mask);
3978 blocks_remaining = block_width - 1;
3979 num_blocks += block_width;
3982 sprite_blocks += block_width;
3985 if(num_blocks > MAX_BLOCKS)
3987 flush_render_block_buffer(psx_gpu);
3988 num_blocks = block_width;
3989 block = psx_gpu->blocks;
3992 while(blocks_remaining)
3994 block->pixels = colors;
3995 block->draw_mask = zero_mask;
3996 block->fb_ptr = fb_ptr;
4003 block->pixels = colors;
4004 block->draw_mask = right_mask;
4005 block->fb_ptr = fb_ptr;
4008 fb_ptr += fb_ptr_pitch;
4011 psx_gpu->num_blocks = num_blocks;
4017 static void __attribute__((noinline))
4018 setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4019 s32 v, s32 width, s32 height, u32 color)
4021 u32 r = color & 0xFF;
4022 u32 g = (color >> 8) & 0xFF;
4023 u32 b = (color >> 16) & 0xFF;
4024 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4026 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4028 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4033 if(psx_gpu->num_blocks)
4035 flush_render_block_buffer(psx_gpu);
4042 vram_ptr = (void *)vram_ptr16;
4043 if((uintptr_t)vram_ptr16 & 2)
4045 *vram_ptr16 = color_32bpp;
4046 vram_ptr = (void *)(vram_ptr16 + 1);
4050 while(num_width >= 4 * 2)
4052 vram_ptr[0] = color_32bpp;
4053 vram_ptr[1] = color_32bpp;
4054 vram_ptr[2] = color_32bpp;
4055 vram_ptr[3] = color_32bpp;
4061 while(num_width >= 2)
4063 *vram_ptr++ = color_32bpp;
4069 *(u16 *)vram_ptr = color_32bpp;
4077 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4078 s32 v, s32 width, s32 height, u32 color);
4080 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4081 s32 v, s32 width, s32 height, u32 color)
4083 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
4084 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
4085 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
4087 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
4093 s32 w1 = width > 512 ? 512 : width;
4094 setup_sprite_untextured_512(psx_gpu, x, y, 0, 0, w1, height, color);
4101 #define setup_sprite_blocks_switch_textured(texture_mode) \
4102 setup_sprite_##texture_mode \
4104 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4105 setup_sprite_untextured \
4107 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4108 setup_sprite_blocks_switch_##texturing(texture_mode) \
4111 #define texture_sprite_blocks_switch_4bpp() \
4112 texture_blocks_untextured \
4114 #define texture_sprite_blocks_switch_8bpp() \
4115 texture_sprite_blocks_8bpp \
4117 #define texture_sprite_blocks_switch_16bpp() \
4118 texture_blocks_untextured \
4120 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4121 texture_blocks_untextured \
4123 #define texture_sprite_blocks_switch_textured(texture_mode) \
4124 texture_sprite_blocks_switch_##texture_mode() \
4126 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4127 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4129 setup_sprite_blocks_switch(texturing, texture_mode), \
4130 texture_sprite_blocks_switch_##texturing(texture_mode), \
4131 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4133 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4136 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4137 mask_evaluate, shading, dithering, texturing, blending) \
4138 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4139 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4140 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4141 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4143 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4144 mask_evaluate, shading, dithering, texturing) \
4145 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4146 mask_evaluate, shading, dithering, texturing, unblended), \
4147 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4148 mask_evaluate, shading, dithering, texturing, blended) \
4150 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4151 mask_evaluate, shading, dithering) \
4152 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4153 mask_evaluate, shading, dithering, untextured), \
4154 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4155 mask_evaluate, shading, dithering, textured) \
4157 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4158 mask_evaluate, shading) \
4159 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4160 mask_evaluate, shading, undithered), \
4161 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4162 mask_evaluate, shading, dithered) \
4164 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4165 blend_mode, mask_evaluate) \
4166 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4167 mask_evaluate, unshaded), \
4168 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4169 mask_evaluate, shaded) \
4171 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4172 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4174 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4177 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4178 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4179 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4180 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4181 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4183 #define render_sprite_blocks_switch_block() \
4184 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4185 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4186 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4187 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4190 render_block_handler_struct render_sprite_block_handlers[] =
4192 render_sprite_blocks_switch_block()
4196 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4197 s32 width, s32 height, u32 flags, u32 color)
4199 s32 x_right = x + width - 1;
4200 s32 y_bottom = y + height - 1;
4206 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4212 if(x < psx_gpu->viewport_start_x)
4214 u32 clip = psx_gpu->viewport_start_x - x;
4220 if(y < psx_gpu->viewport_start_y)
4222 s32 clip = psx_gpu->viewport_start_y - y;
4228 if(x_right > psx_gpu->viewport_end_x)
4229 width -= x_right - psx_gpu->viewport_end_x;
4231 if(y_bottom > psx_gpu->viewport_end_y)
4232 height -= y_bottom - psx_gpu->viewport_end_y;
4234 if((width <= 0) || (height <= 0))
4238 span_pixels += width * height;
4242 u32 render_state = flags &
4243 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4244 RENDER_FLAGS_TEXTURE_MAP);
4246 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4248 if((psx_gpu->render_state != render_state) ||
4249 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4251 psx_gpu->render_state = render_state;
4252 flush_render_block_buffer(psx_gpu);
4258 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4262 if(psx_gpu->triangle_color != color)
4264 flush_render_block_buffer(psx_gpu);
4265 psx_gpu->triangle_color = color;
4268 if(color == 0x808080)
4269 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4271 render_block_handler_struct *render_block_handler =
4272 &(render_sprite_block_handlers[render_state]);
4273 psx_gpu->render_block_handler = render_block_handler;
4275 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4276 (psx_gpu, x, y, u, v, width, height, color);
4279 #define draw_pixel_line_mask_evaluate_yes() \
4280 if((*vram_ptr & 0x8000) == 0) \
4282 #define draw_pixel_line_mask_evaluate_no() \
4285 #define draw_pixel_line_shaded() \
4287 color_r = fixed_to_int(current_r); \
4288 color_g = fixed_to_int(current_g); \
4289 color_b = fixed_to_int(current_b); \
4291 current_r += gradient_r; \
4292 current_g += gradient_g; \
4293 current_b += gradient_b; \
4296 #define draw_pixel_line_unshaded() \
4298 color_r = color & 0xFF; \
4299 color_g = (color >> 8) & 0xFF; \
4300 color_b = (color >> 16) & 0xFF; \
4304 #define draw_pixel_line_dithered(_x, _y) \
4306 u32 dither_xor = _x ^ _y; \
4307 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4308 dither_offset |= (_y & 0x1) << 1; \
4309 dither_offset |= (dither_xor & 0x1) << 2; \
4310 dither_offset -= 4; \
4312 color_r += dither_offset; \
4313 color_g += dither_offset; \
4314 color_b += dither_offset; \
4335 #define draw_pixel_line_undithered(_x, _y) \
4338 #define draw_pixel_line_average() \
4339 color_r = (color_r + fb_r) / 2; \
4340 color_g = (color_g + fb_g) / 2; \
4341 color_b = (color_b + fb_b) / 2 \
4343 #define draw_pixel_line_add() \
4358 #define draw_pixel_line_subtract() \
4359 color_r = fb_r - color_r; \
4360 color_g = fb_g - color_g; \
4361 color_b = fb_b - color_b; \
4372 #define draw_pixel_line_add_fourth() \
4373 color_r = fb_r + (color_r / 4); \
4374 color_g = fb_g + (color_g / 4); \
4375 color_b = fb_b + (color_b / 4); \
4387 #define draw_pixel_line_blended(blend_mode) \
4388 s32 fb_pixel = *vram_ptr; \
4389 s32 fb_r = fb_pixel & 0x1F; \
4390 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4391 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4393 draw_pixel_line_##blend_mode() \
4395 #define draw_pixel_line_unblended(blend_mode) \
4398 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4400 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4401 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4403 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4405 draw_pixel_line_##shading(); \
4406 draw_pixel_line_##dithering(_x, _y); \
4412 draw_pixel_line_##blending(blend_mode); \
4414 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4415 psx_gpu->mask_msb; \
4419 #define update_increment(value) \
4422 #define update_decrement(value) \
4425 #define update_vram_row_increment(value) \
4428 #define update_vram_row_decrement(value) \
4431 #define compare_increment(a, b) \
4434 #define compare_decrement(a, b) \
4437 #define set_line_gradients(minor) \
4439 s32 gradient_divisor = delta_##minor; \
4440 if(gradient_divisor != 0) \
4442 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4443 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4444 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4452 current_r = fixed_center(vertex_a->r); \
4453 current_g = fixed_center(vertex_a->g); \
4454 current_b = fixed_center(vertex_a->b); \
4457 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4458 mask_evaluate, blend_mode) \
4461 error_step = delta_y * 2; \
4462 error_wrap = delta_x * 2; \
4466 set_line_gradients(x); \
4468 for(current_x = x_a; current_x <= x_b; current_x++) \
4470 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4471 mask_evaluate, blend_mode); \
4472 error += error_step; \
4475 if(error >= error_wrap) \
4477 update_##direction(current_y); \
4478 update_vram_row_##direction(); \
4479 error -= error_wrap; \
4484 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4485 mask_evaluate, blend_mode) \
4488 error_step = delta_x * 2; \
4489 error_wrap = delta_y * 2; \
4493 set_line_gradients(y); \
4495 for(current_y = y_a; compare_##direction(current_y, y_b); \
4496 update_##direction(current_y)) \
4498 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4499 mask_evaluate, blend_mode); \
4500 error += error_step; \
4501 update_vram_row_##direction(); \
4503 if(error > error_wrap) \
4507 error -= error_wrap; \
4513 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4519 if(delta_x > delta_y) \
4521 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4522 mask_evaluate, blend_mode); \
4526 draw_line_span_vertical(decrement, shading, blending, dithering, \
4527 mask_evaluate, blend_mode); \
4532 if(delta_x > delta_y) \
4534 draw_line_span_horizontal(increment, shading, blending, dithering, \
4535 mask_evaluate, blend_mode); \
4539 draw_line_span_vertical(increment, shading, blending, dithering, \
4540 mask_evaluate, blend_mode); \
4545 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4546 u32 color, int double_resolution)
4548 s32 color_r, color_g, color_b;
4549 u32 triangle_winding = 0;
4551 fixed_type gradient_r = 0;
4552 fixed_type gradient_g = 0;
4553 fixed_type gradient_b = 0;
4554 fixed_type current_r = 0;
4555 fixed_type current_g = 0;
4556 fixed_type current_b = 0;
4561 s32 delta_x, delta_y;
4572 flush_render_block_buffer(psx_gpu);
4573 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4575 vertex_struct *vertex_a = &(vertexes[0]);
4576 vertex_struct *vertex_b = &(vertexes[1]);
4584 if(vertex_a->x >= vertex_b->x)
4586 vertex_swap(vertex_a, vertex_b);
4587 (void)triangle_winding;
4596 delta_x = x_b - x_a;
4597 delta_y = y_b - y_a;
4599 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4602 if(double_resolution)
4612 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4614 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4618 if(flags & RENDER_FLAGS_SHADE)
4619 control_mask |= 0x1;
4621 if(flags & RENDER_FLAGS_BLEND)
4623 control_mask |= 0x2;
4624 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4627 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4628 control_mask |= 0x4;
4630 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4631 control_mask |= 0x8;
4633 switch(control_mask)
4636 render_line_body(unshaded, unblended, undithered, no, none);
4640 render_line_body(shaded, unblended, undithered, no, none);
4644 render_line_body(unshaded, blended, undithered, no, average);
4648 render_line_body(shaded, blended, undithered, no, average);
4652 render_line_body(unshaded, unblended, dithered, no, none);
4656 render_line_body(shaded, unblended, dithered, no, none);
4660 render_line_body(unshaded, blended, dithered, no, average);
4664 render_line_body(shaded, blended, dithered, no, average);
4668 render_line_body(unshaded, unblended, undithered, yes, none);
4672 render_line_body(shaded, unblended, undithered, yes, none);
4676 render_line_body(unshaded, blended, undithered, yes, average);
4680 render_line_body(shaded, blended, undithered, yes, average);
4684 render_line_body(unshaded, unblended, dithered, yes, none);
4688 render_line_body(shaded, unblended, dithered, yes, none);
4692 render_line_body(unshaded, blended, dithered, yes, average);
4696 render_line_body(shaded, blended, dithered, yes, average);
4700 render_line_body(unshaded, blended, undithered, no, add);
4704 render_line_body(shaded, blended, undithered, no, add);
4708 render_line_body(unshaded, blended, dithered, no, add);
4712 render_line_body(shaded, blended, dithered, no, add);
4716 render_line_body(unshaded, blended, undithered, yes, add);
4720 render_line_body(shaded, blended, undithered, yes, add);
4724 render_line_body(unshaded, blended, dithered, yes, add);
4728 render_line_body(shaded, blended, dithered, yes, add);
4732 render_line_body(unshaded, blended, undithered, no, subtract);
4736 render_line_body(shaded, blended, undithered, no, subtract);
4740 render_line_body(unshaded, blended, dithered, no, subtract);
4744 render_line_body(shaded, blended, dithered, no, subtract);
4748 render_line_body(unshaded, blended, undithered, yes, subtract);
4752 render_line_body(shaded, blended, undithered, yes, subtract);
4756 render_line_body(unshaded, blended, dithered, yes, subtract);
4760 render_line_body(shaded, blended, dithered, yes, subtract);
4764 render_line_body(unshaded, blended, undithered, no, add_fourth);
4768 render_line_body(shaded, blended, undithered, no, add_fourth);
4772 render_line_body(unshaded, blended, dithered, no, add_fourth);
4776 render_line_body(shaded, blended, dithered, no, add_fourth);
4780 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4784 render_line_body(shaded, blended, undithered, yes, add_fourth);
4788 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4792 render_line_body(shaded, blended, dithered, yes, add_fourth);
4798 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4799 u32 width, u32 height)
4801 if((width == 0) || (height == 0))
4804 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4806 u32 r = color & 0xFF;
4807 u32 g = (color >> 8) & 0xFF;
4808 u32 b = (color >> 16) & 0xFF;
4809 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4811 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4813 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4815 u32 pitch = 512 - (width / 2);
4818 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4823 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4832 vram_ptr[0] = color_32bpp;
4833 vram_ptr[1] = color_32bpp;
4834 vram_ptr[2] = color_32bpp;
4835 vram_ptr[3] = color_32bpp;
4836 vram_ptr[4] = color_32bpp;
4837 vram_ptr[5] = color_32bpp;
4838 vram_ptr[6] = color_32bpp;
4839 vram_ptr[7] = color_32bpp;
4850 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4851 u32 width, u32 height)
4853 if((width == 0) || (height == 0))
4859 u32 r = color & 0xFF;
4860 u32 g = (color >> 8) & 0xFF;
4861 u32 b = (color >> 16) & 0xFF;
4862 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4864 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4866 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4868 u32 pitch = 1024 / 2 - (width / 2);
4876 vram_ptr[0] = color_32bpp;
4877 vram_ptr[1] = color_32bpp;
4878 vram_ptr[2] = color_32bpp;
4879 vram_ptr[3] = color_32bpp;
4880 vram_ptr[4] = color_32bpp;
4881 vram_ptr[5] = color_32bpp;
4882 vram_ptr[6] = color_32bpp;
4883 vram_ptr[7] = color_32bpp;
4895 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4896 u32 width, u32 height, u32 pitch)
4898 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4900 u32 mask_msb = psx_gpu->mask_msb;
4902 if((width == 0) || (height == 0))
4905 flush_render_block_buffer(psx_gpu);
4906 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4908 for(draw_y = 0; draw_y < height; draw_y++)
4910 for(draw_x = 0; draw_x < width; draw_x++)
4912 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4920 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4921 u32 dest_x, u32 dest_y, u32 width, u32 height)
4923 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4924 dest_x, dest_y, width, height, 1024);
4928 void initialize_reciprocal_table(void)
4931 u32 height_normalized;
4932 u32 height_reciprocal;
4935 for(height = 1; height < sizeof(reciprocal_table)
4936 / sizeof(reciprocal_table[0]); height++)
4938 shift = __builtin_clz(height);
4939 height_normalized = height << shift;
4940 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4943 shift = 32 - (51 - shift);
4945 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4950 #define dither_table_row(a, b, c, d) \
4951 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4953 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4955 vec_8x16u test_mask =
4956 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4958 psx_gpu->test_mask = test_mask;
4960 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4961 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4962 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4963 psx_gpu->viewport_mask = 0;
4964 psx_gpu->current_texture_page = 0;
4965 psx_gpu->current_texture_mask = 0;
4966 psx_gpu->last_8bpp_texture_page = 0;
4968 psx_gpu->clut_settings = 0;
4969 psx_gpu->texture_settings = 0;
4970 psx_gpu->render_state = 0;
4971 psx_gpu->render_state_base = 0;
4972 psx_gpu->num_blocks = 0;
4973 psx_gpu->uvrgb_phase = 0x8000;
4975 psx_gpu->vram_ptr = vram;
4976 psx_gpu->vram_out_ptr = vram;
4978 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4979 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4980 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4982 psx_gpu->viewport_start_x = psx_gpu->viewport_start_y = 0;
4983 psx_gpu->viewport_end_x = psx_gpu->viewport_end_y = 0;
4984 psx_gpu->mask_msb = 0;
4986 psx_gpu->texture_window_x = 0;
4987 psx_gpu->texture_window_y = 0;
4988 psx_gpu->texture_mask_width = 0xFF;
4989 psx_gpu->texture_mask_height = 0xFF;
4991 psx_gpu->render_mode = 0;
4993 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4995 initialize_reciprocal_table();
4996 psx_gpu->reciprocal_table_ptr = reciprocal_table;
5003 // (minus ones(4) * 4)
5005 // d0: (1 3 5 7): x1 ^ y1
5006 // d1: (2 3 6 7): y0
5007 // d2: (4 5 6 7): x0 ^ y0
5009 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
5010 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
5011 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
5012 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
5014 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
5016 psx_gpu->saved_hres = 256;
5018 // check some offset
5019 psx_gpu->reserved_a[(offsetof(psx_gpu_struct, blocks) == psx_gpu_blocks_offset) - 1] = 0;
5025 gettimeofday(&tv, NULL);
5027 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
5030 #if 0 //def NEON_BUILD
5035 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
5040 void init_counter(void)
5043 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
5044 value |= 5; // master enable, ccnt reset
5045 value &= ~8; // ccnt divider 0
5046 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
5047 // enable cycle counter
5048 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
5051 void triangle_benchmark(psx_gpu_struct *psx_gpu)
5058 const u32 iterations = 500000;
5060 psx_gpu->num_blocks = 64;
5061 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5063 for(i = 0; i < 64; i++)
5065 memset(&(psx_gpu->blocks[i].r), 0, 16);
5070 ticks = get_counter();
5072 for(i = 0; i < iterations; i++)
5074 texture_sprite_blocks_8bpp(psx_gpu);
5077 ticks_elapsed = get_counter() - ticks;
5079 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5084 #include "psx_gpu_4x.c"
5086 // vim:ts=2:sw=2:expandtab