2 * Copyright (C) 2011 Gilead Kutnick "Exophase" <exophase@gmail.com>
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License as
6 * published by the Free Software Foundation; either version 2 of
7 * the License, or (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * General Public License for more details.
23 #include "vector_ops.h"
25 #include "psx_gpu_simd.h"
28 void dump_r_d(const char *name, void *dump);
29 void dump_r_q(const char *name, void *dump);
30 #define dumprd(n) dump_r_d(#n, n.e)
31 #define dumprq(n) dump_r_q(#n, n.e)
35 u32 span_pixel_blocks = 0;
41 u32 sprites_16bpp = 0;
42 u32 sprite_blocks = 0;
43 u32 sprites_untextured = 0;
45 u32 trivial_rejects = 0;
49 u32 texel_blocks_4bpp = 0;
50 u32 texel_blocks_8bpp = 0;
51 u32 texel_blocks_16bpp = 0;
52 u32 texel_blocks_untextured = 0;
54 u32 render_buffer_flushes = 0;
55 u32 state_changes = 0;
56 u32 left_split_triangles = 0;
57 u32 flat_triangles = 0;
58 u32 clipped_triangles = 0;
59 u32 zero_block_spans = 0;
60 u32 texture_cache_loads = 0;
61 u32 false_modulated_blocks = 0;
63 #define stats_add(stat, count) // stat += count
65 /* double size for enhancement */
66 u32 reciprocal_table[512 * 2];
69 typedef s32 fixed_type;
71 #define EDGE_STEP_BITS 32
74 #define fixed_center(value) \
75 ((((fixed_type)(value)) << FIXED_BITS) + (1 << (FIXED_BITS - 1))) \
77 #define int_to_fixed(value) \
78 (((fixed_type)(value)) << FIXED_BITS) \
80 #define fixed_to_int(value) \
81 ((value) >> FIXED_BITS) \
83 #define fixed_to_double(value) \
84 ((value) / (double)(1 << FIXED_BITS)) \
86 #define double_to_fixed(value) \
87 (fixed_type)(((value) * (double)(1 << FIXED_BITS))) \
89 typedef void (setup_blocks_function_type)(psx_gpu_struct *psx_gpu);
90 typedef void (texture_blocks_function_type)(psx_gpu_struct *psx_gpu);
91 typedef void (shade_blocks_function_type)(psx_gpu_struct *psx_gpu);
92 typedef void (blend_blocks_function_type)(psx_gpu_struct *psx_gpu);
94 typedef void (setup_sprite_function_type)(psx_gpu_struct *psx_gpu, s32 x,
95 s32 y, s32 u, s32 v, s32 width, s32 height, u32 color);
97 struct render_block_handler_struct
100 texture_blocks_function_type *texture_blocks;
101 shade_blocks_function_type *shade_blocks;
102 blend_blocks_function_type *blend_blocks;
107 u32 fixed_reciprocal(u32 denominator, u32 *_shift)
109 u32 shift = __builtin_clz(denominator);
110 u32 denominator_normalized = denominator << shift;
112 double numerator = (1ULL << 62) + denominator_normalized;
115 double denominator_normalized_dp_b;
116 u64 denominator_normalized_dp_u64;
119 double reciprocal_dp;
121 u64 numerator_u64 = (denominator_normalized >> 10) |
122 ((u64)(62 + 1023) << 52);
123 *((u64 *)(&numerator_b)) = numerator_u64;
125 denominator_normalized_dp_u64 =
126 (u64)(denominator_normalized << 21) |
127 ((u64)((denominator_normalized >> 11) + ((1022 + 31) << 20)) << 32);
128 *((u64 *)(&denominator_normalized_dp_b)) = denominator_normalized_dp_u64;
130 // Implement with a DP divide
131 reciprocal_dp = numerator / denominator_normalized_dp_b;
132 reciprocal = reciprocal_dp;
134 if(reciprocal == 0x80000001)
135 reciprocal = 0x80000000;
137 *_shift = 62 - shift;
141 double reciprocal_estimate(double a)
146 q = (int)(a * 512.0);
147 /* a in units of 1/512 rounded down */
148 r = 1.0 / (((double)q + 0.5) / 512.0); /* reciprocal r */
149 s = (int)(256.0 * r + 0.5);
151 /* r in units of 1/256 rounded to nearest */
153 return (double)s / 256.0;
156 u32 reciprocal_estimate_u32(u32 value)
159 volatile double dp_value;
160 volatile u64 *dp_value_ptr = (volatile u64 *)&dp_value;
162 if((value >> 31) == 0)
165 dp_value_u64 = (0x3FEULL << (31 + 21)) | ((u64)(value & 0x7FFFFFFF) << 21);
167 *dp_value_ptr = dp_value_u64;
169 dp_value = reciprocal_estimate(dp_value);
170 dp_value_u64 = *dp_value_ptr;
172 return (0x80000000 | ((dp_value_u64 >> 21) & 0x7FFFFFFF));
175 u32 fixed_reciprocal_nr(u32 value, u32 *_shift)
177 u32 shift = __builtin_clz(value);
178 u32 value_normalized = value << shift;
180 *_shift = 62 - shift;
182 value_normalized -= 2;
184 u32 reciprocal_normalized = reciprocal_estimate_u32(value_normalized) >> 1;
186 u32 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
187 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
188 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
189 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
190 temp = -(((u64)value_normalized * (u32)reciprocal_normalized) >> 31);
191 reciprocal_normalized = (((u64)reciprocal_normalized * temp) >> 31);
193 return reciprocal_normalized;
199 s32 triangle_signed_area_x2(s32 x0, s32 y0, s32 x1, s32 y1, s32 x2, s32 y2)
201 return ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0));
204 u32 texture_region_mask(s32 x1, s32 y1, s32 x2, s32 y2)
206 s32 coverage_x, coverage_y;
211 coverage_x = x2 >> 6;
212 coverage_y = y2 >> 8;
220 mask_down_right = ~(0xFFFFFFFF << (coverage_x + 1)) & 0xFFFF;
223 mask_down_right |= mask_down_right << 16;
225 coverage_x = x1 >> 6;
227 mask_up_left = 0xFFFF0000 << coverage_x;
229 mask_up_left = 0xFFFF0000;
231 coverage_y = y1 >> 8;
233 mask_up_left |= mask_up_left >> 16;
235 return mask_up_left & mask_down_right;
238 u32 invalidate_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
241 u32 mask = texture_region_mask(x1, y1, x2, y2);
243 psx_gpu->dirty_textures_4bpp_mask |= mask;
244 psx_gpu->dirty_textures_8bpp_mask |= mask;
245 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
250 u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1,
251 u32 y1, u32 x2, u32 y2)
253 u32 mask = texture_region_mask(x1, y1, x2, y2) &
254 psx_gpu->viewport_mask;
256 psx_gpu->dirty_textures_4bpp_mask |= mask;
257 psx_gpu->dirty_textures_8bpp_mask |= mask;
258 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
263 static void update_texture_cache_region_(psx_gpu_struct *psx_gpu,
264 u32 x1, u32 y1, u32 x2, u32 y2)
266 u32 mask = texture_region_mask(x1, y1, x2, y2);
268 u8 *texture_page_ptr;
273 psx_gpu->dirty_textures_8bpp_mask |= mask;
274 psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
276 if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
277 (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
279 texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
280 texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
281 texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
282 vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
290 texel_block = *vram_ptr;
292 texture_page_ptr[0] = texel_block & 0xF;
293 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
294 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
295 texture_page_ptr[3] = texel_block >> 12;
298 texture_page_ptr += 4;
312 psx_gpu->dirty_textures_4bpp_mask |= mask;
316 void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1,
325 update_texture_cache_region_(psx_gpu, x1, y1, x2, y2);
329 while (unlikely(w > 0));
334 void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu)
336 u32 current_texture_page = psx_gpu->current_texture_page;
337 u8 *texture_page_ptr = psx_gpu->texture_page_base;
338 u16 *vram_ptr = psx_gpu->vram_ptr;
344 vram_ptr += (current_texture_page >> 4) * 256 * 1024;
345 vram_ptr += (current_texture_page & 0xF) * 64;
347 texture_cache_loads++;
354 psx_gpu->dirty_textures_4bpp_mask &= ~(psx_gpu->current_texture_mask);
364 texel_block = *vram_ptr;
366 texture_page_ptr[0] = texel_block & 0xF;
367 texture_page_ptr[1] = (texel_block >> 4) & 0xF;
368 texture_page_ptr[2] = (texel_block >> 8) & 0xF;
369 texture_page_ptr[3] = texel_block >> 12;
372 texture_page_ptr += 4;
386 vram_ptr -= (1024 * 16) - 4;
392 vram_ptr += (16 * 1024) - (4 * 16);
397 void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
400 u16 *texture_page_ptr = psx_gpu->texture_page_base;
401 u16 *vram_ptr = psx_gpu->vram_ptr;
408 texture_cache_loads++;
410 vram_ptr += (texture_page >> 4) * 256 * 1024;
411 vram_ptr += (texture_page & 0xF) * 64;
413 if((texture_page ^ psx_gpu->current_texture_page) & 0x1)
414 texture_page_ptr += (8 * 16) * 8;
427 load_128b(texels, vram_ptr);
428 store_128b(texels, texture_page_ptr);
430 texture_page_ptr += 8;
438 vram_ptr -= (1024 * 16);
447 vram_ptr += (16 * 1024);
449 texture_page_ptr += (8 * 16) * 8;
457 void update_texture_8bpp_cache(psx_gpu_struct *psx_gpu)
459 u32 current_texture_page = psx_gpu->current_texture_page;
460 u32 update_textures =
461 psx_gpu->dirty_textures_8bpp_mask & psx_gpu->current_texture_mask;
463 psx_gpu->dirty_textures_8bpp_mask &= ~update_textures;
465 if(update_textures & (1 << current_texture_page))
467 update_texture_8bpp_cache_slice(psx_gpu, current_texture_page);
468 update_textures &= ~(1 << current_texture_page);
473 u32 adjacent_texture_page = ((current_texture_page + 1) & 0xF) |
474 (current_texture_page & 0x10);
476 update_texture_8bpp_cache_slice(psx_gpu, adjacent_texture_page);
480 void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
482 if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
483 (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
485 u32 num_blocks_dest = 0;
486 block_struct *block_src = psx_gpu->blocks;
487 block_struct *block_dest = psx_gpu->blocks;
489 u16 *vram_ptr = psx_gpu->vram_ptr;
492 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
494 for(i = 0; i < psx_gpu->num_blocks; i++)
496 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
497 if(fb_offset & (1 << 11))
499 *block_dest = *block_src;
508 for(i = 0; i < psx_gpu->num_blocks; i++)
510 u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
511 if((fb_offset & (1 << 11)) == 0)
513 *block_dest = *block_src;
521 psx_gpu->num_blocks = num_blocks_dest;
524 if(psx_gpu->num_blocks)
526 render_block_handler_struct *render_block_handler =
527 psx_gpu->render_block_handler;
529 render_block_handler->texture_blocks(psx_gpu);
530 render_block_handler->shade_blocks(psx_gpu);
531 render_block_handler->blend_blocks(psx_gpu);
534 span_pixel_blocks += psx_gpu->num_blocks;
535 render_buffer_flushes++;
538 psx_gpu->num_blocks = 0;
545 #define setup_gradient_calculation_input(set, vertex) \
546 /* First type is: uvrg bxxx xxxx */\
547 /* Second type is: yyyy ybyy uvrg */\
548 /* Since x_a and y_c are the same the same variable is used for both. */\
549 x##set##_a_y##set##_c.e[0] = vertex->u; \
550 x##set##_a_y##set##_c.e[1] = vertex->v; \
551 x##set##_a_y##set##_c.e[2] = vertex->r; \
552 x##set##_a_y##set##_c.e[3] = vertex->g; \
553 dup_4x16b(x##set##_b, vertex->x); \
554 dup_4x16b(x##set##_c, vertex->x); \
555 dup_4x16b(y##set##_a, vertex->y); \
556 dup_4x16b(y##set##_b, vertex->y); \
557 x##set##_b.e[0] = vertex->b; \
558 y##set##_b.e[1] = vertex->b \
561 void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a,
562 vertex_struct *b, vertex_struct *c)
564 u32 triangle_area = psx_gpu->triangle_area;
565 u32 winding_mask_scalar;
567 u32 triangle_area_shift;
568 u64 triangle_area_reciprocal =
569 fixed_reciprocal(triangle_area, &triangle_area_shift);
570 triangle_area_shift = -(triangle_area_shift - FIXED_BITS);
572 // ((x1 - x0) * (y2 - y1)) - ((x2 - x1) * (y1 - y0)) =
573 // ( d0 * d1 ) - ( d2 * d3 ) =
574 // ( m0 ) - ( m1 ) = gradient
576 // This is split to do 12 elements at a time over three sets: a, b, and c.
577 // Technically we only need to do 10 elements (uvrgb_x and uvrgb_y), so
578 // two of the slots are unused.
580 // Inputs are all 16-bit signed. The m0/m1 results are 32-bit signed, as
583 vec_4x16s x0_a_y0_c, x0_b, x0_c;
584 vec_4x16s y0_a, y0_b;
585 vec_4x16s x1_a_y1_c, x1_b, x1_c;
586 vec_4x16s y1_a, y1_b;
587 vec_4x16s x2_a_y2_c, x2_b, x2_c;
588 vec_4x16s y2_a, y2_b;
592 vec_4x32u uvrgb_phase;
594 vec_4x16s d0_a_d3_c, d0_b, d0_c;
595 vec_4x16s d1_a, d1_b, d1_c_d2_a;
596 vec_4x16s d2_b, d2_c;
597 vec_4x16s d3_a, d3_b;
599 vec_4x32s m0_a, m0_b, m0_c;
600 vec_4x32s m1_a, m1_b, m1_c;
602 vec_4x32u gradient_area_a, gradient_area_c;
603 vec_2x32u gradient_area_b;
605 vec_4x32u gradient_area_sign_a, gradient_area_sign_c;
606 vec_2x32u gradient_area_sign_b;
607 vec_4x32u winding_mask;
609 vec_2x64u gradient_wide_a0, gradient_wide_a1;
610 vec_2x64u gradient_wide_c0, gradient_wide_c1;
611 vec_2x64u gradient_wide_b;
613 vec_4x32u gradient_a, gradient_c;
614 vec_2x32u gradient_b;
615 vec_16x8s gradient_shift;
617 setup_gradient_calculation_input(0, a);
618 setup_gradient_calculation_input(1, b);
619 setup_gradient_calculation_input(2, c);
621 dup_4x32b(uvrgb_phase, psx_gpu->uvrgb_phase);
622 shl_long_4x16b(uvrg_base, x0_a_y0_c, 16);
623 shl_long_4x16b(b_base, x0_b, 16);
625 add_4x32b(uvrg_base, uvrg_base, uvrgb_phase);
626 add_4x32b(b_base, b_base, uvrgb_phase);
628 // Can probably pair these, but it'll require careful register allocation
629 sub_4x16b(d0_a_d3_c, x1_a_y1_c, x0_a_y0_c);
630 sub_4x16b(d1_c_d2_a, x2_a_y2_c, x1_a_y1_c);
632 sub_4x16b(d0_b, x1_b, x0_b);
633 sub_4x16b(d0_c, x1_c, x0_c);
635 sub_4x16b(d1_a, y2_a, y1_a);
636 sub_4x16b(d1_b, y2_b, y1_b);
638 sub_4x16b(d2_b, x2_b, x1_b);
639 sub_4x16b(d2_c, x2_c, x1_c);
641 sub_4x16b(d3_a, y1_a, y0_a);
642 sub_4x16b(d3_b, y1_b, y0_b);
644 mul_long_4x16b(m0_a, d0_a_d3_c, d1_a);
645 mul_long_4x16b(m0_b, d0_b, d1_b);
646 mul_long_4x16b(m0_c, d0_c, d1_c_d2_a);
648 mul_long_4x16b(m1_a, d1_c_d2_a, d3_a);
649 mul_long_4x16b(m1_b, d2_b, d3_b);
650 mul_long_4x16b(m1_c, d2_c, d0_a_d3_c);
652 sub_4x32b(gradient_area_a, m0_a, m1_a);
653 sub_2x32b(gradient_area_b, m0_b.low, m1_b.low);
654 sub_4x32b(gradient_area_c, m0_c, m1_c);
656 cmpltz_4x32b(gradient_area_sign_a, gradient_area_a);
657 cmpltz_2x32b(gradient_area_sign_b, gradient_area_b);
658 cmpltz_4x32b(gradient_area_sign_c, gradient_area_c);
660 abs_4x32b(gradient_area_a, gradient_area_a);
661 abs_2x32b(gradient_area_b, gradient_area_b);
662 abs_4x32b(gradient_area_c, gradient_area_c);
664 winding_mask_scalar = -psx_gpu->triangle_winding;
666 dup_4x32b(winding_mask, winding_mask_scalar);
667 eor_4x32b(gradient_area_sign_a, gradient_area_sign_a, winding_mask);
668 eor_2x32b(gradient_area_sign_b, gradient_area_sign_b, winding_mask);
669 eor_4x32b(gradient_area_sign_c, gradient_area_sign_c, winding_mask);
671 mul_scalar_long_2x32b(gradient_wide_a0,
672 vector_cast(vec_2x32s, gradient_area_a.low),
673 (s64)triangle_area_reciprocal);
674 mul_scalar_long_2x32b(gradient_wide_a1,
675 vector_cast(vec_2x32s, gradient_area_a.high),
676 (s64)triangle_area_reciprocal);
677 mul_scalar_long_2x32b(gradient_wide_b,
678 vector_cast(vec_2x32s, gradient_area_b),
679 (s64)triangle_area_reciprocal);
680 mul_scalar_long_2x32b(gradient_wide_c0,
681 vector_cast(vec_2x32s, gradient_area_c.low),
682 (s64)triangle_area_reciprocal);
683 mul_scalar_long_2x32b(gradient_wide_c1,
684 vector_cast(vec_2x32s, gradient_area_c.high),
685 (s64)triangle_area_reciprocal);
687 dup_16x8b(gradient_shift, triangle_area_shift);
688 shl_reg_2x64b(gradient_wide_a0, gradient_wide_a0,
689 vector_cast(vec_2x64u, gradient_shift));
690 shl_reg_2x64b(gradient_wide_a1, gradient_wide_a1,
691 vector_cast(vec_2x64u, gradient_shift));
692 shl_reg_2x64b(gradient_wide_b, gradient_wide_b,
693 vector_cast(vec_2x64u, gradient_shift));
694 shl_reg_2x64b(gradient_wide_c0, gradient_wide_c0,
695 vector_cast(vec_2x64u, gradient_shift));
696 shl_reg_2x64b(gradient_wide_c1, gradient_wide_c1,
697 vector_cast(vec_2x64u, gradient_shift));
699 mov_narrow_2x64b(gradient_a.low, gradient_wide_a0);
700 mov_narrow_2x64b(gradient_a.high, gradient_wide_a1);
701 mov_narrow_2x64b(gradient_b, gradient_wide_b);
702 mov_narrow_2x64b(gradient_c.low, gradient_wide_c0);
703 mov_narrow_2x64b(gradient_c.high, gradient_wide_c1);
705 shl_4x32b(gradient_a, gradient_a, 4);
706 shl_2x32b(gradient_b, gradient_b, 4);
707 shl_4x32b(gradient_c, gradient_c, 4);
709 eor_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
710 eor_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
711 eor_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
713 sub_4x32b(gradient_a, gradient_a, gradient_area_sign_a);
714 sub_2x32b(gradient_b, gradient_b, gradient_area_sign_b);
715 sub_4x32b(gradient_c, gradient_c, gradient_area_sign_c);
717 u32 left_adjust = a->x;
718 mls_scalar_4x32b(uvrg_base, gradient_a, left_adjust);
719 mls_scalar_2x32b(b_base.low, gradient_b, left_adjust);
729 eor_4x32b(zero, zero, zero);
730 add_4x32b(uvrg_dx2, gradient_a, gradient_a);
731 add_2x32b(b_dx2, gradient_b, gradient_b);
732 add_4x32b(uvrg_dx3, gradient_a, uvrg_dx2);
733 add_2x32b(b_dx3, gradient_b, b_dx2);
735 // Can be done with vst4, assuming that the zero, dx, dx2, and dx3 are
737 psx_gpu->u_block_span.e[0] = zero.e[0];
738 psx_gpu->u_block_span.e[1] = gradient_a.e[0];
739 psx_gpu->u_block_span.e[2] = uvrg_dx2.e[0];
740 psx_gpu->u_block_span.e[3] = uvrg_dx3.e[0];
742 psx_gpu->v_block_span.e[0] = zero.e[1];
743 psx_gpu->v_block_span.e[1] = gradient_a.e[1];
744 psx_gpu->v_block_span.e[2] = uvrg_dx2.e[1];
745 psx_gpu->v_block_span.e[3] = uvrg_dx3.e[1];
747 psx_gpu->r_block_span.e[0] = zero.e[2];
748 psx_gpu->r_block_span.e[1] = gradient_a.e[2];
749 psx_gpu->r_block_span.e[2] = uvrg_dx2.e[2];
750 psx_gpu->r_block_span.e[3] = uvrg_dx3.e[2];
752 psx_gpu->g_block_span.e[0] = zero.e[3];
753 psx_gpu->g_block_span.e[1] = gradient_a.e[3];
754 psx_gpu->g_block_span.e[2] = uvrg_dx2.e[3];
755 psx_gpu->g_block_span.e[3] = uvrg_dx3.e[3];
757 psx_gpu->b_block_span.e[0] = zero.e[0];
758 psx_gpu->b_block_span.e[1] = gradient_b.e[0];
759 psx_gpu->b_block_span.e[2] = b_dx2.e[0];
760 psx_gpu->b_block_span.e[3] = b_dx3.e[0];
762 psx_gpu->uvrg = uvrg_base;
763 psx_gpu->b = b_base.e[0];
765 psx_gpu->uvrg_dx = gradient_a;
766 psx_gpu->uvrg_dy = gradient_c;
767 psx_gpu->b_dy = gradient_b.e[1];
771 #define vector_check(_a, _b) \
772 if(memcmp(&_a, &_b, sizeof(_b))) \
774 if(sizeof(_b) == 8) \
776 printf("mismatch on %s vs %s: (%x %x) vs (%x %x)\n", \
777 #_a, #_b, _a.e[0], _a.e[1], _b.e[0], _b.e[1]); \
781 printf("mismatch on %s vs %s: (%x %x %x %x) vs (%x %x %x %x)\n", \
782 #_a, #_b, _a.e[0], _a.e[1], _a.e[2], _a.e[3], _b.e[0], _b.e[1], \
787 #define scalar_check(_a, _b) \
789 printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \
792 #if !defined(NEON_BUILD) && !defined(NDEBUG)
793 static void setup_spans_debug_check(psx_gpu_struct *psx_gpu,
794 edge_data_struct *span_edge_data_element)
796 u32 _num_spans = span_edge_data_element - psx_gpu->span_edge_data;
797 if (_num_spans > MAX_SPANS)
798 *(volatile int *)0 = 1;
799 if (_num_spans < psx_gpu->num_spans)
801 if(span_edge_data_element->num_blocks > MAX_BLOCKS_PER_ROW)
802 *(volatile int *)0 = 2;
803 if(span_edge_data_element->y >= 2048)
804 *(volatile int *)0 = 3;
808 #define setup_spans_debug_check(psx_gpu, span_edge_data_element)
811 #define setup_spans_prologue_alternate_yes() \
812 vec_2x64s alternate_x; \
813 vec_2x64s alternate_dx_dy; \
814 vec_4x32s alternate_x_32; \
815 vec_4x16u alternate_x_16; \
817 vec_4x16u alternate_select; \
818 vec_4x16s y_mid_point; \
822 s32 edge_dx_dy_alt; \
825 #define setup_spans_prologue_alternate_no() \
827 #define setup_spans_prologue(alternate_active) \
828 edge_data_struct *span_edge_data; \
829 vec_4x32u *span_uvrg_offset; \
830 u32 *span_b_offset; \
834 vec_2x64s edges_xy; \
835 vec_2x32s edges_dx_dy; \
836 vec_2x32u edge_shifts; \
838 vec_2x64s left_x, right_x; \
839 vec_2x64s left_dx_dy, right_dx_dy; \
840 vec_4x32s left_x_32, right_x_32; \
841 vec_8x16s left_right_x_16; \
843 vec_8x16s left_edge; \
844 vec_8x16s right_edge; \
845 vec_4x16u span_shift; \
849 vec_4x16u c_0xFFFE; \
852 vec_2x32s x_starts; \
861 vec_4x32u uvrg = psx_gpu->uvrg; \
862 vec_4x32u uvrg_dy = psx_gpu->uvrg_dy; \
863 u32 b = psx_gpu->b; \
864 u32 b_dy = psx_gpu->b_dy; \
866 dup_2x32b(c_0x01, 0x01); \
867 setup_spans_prologue_alternate_##alternate_active() \
869 #define setup_spans_prologue_b() \
870 span_edge_data = psx_gpu->span_edge_data; \
871 span_uvrg_offset = psx_gpu->span_uvrg_offset; \
872 span_b_offset = psx_gpu->span_b_offset; \
874 vec_8x16u c_0x0001; \
875 vec_4x16u c_max_blocks_per_row; \
877 dup_8x16b(c_0x0001, 0x0001); \
878 dup_8x16b(left_edge, psx_gpu->viewport_start_x); \
879 dup_8x16b(right_edge, psx_gpu->viewport_end_x); \
880 add_8x16b(right_edge, right_edge, c_0x0001); \
881 dup_4x16b(c_0x04, 0x04); \
882 dup_4x16b(c_0x07, 0x07); \
883 dup_4x16b(c_0xFFFE, 0xFFFE); \
884 dup_4x16b(c_max_blocks_per_row, MAX_BLOCKS_PER_ROW); \
887 #define compute_edge_delta_x2() \
890 vec_2x32s height_reciprocals; \
891 vec_2x32s heights_b; \
894 u32 edge_shift = reciprocal_table[height]; \
896 dup_2x32b(heights, height); \
897 sub_2x32b(widths, x_ends, x_starts); \
899 dup_2x32b(edge_shifts, edge_shift); \
900 sub_2x32b(heights_b, heights, c_0x01); \
901 shr_2x32b(height_reciprocals, edge_shifts, 10); \
903 mla_2x32b(heights_b, x_starts, heights); \
904 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
905 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
906 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
909 #define compute_edge_delta_x3(start_c, height_a, height_b) \
912 vec_2x32s height_reciprocals; \
913 vec_2x32s heights_b; \
918 u32 height_reciprocal_alt; \
920 heights.e[0] = height_a; \
921 heights.e[1] = height_b; \
923 edge_shifts.e[0] = reciprocal_table[height_a]; \
924 edge_shifts.e[1] = reciprocal_table[height_b]; \
925 edge_shift_alt = reciprocal_table[height_minor_b]; \
927 sub_2x32b(widths, x_ends, x_starts); \
928 width_alt = x_c - start_c; \
930 shr_2x32b(height_reciprocals, edge_shifts, 10); \
931 height_reciprocal_alt = edge_shift_alt >> 10; \
933 bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
934 edge_shift_alt &= 0x1F; \
936 sub_2x32b(heights_b, heights, c_0x01); \
937 height_b_alt = height_minor_b - 1; \
939 mla_2x32b(heights_b, x_starts, heights); \
940 height_b_alt += height_minor_b * start_c; \
942 mul_long_2x32b(edges_xy, heights_b, height_reciprocals); \
943 edge_alt = (s64)height_b_alt * height_reciprocal_alt; \
945 mul_2x32b(edges_dx_dy, widths, height_reciprocals); \
946 edge_dx_dy_alt = width_alt * height_reciprocal_alt; \
950 #define setup_spans_adjust_y_up() \
951 sub_4x32b(y_x4, y_x4, c_0x04) \
953 #define setup_spans_adjust_y_down() \
954 add_4x32b(y_x4, y_x4, c_0x04) \
956 #define setup_spans_adjust_interpolants_up() \
957 sub_4x32b(uvrg, uvrg, uvrg_dy); \
960 #define setup_spans_adjust_interpolants_down() \
961 add_4x32b(uvrg, uvrg, uvrg_dy); \
965 #define setup_spans_clip_interpolants_increment() \
966 mla_scalar_4x32b(uvrg, uvrg_dy, clip); \
969 #define setup_spans_clip_interpolants_decrement() \
970 mls_scalar_4x32b(uvrg, uvrg_dy, clip); \
973 #define setup_spans_clip_alternate_yes() \
974 edge_alt += edge_dx_dy_alt * (s64)(clip) \
976 #define setup_spans_clip_alternate_no() \
978 #define setup_spans_clip(direction, alternate_active) \
980 clipped_triangles++; \
981 mla_scalar_long_2x32b(edges_xy, edges_dx_dy, (s64)clip); \
982 setup_spans_clip_alternate_##alternate_active(); \
983 setup_spans_clip_interpolants_##direction(); \
987 #define setup_spans_adjust_edges_alternate_no(left_index, right_index) \
989 vec_2x64u edge_shifts_64; \
990 vec_2x64s edges_dx_dy_64; \
992 mov_wide_2x32b(edge_shifts_64, edge_shifts); \
993 shl_variable_2x64b(edges_xy, edges_xy, edge_shifts_64); \
995 mov_wide_2x32b(edges_dx_dy_64, edges_dx_dy); \
996 shl_variable_2x64b(edges_dx_dy_64, edges_dx_dy_64, edge_shifts_64); \
998 left_x.e[0] = edges_xy.e[left_index]; \
999 right_x.e[0] = edges_xy.e[right_index]; \
1001 left_dx_dy.e[0] = edges_dx_dy_64.e[left_index]; \
1002 left_dx_dy.e[1] = edges_dx_dy_64.e[left_index]; \
1003 right_dx_dy.e[0] = edges_dx_dy_64.e[right_index]; \
1004 right_dx_dy.e[1] = edges_dx_dy_64.e[right_index]; \
1006 add_1x64b(left_x.high, left_x.low, left_dx_dy.low); \
1007 add_1x64b(right_x.high, right_x.low, right_dx_dy.low); \
1009 add_2x64b(left_dx_dy, left_dx_dy, left_dx_dy); \
1010 add_2x64b(right_dx_dy, right_dx_dy, right_dx_dy); \
1013 #define setup_spans_adjust_edges_alternate_yes(left_index, right_index) \
1015 setup_spans_adjust_edges_alternate_no(left_index, right_index); \
1016 s64 edge_dx_dy_alt_64; \
1018 dup_4x16b(y_mid_point, y_b); \
1020 edge_alt <<= edge_shift_alt; \
1021 edge_dx_dy_alt_64 = (s64)edge_dx_dy_alt << edge_shift_alt; \
1023 alternate_x.e[0] = edge_alt; \
1024 alternate_dx_dy.e[0] = edge_dx_dy_alt_64; \
1025 alternate_dx_dy.e[1] = edge_dx_dy_alt_64; \
1027 add_1x64b(alternate_x.high, alternate_x.low, alternate_dx_dy.low); \
1028 add_2x64b(alternate_dx_dy, alternate_dx_dy, alternate_dx_dy); \
1032 #define setup_spans_y_select_up() \
1033 cmplt_4x16b(alternate_select, y_x4, y_mid_point) \
1035 #define setup_spans_y_select_down() \
1036 cmpgt_4x16b(alternate_select, y_x4, y_mid_point) \
1038 #define setup_spans_y_select_alternate_yes(direction) \
1039 setup_spans_y_select_##direction() \
1041 #define setup_spans_y_select_alternate_no(direction) \
1043 #define setup_spans_alternate_select_left() \
1044 bit_4x16b(left_right_x_16.low, alternate_x_16, alternate_select) \
1046 #define setup_spans_alternate_select_right() \
1047 bit_4x16b(left_right_x_16.high, alternate_x_16, alternate_select) \
1049 #define setup_spans_alternate_select_none() \
1051 #define setup_spans_increment_alternate_yes() \
1052 shr_narrow_2x64b(alternate_x_32.low, alternate_x, 32); \
1053 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1054 shr_narrow_2x64b(alternate_x_32.high, alternate_x, 32); \
1055 add_2x64b(alternate_x, alternate_x, alternate_dx_dy); \
1056 mov_narrow_4x32b(alternate_x_16, alternate_x_32) \
1058 #define setup_spans_increment_alternate_no() \
1060 #define setup_spans_set_x4(alternate, direction, alternate_active) \
1062 span_uvrg_offset[0] = uvrg; \
1063 span_b_offset[0] = b; \
1064 setup_spans_adjust_interpolants_##direction(); \
1066 span_uvrg_offset[1] = uvrg; \
1067 span_b_offset[1] = b; \
1068 setup_spans_adjust_interpolants_##direction(); \
1070 span_uvrg_offset[2] = uvrg; \
1071 span_b_offset[2] = b; \
1072 setup_spans_adjust_interpolants_##direction(); \
1074 span_uvrg_offset[3] = uvrg; \
1075 span_b_offset[3] = b; \
1076 setup_spans_adjust_interpolants_##direction(); \
1078 span_uvrg_offset += 4; \
1079 span_b_offset += 4; \
1081 shr_narrow_2x64b(left_x_32.low, left_x, 32); \
1082 shr_narrow_2x64b(right_x_32.low, right_x, 32); \
1084 add_2x64b(left_x, left_x, left_dx_dy); \
1085 add_2x64b(right_x, right_x, right_dx_dy); \
1087 shr_narrow_2x64b(left_x_32.high, left_x, 32); \
1088 shr_narrow_2x64b(right_x_32.high, right_x, 32); \
1090 add_2x64b(left_x, left_x, left_dx_dy); \
1091 add_2x64b(right_x, right_x, right_dx_dy); \
1093 mov_narrow_4x32b(left_right_x_16.low, left_x_32); \
1094 mov_narrow_4x32b(left_right_x_16.high, right_x_32); \
1096 setup_spans_increment_alternate_##alternate_active(); \
1097 setup_spans_y_select_alternate_##alternate_active(direction); \
1098 setup_spans_alternate_select_##alternate(); \
1100 max_8x16b(left_right_x_16, left_right_x_16, left_edge); \
1101 min_8x16b(left_right_x_16, left_right_x_16, right_edge); \
1103 sub_4x16b(left_right_x_16.high, left_right_x_16.high, left_right_x_16.low); \
1104 add_4x16b(left_right_x_16.high, left_right_x_16.high, c_0x07); \
1105 and_4x16b(span_shift, left_right_x_16.high, c_0x07); \
1106 shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \
1107 shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \
1108 min_4x16b(left_right_x_16.high, left_right_x_16.high, c_max_blocks_per_row); \
1111 for(i = 0; i < 4; i++) \
1113 span_edge_data[i].left_x = left_right_x_16.low.e[i]; \
1114 span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \
1115 span_edge_data[i].right_mask = span_shift.e[i]; \
1116 span_edge_data[i].y = y_x4.e[i]; \
1117 setup_spans_debug_check(psx_gpu, &span_edge_data[i]); \
1120 span_edge_data += 4; \
1122 setup_spans_adjust_y_##direction(); \
1126 #define setup_spans_alternate_adjust_yes() \
1127 edge_alt -= edge_dx_dy_alt * (s64)height_minor_a \
1129 #define setup_spans_alternate_adjust_no() \
1132 #define setup_spans_down(left_index, right_index, alternate, alternate_active) \
1133 setup_spans_alternate_adjust_##alternate_active(); \
1134 if(y_c > psx_gpu->viewport_end_y) \
1135 height -= y_c - psx_gpu->viewport_end_y - 1; \
1137 clip = psx_gpu->viewport_start_y - y_a; \
1142 setup_spans_clip(increment, alternate_active); \
1145 setup_spans_prologue_b(); \
1152 y_x4.e[1] = y_a + 1; \
1153 y_x4.e[2] = y_a + 2; \
1154 y_x4.e[3] = y_a + 3; \
1155 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1158 psx_gpu->num_spans = height; \
1161 setup_spans_set_x4(alternate, down, alternate_active); \
1163 } while(height > 0); \
1167 #define setup_spans_alternate_pre_increment_yes() \
1168 edge_alt += edge_dx_dy_alt \
1170 #define setup_spans_alternate_pre_increment_no() \
1172 #define setup_spans_up_decrement_height_yes() \
1175 #define setup_spans_up_decrement_height_no() \
1178 #define setup_spans_up(left_index, right_index, alternate, alternate_active) \
1179 setup_spans_alternate_adjust_##alternate_active(); \
1182 if(y_c < psx_gpu->viewport_start_y) \
1183 height -= psx_gpu->viewport_start_y - y_c; \
1185 setup_spans_up_decrement_height_##alternate_active(); \
1187 clip = y_a - psx_gpu->viewport_end_y; \
1192 setup_spans_clip(decrement, alternate_active); \
1195 setup_spans_prologue_b(); \
1202 y_x4.e[1] = y_a - 1; \
1203 y_x4.e[2] = y_a - 2; \
1204 y_x4.e[3] = y_a - 3; \
1205 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy); \
1206 setup_spans_alternate_pre_increment_##alternate_active(); \
1207 setup_spans_adjust_edges_alternate_##alternate_active(left_index, \
1209 setup_spans_adjust_interpolants_up(); \
1211 psx_gpu->num_spans = height; \
1214 setup_spans_set_x4(alternate, up, alternate_active); \
1219 #define index_left 0
1220 #define index_right 1
1222 #define setup_spans_up_up(minor, major) \
1223 setup_spans_prologue(yes); \
1224 s32 height_minor_a = y_a - y_b; \
1225 s32 height_minor_b = y_b - y_c; \
1226 s32 height = y_a - y_c; \
1228 dup_2x32b(x_starts, x_a); \
1229 x_ends.e[0] = x_c; \
1230 x_ends.e[1] = x_b; \
1232 compute_edge_delta_x3(x_b, height, height_minor_a); \
1233 setup_spans_up(index_##major, index_##minor, minor, yes) \
1238 void setup_spans_up_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1239 vertex_struct *v_b, vertex_struct *v_c)
1241 setup_spans_up_up(left, right);
1244 void setup_spans_up_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1245 vertex_struct *v_b, vertex_struct *v_c)
1247 setup_spans_up_up(right, left);
1250 #define setup_spans_down_down(minor, major) \
1251 setup_spans_prologue(yes); \
1252 s32 height_minor_a = y_b - y_a; \
1253 s32 height_minor_b = y_c - y_b; \
1254 s32 height = y_c - y_a; \
1256 dup_2x32b(x_starts, x_a); \
1257 x_ends.e[0] = x_c; \
1258 x_ends.e[1] = x_b; \
1260 compute_edge_delta_x3(x_b, height, height_minor_a); \
1261 setup_spans_down(index_##major, index_##minor, minor, yes) \
1263 void setup_spans_down_left(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1264 vertex_struct *v_b, vertex_struct *v_c)
1266 setup_spans_down_down(left, right);
1269 void setup_spans_down_right(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1270 vertex_struct *v_b, vertex_struct *v_c)
1272 setup_spans_down_down(right, left);
1275 #define setup_spans_up_flat() \
1276 s32 height = y_a - y_c; \
1279 compute_edge_delta_x2(); \
1280 setup_spans_up(index_left, index_right, none, no) \
1282 void setup_spans_up_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1283 vertex_struct *v_b, vertex_struct *v_c)
1285 setup_spans_prologue(no);
1286 x_starts.e[0] = x_a;
1287 x_starts.e[1] = x_b;
1288 dup_2x32b(x_ends, x_c);
1290 setup_spans_up_flat();
1293 void setup_spans_up_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1294 vertex_struct *v_b, vertex_struct *v_c)
1296 setup_spans_prologue(no);
1297 dup_2x32b(x_starts, x_a);
1301 setup_spans_up_flat();
1304 #define setup_spans_down_flat() \
1305 s32 height = y_c - y_a; \
1308 compute_edge_delta_x2(); \
1309 setup_spans_down(index_left, index_right, none, no) \
1311 void setup_spans_down_a(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1312 vertex_struct *v_b, vertex_struct *v_c)
1314 setup_spans_prologue(no);
1315 x_starts.e[0] = x_a;
1316 x_starts.e[1] = x_b;
1317 dup_2x32b(x_ends, x_c);
1319 setup_spans_down_flat();
1322 void setup_spans_down_b(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1323 vertex_struct *v_b, vertex_struct *v_c)
1325 setup_spans_prologue(no);
1326 dup_2x32b(x_starts, x_a);
1330 setup_spans_down_flat();
1333 void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a,
1334 vertex_struct *v_b, vertex_struct *v_c)
1336 setup_spans_prologue(no);
1344 s32 height_minor_a = y_a - y_b;
1345 s32 height_minor_b = y_c - y_a;
1346 s32 height_major = y_c - y_b;
1348 vec_2x64s edges_xy_b;
1349 vec_2x32s edges_dx_dy_b;
1350 vec_2x32u edge_shifts_b;
1352 vec_2x32s height_increment;
1354 x_starts.e[0] = x_a;
1355 x_starts.e[1] = x_c;
1356 dup_2x32b(x_ends, x_b);
1358 compute_edge_delta_x3(x_a, height_minor_a, height_major);
1360 height_increment.e[0] = 0;
1361 height_increment.e[1] = height_minor_b;
1363 mla_long_2x32b(edges_xy, edges_dx_dy, height_increment);
1365 edges_xy_b.e[0] = edge_alt;
1366 edges_xy_b.e[1] = edges_xy.e[1];
1368 edge_shifts_b = edge_shifts;
1369 edge_shifts_b.e[0] = edge_shift_alt;
1371 neg_2x32b(edges_dx_dy_b, edges_dx_dy);
1372 edges_dx_dy_b.e[0] = edge_dx_dy_alt;
1376 if(y_b < psx_gpu->viewport_start_y)
1377 height_minor_a -= psx_gpu->viewport_start_y - y_b;
1379 clip = y_a - psx_gpu->viewport_end_y;
1382 height_minor_a -= clip;
1384 setup_spans_clip(decrement, no);
1387 setup_spans_prologue_b();
1389 if (height_minor_a > 512)
1390 height_minor_a = 512;
1391 if (height_minor_a > 0)
1394 y_x4.e[1] = y_a - 1;
1395 y_x4.e[2] = y_a - 2;
1396 y_x4.e[3] = y_a - 3;
1397 add_wide_2x32b(edges_xy, edges_xy, edges_dx_dy);
1398 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1399 setup_spans_adjust_interpolants_up();
1401 psx_gpu->num_spans = height_minor_a;
1402 while(height_minor_a > 0)
1404 setup_spans_set_x4(none, up, no);
1405 height_minor_a -= 4;
1408 span_edge_data += height_minor_a;
1409 span_uvrg_offset += height_minor_a;
1410 span_b_offset += height_minor_a;
1413 edges_xy = edges_xy_b;
1414 edges_dx_dy = edges_dx_dy_b;
1415 edge_shifts = edge_shifts_b;
1417 uvrg = psx_gpu->uvrg;
1422 if(y_c > psx_gpu->viewport_end_y)
1423 height_minor_b -= y_c - psx_gpu->viewport_end_y - 1;
1425 clip = psx_gpu->viewport_start_y - y_a;
1428 height_minor_b -= clip;
1430 setup_spans_clip(increment, no);
1433 if (height_minor_b > 512)
1434 height_minor_b = 512;
1435 if (height_minor_b > 0)
1438 y_x4.e[1] = y_a + 1;
1439 y_x4.e[2] = y_a + 2;
1440 y_x4.e[3] = y_a + 3;
1441 setup_spans_adjust_edges_alternate_no(index_left, index_right);
1443 // FIXME: overflow corner case
1444 if(psx_gpu->num_spans + height_minor_b == MAX_SPANS)
1445 height_minor_b &= ~3;
1447 psx_gpu->num_spans += height_minor_b;
1448 while(height_minor_b > 0)
1450 setup_spans_set_x4(none, down, no);
1451 height_minor_b -= 4;
1455 left_split_triangles++;
1461 #define dither_table_entry_normal(value) \
1465 #define setup_blocks_load_msb_mask_indirect() \
1467 #define setup_blocks_load_msb_mask_direct() \
1468 vec_8x16u msb_mask; \
1469 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
1472 #define setup_blocks_variables_shaded_textured(target) \
1473 vec_4x32u u_block; \
1474 vec_4x32u v_block; \
1475 vec_4x32u r_block; \
1476 vec_4x32u g_block; \
1477 vec_4x32u b_block; \
1478 vec_4x32u uvrg_dx = psx_gpu->uvrg_dx; \
1479 vec_4x32u uvrg_dx4; \
1480 vec_4x32u uvrg_dx8; \
1482 u32 b_dx = psx_gpu->b_block_span.e[1]; \
1483 u32 b_dx4 = b_dx << 2; \
1484 u32 b_dx8 = b_dx << 3; \
1487 vec_16x8u texture_mask; \
1488 shl_4x32b(uvrg_dx4, uvrg_dx, 2); \
1489 shl_4x32b(uvrg_dx8, uvrg_dx, 3); \
1490 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1491 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1493 #define setup_blocks_variables_shaded_untextured(target) \
1494 vec_4x32u r_block; \
1495 vec_4x32u g_block; \
1496 vec_4x32u b_block; \
1498 vec_4x32u rgb_dx4; \
1499 vec_4x32u rgb_dx8; \
1502 vec_8x8u d64_0x07; \
1507 dup_8x8b(d64_0x07, 0x07); \
1508 dup_8x8b(d64_1, 1); \
1509 dup_8x8b(d64_4, 4); \
1510 dup_8x8b(d64_128, 128); \
1512 rgb_dx.low = psx_gpu->uvrg_dx.high; \
1513 rgb_dx.e[2] = psx_gpu->b_block_span.e[1]; \
1514 shl_4x32b(rgb_dx4, rgb_dx, 2); \
1515 shl_4x32b(rgb_dx8, rgb_dx, 3) \
1517 #define setup_blocks_variables_unshaded_textured(target) \
1518 vec_4x32u u_block; \
1519 vec_4x32u v_block; \
1520 vec_2x32u uv_dx = psx_gpu->uvrg_dx.low; \
1523 vec_2x32u uv = psx_gpu->uvrg.low; \
1525 vec_16x8u texture_mask; \
1526 shl_2x32b(uv_dx4, uv_dx, 2); \
1527 shl_2x32b(uv_dx8, uv_dx, 3); \
1528 dup_8x8b(texture_mask.low, psx_gpu->texture_mask_width); \
1529 dup_8x8b(texture_mask.high, psx_gpu->texture_mask_height) \
1532 #define setup_blocks_variables_unshaded_untextured_direct() \
1533 or_8x16b(colors, colors, msb_mask) \
1535 #define setup_blocks_variables_unshaded_untextured_indirect() \
1537 #define setup_blocks_variables_unshaded_untextured(target) \
1538 u32 color = psx_gpu->triangle_color; \
1541 u32 color_r = color & 0xFF; \
1542 u32 color_g = (color >> 8) & 0xFF; \
1543 u32 color_b = (color >> 16) & 0xFF; \
1545 color = (color_r >> 3) | ((color_g >> 3) << 5) | \
1546 ((color_b >> 3) << 10); \
1547 dup_8x16b(colors, color); \
1548 setup_blocks_variables_unshaded_untextured_##target() \
1550 #define setup_blocks_span_initialize_dithered_textured() \
1551 vec_8x16u dither_offsets; \
1552 shl_long_8x8b(dither_offsets, dither_offsets_short, 4) \
1554 #define setup_blocks_span_initialize_dithered_untextured() \
1555 vec_8x8u dither_offsets; \
1556 add_8x8b(dither_offsets, dither_offsets_short, d64_4) \
1558 #define setup_blocks_span_initialize_dithered(texturing) \
1559 u32 dither_row = psx_gpu->dither_table[y & 0x3]; \
1560 u32 dither_shift = (span_edge_data->left_x & 0x3) * 8; \
1561 vec_8x8s dither_offsets_short; \
1564 (dither_row >> dither_shift) | (dither_row << (32 - dither_shift)); \
1565 dup_2x32b(vector_cast(vec_2x32u, dither_offsets_short), dither_row); \
1566 setup_blocks_span_initialize_dithered_##texturing() \
1568 #define setup_blocks_span_initialize_undithered(texturing) \
1571 #define setup_blocks_span_initialize_shaded_textured() \
1573 vec_4x32u block_span; \
1574 u32 offset = span_edge_data->left_x; \
1576 uvrg = *span_uvrg_offset; \
1577 mla_scalar_4x32b(uvrg, uvrg_dx, offset); \
1578 b = *span_b_offset; \
1579 b += b_dx * offset; \
1581 dup_4x32b(u_block, uvrg.e[0]); \
1582 dup_4x32b(v_block, uvrg.e[1]); \
1583 dup_4x32b(r_block, uvrg.e[2]); \
1584 dup_4x32b(g_block, uvrg.e[3]); \
1585 dup_4x32b(b_block, b); \
1587 block_span = psx_gpu->u_block_span; \
1588 add_4x32b(u_block, u_block, block_span); \
1589 block_span = psx_gpu->v_block_span; \
1590 add_4x32b(v_block, v_block, block_span); \
1591 block_span = psx_gpu->r_block_span; \
1592 add_4x32b(r_block, r_block, block_span); \
1593 block_span = psx_gpu->g_block_span; \
1594 add_4x32b(g_block, g_block, block_span); \
1595 block_span = psx_gpu->b_block_span; \
1596 add_4x32b(b_block, b_block, block_span); \
1599 #define setup_blocks_span_initialize_shaded_untextured() \
1601 vec_4x32u block_span; \
1602 u32 offset = span_edge_data->left_x; \
1604 rgb.low = span_uvrg_offset->high; \
1605 rgb.high.e[0] = *span_b_offset; \
1606 mla_scalar_4x32b(rgb, rgb_dx, offset); \
1608 dup_4x32b(r_block, rgb.e[0]); \
1609 dup_4x32b(g_block, rgb.e[1]); \
1610 dup_4x32b(b_block, rgb.e[2]); \
1612 block_span = psx_gpu->r_block_span; \
1613 add_4x32b(r_block, r_block, block_span); \
1614 block_span = psx_gpu->g_block_span; \
1615 add_4x32b(g_block, g_block, block_span); \
1616 block_span = psx_gpu->b_block_span; \
1617 add_4x32b(b_block, b_block, block_span); \
1620 #define setup_blocks_span_initialize_unshaded_textured() \
1622 vec_4x32u block_span; \
1623 u32 offset = span_edge_data->left_x; \
1625 uv = span_uvrg_offset->low; \
1626 mla_scalar_2x32b(uv, uv_dx, offset); \
1628 dup_4x32b(u_block, uv.e[0]); \
1629 dup_4x32b(v_block, uv.e[1]); \
1631 block_span = psx_gpu->u_block_span; \
1632 add_4x32b(u_block, u_block, block_span); \
1633 block_span = psx_gpu->v_block_span; \
1634 add_4x32b(v_block, v_block, block_span); \
1637 #define setup_blocks_span_initialize_unshaded_untextured() \
1640 #define setup_blocks_texture_swizzled() \
1642 vec_8x8u u_saved = u; \
1643 sli_8x8b(u, v, 4); \
1644 sri_8x8b(v, u_saved, 4); \
1647 #define setup_blocks_texture_unswizzled() \
1649 #define setup_blocks_store_shaded_textured(swizzling, dithering, target, \
1652 vec_8x16u u_whole; \
1653 vec_8x16u v_whole; \
1654 vec_8x16u r_whole; \
1655 vec_8x16u g_whole; \
1656 vec_8x16u b_whole; \
1668 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1669 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1670 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1671 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1672 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1674 dup_4x32b(dx4, uvrg_dx4.e[0]); \
1675 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1676 dup_4x32b(dx4, uvrg_dx4.e[1]); \
1677 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1678 dup_4x32b(dx4, uvrg_dx4.e[2]); \
1679 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1680 dup_4x32b(dx4, uvrg_dx4.e[3]); \
1681 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1682 dup_4x32b(dx4, b_dx4); \
1683 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1685 mov_narrow_8x16b(u, u_whole); \
1686 mov_narrow_8x16b(v, v_whole); \
1687 mov_narrow_8x16b(r, r_whole); \
1688 mov_narrow_8x16b(g, g_whole); \
1689 mov_narrow_8x16b(b, b_whole); \
1691 dup_4x32b(dx8, uvrg_dx8.e[0]); \
1692 add_4x32b(u_block, u_block, dx8); \
1693 dup_4x32b(dx8, uvrg_dx8.e[1]); \
1694 add_4x32b(v_block, v_block, dx8); \
1695 dup_4x32b(dx8, uvrg_dx8.e[2]); \
1696 add_4x32b(r_block, r_block, dx8); \
1697 dup_4x32b(dx8, uvrg_dx8.e[3]); \
1698 add_4x32b(g_block, g_block, dx8); \
1699 dup_4x32b(dx8, b_dx8); \
1700 add_4x32b(b_block, b_block, dx8); \
1702 and_8x8b(u, u, texture_mask.low); \
1703 and_8x8b(v, v, texture_mask.high); \
1704 setup_blocks_texture_##swizzling(); \
1706 zip_8x16b(uv, u, v); \
1711 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1712 block->fb_ptr = fb_ptr; \
1715 #define setup_blocks_store_unshaded_textured(swizzling, dithering, target, \
1718 vec_8x16u u_whole; \
1719 vec_8x16u v_whole; \
1728 shr_narrow_4x32b(u_whole.low, u_block, 16); \
1729 shr_narrow_4x32b(v_whole.low, v_block, 16); \
1731 dup_4x32b(dx4, uv_dx4.e[0]); \
1732 add_high_narrow_4x32b(u_whole.high, u_block, dx4); \
1733 dup_4x32b(dx4, uv_dx4.e[1]); \
1734 add_high_narrow_4x32b(v_whole.high, v_block, dx4); \
1736 mov_narrow_8x16b(u, u_whole); \
1737 mov_narrow_8x16b(v, v_whole); \
1739 dup_4x32b(dx8, uv_dx8.e[0]); \
1740 add_4x32b(u_block, u_block, dx8); \
1741 dup_4x32b(dx8, uv_dx8.e[1]); \
1742 add_4x32b(v_block, v_block, dx8); \
1744 and_8x8b(u, u, texture_mask.low); \
1745 and_8x8b(v, v, texture_mask.high); \
1746 setup_blocks_texture_##swizzling(); \
1748 zip_8x16b(uv, u, v); \
1750 block->dither_offsets = vector_cast(vec_8x16u, dither_offsets); \
1751 block->fb_ptr = fb_ptr; \
1754 #define setup_blocks_store_shaded_untextured_dithered() \
1755 addq_8x8b(r, r, dither_offsets); \
1756 addq_8x8b(g, g, dither_offsets); \
1757 addq_8x8b(b, b, dither_offsets); \
1759 subq_8x8b(r, r, d64_4); \
1760 subq_8x8b(g, g, d64_4); \
1761 subq_8x8b(b, b, d64_4) \
1763 #define setup_blocks_store_shaded_untextured_undithered() \
1766 #define setup_blocks_store_untextured_pixels_indirect_full(_pixels) \
1767 block->pixels = _pixels; \
1768 block->fb_ptr = fb_ptr \
1770 #define setup_blocks_store_untextured_pixels_indirect_edge(_pixels) \
1771 block->pixels = _pixels; \
1772 block->fb_ptr = fb_ptr \
1774 #define setup_blocks_store_shaded_untextured_seed_pixels_indirect() \
1775 mul_long_8x8b(pixels, r, d64_1) \
1778 #define setup_blocks_store_untextured_pixels_direct_full(_pixels) \
1779 store_8x16b(_pixels, fb_ptr) \
1781 #define setup_blocks_store_untextured_pixels_direct_edge(_pixels) \
1783 vec_8x16u fb_pixels; \
1784 vec_8x16u draw_mask; \
1785 vec_8x16u test_mask = psx_gpu->test_mask; \
1787 load_8x16b(fb_pixels, fb_ptr); \
1788 dup_8x16b(draw_mask, span_edge_data->right_mask); \
1789 tst_8x16b(draw_mask, draw_mask, test_mask); \
1790 bif_8x16b(fb_pixels, _pixels, draw_mask); \
1791 store_8x16b(fb_pixels, fb_ptr); \
1794 #define setup_blocks_store_shaded_untextured_seed_pixels_direct() \
1795 pixels = msb_mask; \
1796 mla_long_8x8b(pixels, r, d64_1) \
1799 #define setup_blocks_store_shaded_untextured(swizzling, dithering, target, \
1802 vec_8x16u r_whole; \
1803 vec_8x16u g_whole; \
1804 vec_8x16u b_whole; \
1815 shr_narrow_4x32b(r_whole.low, r_block, 16); \
1816 shr_narrow_4x32b(g_whole.low, g_block, 16); \
1817 shr_narrow_4x32b(b_whole.low, b_block, 16); \
1819 dup_4x32b(dx4, rgb_dx4.e[0]); \
1820 add_high_narrow_4x32b(r_whole.high, r_block, dx4); \
1821 dup_4x32b(dx4, rgb_dx4.e[1]); \
1822 add_high_narrow_4x32b(g_whole.high, g_block, dx4); \
1823 dup_4x32b(dx4, rgb_dx4.e[2]); \
1824 add_high_narrow_4x32b(b_whole.high, b_block, dx4); \
1826 mov_narrow_8x16b(r, r_whole); \
1827 mov_narrow_8x16b(g, g_whole); \
1828 mov_narrow_8x16b(b, b_whole); \
1830 dup_4x32b(dx8, rgb_dx8.e[0]); \
1831 add_4x32b(r_block, r_block, dx8); \
1832 dup_4x32b(dx8, rgb_dx8.e[1]); \
1833 add_4x32b(g_block, g_block, dx8); \
1834 dup_4x32b(dx8, rgb_dx8.e[2]); \
1835 add_4x32b(b_block, b_block, dx8); \
1837 setup_blocks_store_shaded_untextured_##dithering(); \
1839 shr_8x8b(r, r, 3); \
1840 bic_8x8b(g, g, d64_0x07); \
1841 bic_8x8b(b, b, d64_0x07); \
1843 setup_blocks_store_shaded_untextured_seed_pixels_##target(); \
1844 mla_long_8x8b(pixels, g, d64_4); \
1845 mla_long_8x8b(pixels, b, d64_128) \
1847 setup_blocks_store_untextured_pixels_##target##_##edge_type(pixels); \
1850 #define setup_blocks_store_unshaded_untextured(swizzling, dithering, target, \
1852 setup_blocks_store_untextured_pixels_##target##_##edge_type(colors) \
1855 #define setup_blocks_store_draw_mask_textured_indirect(_block, bits) \
1856 (_block)->draw_mask_bits = bits \
1858 #define setup_blocks_store_draw_mask_untextured_indirect(_block, bits) \
1860 vec_8x16u bits_mask; \
1861 vec_8x16u test_mask = psx_gpu->test_mask; \
1862 dup_8x16b(bits_mask, bits); \
1863 tst_8x16b(bits_mask, bits_mask, test_mask); \
1864 (_block)->draw_mask = bits_mask; \
1867 #define setup_blocks_store_draw_mask_untextured_direct(_block, bits) \
1870 #define setup_blocks_add_blocks_indirect() \
1871 num_blocks += span_num_blocks; \
1873 if(num_blocks > MAX_BLOCKS) \
1875 psx_gpu->num_blocks = num_blocks - span_num_blocks; \
1876 flush_render_block_buffer(psx_gpu); \
1877 num_blocks = span_num_blocks; \
1878 block = psx_gpu->blocks; \
1881 #define setup_blocks_add_blocks_direct() \
1882 stats_add(texel_blocks_untextured, span_num_blocks); \
1883 span_pixel_blocks += span_num_blocks \
1886 #define setup_blocks_builder(shading, texturing, dithering, sw, target) \
1887 void setup_blocks_##shading##_##texturing##_##dithering##_##sw##_##target( \
1888 psx_gpu_struct *psx_gpu) \
1890 setup_blocks_load_msb_mask_##target(); \
1891 setup_blocks_variables_##shading##_##texturing(target); \
1893 edge_data_struct *span_edge_data = psx_gpu->span_edge_data; \
1894 vec_4x32u *span_uvrg_offset = psx_gpu->span_uvrg_offset; \
1895 u32 *span_b_offset = psx_gpu->span_b_offset; \
1897 block_struct *block = psx_gpu->blocks + psx_gpu->num_blocks; \
1899 u32 num_spans = psx_gpu->num_spans; \
1904 u32 num_blocks = psx_gpu->num_blocks; \
1905 u32 span_num_blocks; \
1909 span_num_blocks = span_edge_data->num_blocks; \
1910 if(span_num_blocks) \
1912 y = span_edge_data->y; \
1913 fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
1915 setup_blocks_span_initialize_##shading##_##texturing(); \
1916 setup_blocks_span_initialize_##dithering(texturing); \
1918 setup_blocks_add_blocks_##target(); \
1920 s32 pixel_span = span_num_blocks * 8; \
1921 pixel_span -= __builtin_popcount(span_edge_data->right_mask & 0xFF); \
1922 span_pixels += pixel_span; \
1924 span_num_blocks--; \
1925 while(span_num_blocks) \
1927 setup_blocks_store_##shading##_##texturing(sw, dithering, target, \
1929 setup_blocks_store_draw_mask_##texturing##_##target(block, 0x00); \
1933 span_num_blocks--; \
1936 setup_blocks_store_##shading##_##texturing(sw, dithering, target, edge); \
1937 setup_blocks_store_draw_mask_##texturing##_##target(block, \
1938 span_edge_data->right_mask); \
1944 zero_block_spans++; \
1949 span_uvrg_offset++; \
1953 psx_gpu->num_blocks = num_blocks; \
1957 //setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1961 setup_blocks_builder(shaded, textured, dithered, swizzled, indirect);
1962 setup_blocks_builder(shaded, textured, dithered, unswizzled, indirect);
1964 setup_blocks_builder(unshaded, textured, dithered, unswizzled, indirect);
1965 setup_blocks_builder(unshaded, textured, dithered, swizzled, indirect);
1967 setup_blocks_builder(shaded, untextured, undithered, unswizzled, indirect);
1968 setup_blocks_builder(shaded, untextured, dithered, unswizzled, indirect);
1969 setup_blocks_builder(shaded, untextured, undithered, unswizzled, direct);
1970 setup_blocks_builder(shaded, untextured, dithered, unswizzled, direct);
1972 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, indirect);
1973 setup_blocks_builder(unshaded, untextured, undithered, unswizzled, direct);
1975 void texture_blocks_untextured(psx_gpu_struct *psx_gpu)
1977 if(psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE)
1978 stats_add(texel_blocks_untextured, psx_gpu->num_blocks);
1981 void texture_blocks_4bpp(psx_gpu_struct *psx_gpu)
1983 block_struct *block = psx_gpu->blocks;
1984 u32 num_blocks = psx_gpu->num_blocks;
1985 stats_add(texel_blocks_4bpp, num_blocks);
1987 vec_8x8u texels_low;
1988 vec_8x8u texels_high;
1995 vec_16x8u clut_high;
1997 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
1998 u16 *clut_ptr = psx_gpu->clut_ptr;
2000 // Can be done with one deinterleaving load on NEON
2001 load_8x16b(clut_a, clut_ptr);
2002 load_8x16b(clut_b, clut_ptr + 8);
2003 unzip_16x8b(clut_low, clut_high, clut_a, clut_b);
2005 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask)
2006 update_texture_4bpp_cache(psx_gpu);
2010 texels.e[0] = texture_ptr_8bpp[block->uv.e[0]];
2011 texels.e[1] = texture_ptr_8bpp[block->uv.e[1]];
2012 texels.e[2] = texture_ptr_8bpp[block->uv.e[2]];
2013 texels.e[3] = texture_ptr_8bpp[block->uv.e[3]];
2014 texels.e[4] = texture_ptr_8bpp[block->uv.e[4]];
2015 texels.e[5] = texture_ptr_8bpp[block->uv.e[5]];
2016 texels.e[6] = texture_ptr_8bpp[block->uv.e[6]];
2017 texels.e[7] = texture_ptr_8bpp[block->uv.e[7]];
2019 tbl_16(texels_low, texels, clut_low);
2020 tbl_16(texels_high, texels, clut_high);
2022 // Can be done with an interleaving store on NEON
2023 zip_8x16b(pixels, texels_low, texels_high);
2025 block->texels = pixels;
2032 void texture_blocks_8bpp(psx_gpu_struct *psx_gpu)
2034 block_struct *block = psx_gpu->blocks;
2035 u32 num_blocks = psx_gpu->num_blocks;
2037 stats_add(texel_blocks_8bpp, num_blocks);
2039 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask)
2040 update_texture_8bpp_cache(psx_gpu);
2043 u8 *texture_ptr_8bpp = psx_gpu->texture_page_ptr;
2051 for(i = 0; i < 8; i++)
2053 offset = block->uv.e[i];
2055 texel = texture_ptr_8bpp[offset];
2056 texels.e[i] = psx_gpu->clut_ptr[texel];
2059 block->texels = texels;
2066 void texture_blocks_16bpp(psx_gpu_struct *psx_gpu)
2068 block_struct *block = psx_gpu->blocks;
2069 u32 num_blocks = psx_gpu->num_blocks;
2071 stats_add(texel_blocks_16bpp, num_blocks);
2075 u16 *texture_ptr_16bpp = psx_gpu->texture_page_ptr;
2081 for(i = 0; i < 8; i++)
2083 offset = block->uv.e[i];
2084 offset += ((offset & 0xFF00) * 3);
2086 texels.e[i] = texture_ptr_16bpp[offset];
2089 block->texels = texels;
2099 #define shade_blocks_load_msb_mask_indirect() \
2101 #define shade_blocks_load_msb_mask_direct() \
2102 vec_8x16u msb_mask; \
2103 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2105 #define shade_blocks_store_indirect(_draw_mask, _pixels) \
2106 block->draw_mask = _draw_mask; \
2107 block->pixels = _pixels \
2109 #define shade_blocks_store_direct(_draw_mask, _pixels) \
2111 vec_8x16u fb_pixels; \
2112 or_8x16b(_pixels, _pixels, msb_mask); \
2113 load_8x16b(fb_pixels, block->fb_ptr); \
2114 bif_8x16b(fb_pixels, _pixels, _draw_mask); \
2115 store_8x16b(fb_pixels, block->fb_ptr); \
2119 #define shade_blocks_textured_false_modulated_check_dithered(target) \
2120 if(psx_gpu->triangle_color == 0x808080) \
2122 false_modulated_blocks += num_blocks; \
2125 #define shade_blocks_textured_false_modulated_check_undithered(target) \
2126 if(psx_gpu->triangle_color == 0x808080) \
2129 shade_blocks_textured_unmodulated_##target(psx_gpu); \
2130 false_modulated_blocks += num_blocks; \
2135 #define shade_blocks_textured_modulated_shaded_primitive_load(dithering, \
2138 #define shade_blocks_textured_modulated_unshaded_primitive_load(dithering, \
2141 u32 color = psx_gpu->triangle_color; \
2142 dup_8x8b(colors_r, color); \
2143 dup_8x8b(colors_g, color >> 8); \
2144 dup_8x8b(colors_b, color >> 16); \
2145 shade_blocks_textured_false_modulated_check_##dithering(target); \
2148 #define shade_blocks_textured_modulated_shaded_block_load() \
2149 colors_r = block->r; \
2150 colors_g = block->g; \
2151 colors_b = block->b \
2153 #define shade_blocks_textured_modulated_unshaded_block_load() \
2155 #define shade_blocks_textured_modulate_dithered(component) \
2156 pixels_##component = block->dither_offsets; \
2157 mla_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2159 #define shade_blocks_textured_modulate_undithered(component) \
2160 mul_long_8x8b(pixels_##component, texels_##component, colors_##component) \
2162 #define shade_blocks_textured_modulated_builder(shading, dithering, target) \
2163 void shade_blocks_##shading##_textured_modulated_##dithering##_##target( \
2164 psx_gpu_struct *psx_gpu) \
2166 block_struct *block = psx_gpu->blocks; \
2167 u32 num_blocks = psx_gpu->num_blocks; \
2170 vec_8x8u texels_r; \
2171 vec_8x8u texels_g; \
2172 vec_8x8u texels_b; \
2174 vec_8x8u colors_r; \
2175 vec_8x8u colors_g; \
2176 vec_8x8u colors_b; \
2178 vec_8x8u pixels_r_low; \
2179 vec_8x8u pixels_g_low; \
2180 vec_8x8u pixels_b_low; \
2183 vec_8x16u pixels_r; \
2184 vec_8x16u pixels_g; \
2185 vec_8x16u pixels_b; \
2187 vec_8x16u draw_mask; \
2188 vec_8x16u zero_mask; \
2190 vec_8x8u d64_0x07; \
2191 vec_8x8u d64_0x1F; \
2196 vec_8x16u d128_0x8000; \
2198 vec_8x16u test_mask = psx_gpu->test_mask; \
2199 u32 draw_mask_bits; \
2200 shade_blocks_load_msb_mask_##target(); \
2202 dup_8x8b(d64_0x07, 0x07); \
2203 dup_8x8b(d64_0x1F, 0x1F); \
2204 dup_8x8b(d64_1, 1); \
2205 dup_8x8b(d64_4, 4); \
2206 dup_8x8b(d64_128, 128); \
2208 dup_8x16b(d128_0x8000, 0x8000); \
2210 shade_blocks_textured_modulated_##shading##_primitive_load(dithering, \
2215 draw_mask_bits = block->draw_mask_bits; \
2216 dup_8x16b(draw_mask, draw_mask_bits); \
2217 tst_8x16b(draw_mask, draw_mask, test_mask); \
2219 shade_blocks_textured_modulated_##shading##_block_load(); \
2221 texels = block->texels; \
2223 mov_narrow_8x16b(texels_r, texels); \
2224 shr_narrow_8x16b(texels_g, texels, 5); \
2225 shr_narrow_8x16b(texels_b, texels, 7); \
2227 and_8x8b(texels_r, texels_r, d64_0x1F); \
2228 and_8x8b(texels_g, texels_g, d64_0x1F); \
2229 shr_8x8b(texels_b, texels_b, 3); \
2231 shade_blocks_textured_modulate_##dithering(r); \
2232 shade_blocks_textured_modulate_##dithering(g); \
2233 shade_blocks_textured_modulate_##dithering(b); \
2235 cmpeqz_8x16b(zero_mask, texels); \
2236 and_8x16b(pixels, texels, d128_0x8000); \
2238 shrq_narrow_signed_8x16b(pixels_r_low, pixels_r, 4); \
2239 shrq_narrow_signed_8x16b(pixels_g_low, pixels_g, 4); \
2240 shrq_narrow_signed_8x16b(pixels_b_low, pixels_b, 4); \
2242 or_8x16b(zero_mask, draw_mask, zero_mask); \
2244 shr_8x8b(pixels_r_low, pixels_r_low, 3); \
2245 bic_8x8b(pixels_g_low, pixels_g_low, d64_0x07); \
2246 bic_8x8b(pixels_b_low, pixels_b_low, d64_0x07); \
2248 mla_long_8x8b(pixels, pixels_r_low, d64_1); \
2249 mla_long_8x8b(pixels, pixels_g_low, d64_4); \
2250 mla_long_8x8b(pixels, pixels_b_low, d64_128); \
2252 shade_blocks_store_##target(zero_mask, pixels); \
2261 shade_blocks_textured_modulated_builder(shaded, dithered, direct);
2262 shade_blocks_textured_modulated_builder(shaded, undithered, direct);
2263 shade_blocks_textured_modulated_builder(unshaded, dithered, direct);
2264 shade_blocks_textured_modulated_builder(unshaded, undithered, direct);
2266 shade_blocks_textured_modulated_builder(shaded, dithered, indirect);
2267 shade_blocks_textured_modulated_builder(shaded, undithered, indirect);
2268 shade_blocks_textured_modulated_builder(unshaded, dithered, indirect);
2269 shade_blocks_textured_modulated_builder(unshaded, undithered, indirect);
2274 #define shade_blocks_textured_unmodulated_builder(target) \
2275 void shade_blocks_textured_unmodulated_##target(psx_gpu_struct *psx_gpu) \
2277 block_struct *block = psx_gpu->blocks; \
2278 u32 num_blocks = psx_gpu->num_blocks; \
2279 vec_8x16u draw_mask; \
2280 vec_8x16u test_mask = psx_gpu->test_mask; \
2281 u32 draw_mask_bits; \
2284 shade_blocks_load_msb_mask_##target(); \
2288 vec_8x16u zero_mask; \
2290 draw_mask_bits = block->draw_mask_bits; \
2291 dup_8x16b(draw_mask, draw_mask_bits); \
2292 tst_8x16b(draw_mask, draw_mask, test_mask); \
2294 pixels = block->texels; \
2296 cmpeqz_8x16b(zero_mask, pixels); \
2297 or_8x16b(zero_mask, draw_mask, zero_mask); \
2299 shade_blocks_store_##target(zero_mask, pixels); \
2306 #define shade_blocks_textured_unmodulated_dithered_builder(target) \
2307 void shade_blocks_textured_unmodulated_dithered_##target(psx_gpu_struct \
2310 block_struct *block = psx_gpu->blocks; \
2311 u32 num_blocks = psx_gpu->num_blocks; \
2312 vec_8x16u draw_mask; \
2313 vec_8x16u test_mask = psx_gpu->test_mask; \
2314 u32 draw_mask_bits; \
2317 shade_blocks_load_msb_mask_##target(); \
2321 vec_8x16u zero_mask; \
2323 draw_mask_bits = block->draw_mask_bits; \
2324 dup_8x16b(draw_mask, draw_mask_bits); \
2325 tst_8x16b(draw_mask, draw_mask, test_mask); \
2327 pixels = block->texels; \
2329 cmpeqz_8x16b(zero_mask, pixels); \
2330 or_8x16b(zero_mask, draw_mask, zero_mask); \
2332 shade_blocks_store_##target(zero_mask, pixels); \
2341 shade_blocks_textured_unmodulated_builder(indirect)
2342 shade_blocks_textured_unmodulated_builder(direct)
2344 void shade_blocks_unshaded_untextured_indirect(psx_gpu_struct *psx_gpu)
2348 void shade_blocks_unshaded_untextured_direct(psx_gpu_struct *psx_gpu)
2350 block_struct *block = psx_gpu->blocks;
2351 u32 num_blocks = psx_gpu->num_blocks;
2353 vec_8x16u pixels = block->pixels;
2354 shade_blocks_load_msb_mask_direct();
2358 shade_blocks_store_direct(block->draw_mask, pixels);
2367 void shade_blocks_shaded_untextured(psx_gpu_struct *psx_gpu)
2372 #define blend_blocks_mask_evaluate_on() \
2373 vec_8x16u mask_pixels; \
2374 cmpltz_8x16b(mask_pixels, framebuffer_pixels); \
2375 or_8x16b(draw_mask, draw_mask, mask_pixels) \
2377 #define blend_blocks_mask_evaluate_off() \
2379 #define blend_blocks_average() \
2381 vec_8x16u pixels_no_msb; \
2382 vec_8x16u fb_pixels_no_msb; \
2384 vec_8x16u d128_0x0421; \
2385 vec_8x16u d128_0x8000; \
2387 dup_8x16b(d128_0x0421, 0x0421); \
2388 dup_8x16b(d128_0x8000, 0x8000); \
2390 eor_8x16b(blend_pixels, pixels, framebuffer_pixels); \
2391 bic_8x16b(pixels_no_msb, pixels, d128_0x8000); \
2392 and_8x16b(blend_pixels, blend_pixels, d128_0x0421); \
2393 sub_8x16b(blend_pixels, pixels_no_msb, blend_pixels); \
2394 bic_8x16b(fb_pixels_no_msb, framebuffer_pixels, d128_0x8000); \
2395 average_8x16b(blend_pixels, fb_pixels_no_msb, blend_pixels); \
2398 #define blend_blocks_add() \
2400 vec_8x16u pixels_rb, pixels_g; \
2401 vec_8x16u fb_rb, fb_g; \
2403 vec_8x16u d128_0x7C1F; \
2404 vec_8x16u d128_0x03E0; \
2406 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2407 dup_8x16b(d128_0x03E0, 0x03E0); \
2409 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2410 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2412 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2413 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2415 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2416 add_8x16b(fb_g, fb_g, pixels_g); \
2418 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2419 vector_cast(vec_16x8u, d128_0x7C1F)); \
2420 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2422 or_8x16b(blend_pixels, fb_rb, fb_g); \
2425 #define blend_blocks_subtract() \
2427 vec_8x16u pixels_rb, pixels_g; \
2428 vec_8x16u fb_rb, fb_g; \
2430 vec_8x16u d128_0x7C1F; \
2431 vec_8x16u d128_0x03E0; \
2433 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2434 dup_8x16b(d128_0x03E0, 0x03E0); \
2436 and_8x16b(pixels_rb, pixels, d128_0x7C1F); \
2437 and_8x16b(pixels_g, pixels, d128_0x03E0); \
2439 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2440 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2442 subs_16x8b(vector_cast(vec_16x8u, fb_rb), \
2443 vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, pixels_rb)); \
2444 subs_8x16b(fb_g, fb_g, pixels_g); \
2446 or_8x16b(blend_pixels, fb_rb, fb_g); \
2449 #define blend_blocks_add_fourth() \
2451 vec_8x16u pixels_rb, pixels_g; \
2452 vec_8x16u pixels_fourth; \
2453 vec_8x16u fb_rb, fb_g; \
2455 vec_8x16u d128_0x7C1F; \
2456 vec_8x16u d128_0x1C07; \
2457 vec_8x16u d128_0x03E0; \
2458 vec_8x16u d128_0x00E0; \
2460 dup_8x16b(d128_0x7C1F, 0x7C1F); \
2461 dup_8x16b(d128_0x1C07, 0x1C07); \
2462 dup_8x16b(d128_0x03E0, 0x03E0); \
2463 dup_8x16b(d128_0x00E0, 0x00E0); \
2465 shr_8x16b(pixels_fourth, vector_cast(vec_8x16s, pixels), 2); \
2467 and_8x16b(fb_rb, framebuffer_pixels, d128_0x7C1F); \
2468 and_8x16b(fb_g, framebuffer_pixels, d128_0x03E0); \
2470 and_8x16b(pixels_rb, pixels_fourth, d128_0x1C07); \
2471 and_8x16b(pixels_g, pixels_fourth, d128_0x00E0); \
2473 add_8x16b(fb_rb, fb_rb, pixels_rb); \
2474 add_8x16b(fb_g, fb_g, pixels_g); \
2476 min_16x8b(vector_cast(vec_16x8u, fb_rb), vector_cast(vec_16x8u, fb_rb), \
2477 vector_cast(vec_16x8u, d128_0x7C1F)); \
2478 min_8x16b(fb_g, fb_g, d128_0x03E0); \
2480 or_8x16b(blend_pixels, fb_rb, fb_g); \
2483 #define blend_blocks_blended_combine_textured() \
2485 vec_8x16u blend_mask; \
2486 cmpltz_8x16b(blend_mask, pixels); \
2488 or_immediate_8x16b(blend_pixels, blend_pixels, 0x8000); \
2489 bif_8x16b(blend_pixels, pixels, blend_mask); \
2492 #define blend_blocks_blended_combine_untextured() \
2495 #define blend_blocks_body_blend(blend_mode, texturing) \
2497 blend_blocks_##blend_mode(); \
2498 blend_blocks_blended_combine_##texturing(); \
2501 #define blend_blocks_body_average(texturing) \
2502 blend_blocks_body_blend(average, texturing) \
2504 #define blend_blocks_body_add(texturing) \
2505 blend_blocks_body_blend(add, texturing) \
2507 #define blend_blocks_body_subtract(texturing) \
2508 blend_blocks_body_blend(subtract, texturing) \
2510 #define blend_blocks_body_add_fourth(texturing) \
2511 blend_blocks_body_blend(add_fourth, texturing) \
2513 #define blend_blocks_body_unblended(texturing) \
2514 blend_pixels = pixels \
2517 #define blend_blocks_builder(texturing, blend_mode, mask_evaluate) \
2519 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate(psx_gpu_struct \
2522 block_struct *block = psx_gpu->blocks; \
2523 u32 num_blocks = psx_gpu->num_blocks; \
2524 vec_8x16u draw_mask; \
2526 vec_8x16u blend_pixels; \
2527 vec_8x16u framebuffer_pixels; \
2528 vec_8x16u msb_mask; \
2532 dup_8x16b(msb_mask, psx_gpu->mask_msb); \
2536 pixels = block->pixels; \
2537 draw_mask = block->draw_mask; \
2538 fb_ptr = block->fb_ptr; \
2540 load_8x16b(framebuffer_pixels, fb_ptr); \
2542 blend_blocks_mask_evaluate_##mask_evaluate(); \
2543 blend_blocks_body_##blend_mode(texturing); \
2545 or_8x16b(blend_pixels, blend_pixels, msb_mask); \
2546 bif_8x16b(framebuffer_pixels, blend_pixels, draw_mask); \
2547 store_8x16b(framebuffer_pixels, fb_ptr); \
2557 void blend_blocks_textured_unblended_off(psx_gpu_struct *psx_gpu)
2561 blend_blocks_builder(textured, average, off);
2562 blend_blocks_builder(textured, average, on);
2563 blend_blocks_builder(textured, add, off);
2564 blend_blocks_builder(textured, add, on);
2565 blend_blocks_builder(textured, subtract, off);
2566 blend_blocks_builder(textured, subtract, on);
2567 blend_blocks_builder(textured, add_fourth, off);
2568 blend_blocks_builder(textured, add_fourth, on);
2570 blend_blocks_builder(untextured, average, off);
2571 blend_blocks_builder(untextured, average, on);
2572 blend_blocks_builder(untextured, add, off);
2573 blend_blocks_builder(untextured, add, on);
2574 blend_blocks_builder(untextured, subtract, off);
2575 blend_blocks_builder(untextured, subtract, on);
2576 blend_blocks_builder(untextured, add_fourth, off);
2577 blend_blocks_builder(untextured, add_fourth, on);
2579 blend_blocks_builder(textured, unblended, on);
2584 #define vertex_swap(_a, _b) \
2586 vertex_struct *temp_vertex = _a; \
2589 triangle_winding ^= 1; \
2593 // Setup blocks parametric-variables:
2594 // SHADE TEXTURE_MAP SWIZZLING
2601 // 8 inputs, 6 combinations
2603 #define setup_blocks_switch_untextured_unshaded(dithering, target) \
2604 setup_blocks_unshaded_untextured_undithered_unswizzled_##target \
2606 #define setup_blocks_switch_untextured_shaded(dithering, target) \
2607 setup_blocks_shaded_untextured_##dithering##_unswizzled_##target \
2609 #define setup_blocks_switch_untextured(shading, texture_mode, dithering, \
2611 setup_blocks_switch_untextured_##shading(dithering, target) \
2613 #define setup_blocks_switch_texture_mode_4bpp(shading) \
2614 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2616 #define setup_blocks_switch_texture_mode_8bpp(shading) \
2617 setup_blocks_##shading##_textured_dithered_swizzled_indirect \
2619 #define setup_blocks_switch_texture_mode_16bpp(shading) \
2620 setup_blocks_##shading##_textured_dithered_unswizzled_indirect \
2622 #define setup_blocks_switch_textured(shading, texture_mode, dithering, target) \
2623 setup_blocks_switch_texture_mode_##texture_mode(shading) \
2625 #define setup_blocks_switch_blended(shading, texturing, texture_mode, \
2626 dithering, mask_evaluate) \
2627 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2629 #define setup_blocks_switch_unblended_on(shading, texturing, texture_mode, \
2631 setup_blocks_switch_##texturing(shading, texture_mode, dithering, indirect) \
2633 #define setup_blocks_switch_unblended_off(shading, texturing, texture_mode, \
2635 setup_blocks_switch_##texturing(shading, texture_mode, dithering, direct) \
2637 #define setup_blocks_switch_unblended(shading, texturing, texture_mode, \
2638 dithering, mask_evaluate) \
2639 setup_blocks_switch_unblended_##mask_evaluate(shading, texturing, \
2640 texture_mode, dithering) \
2642 #define setup_blocks_switch(shading, texturing, texture_mode, dithering, \
2643 blending, mask_evaluate) \
2644 setup_blocks_switch_##blending(shading, texturing, texture_mode, \
2645 dithering, mask_evaluate) \
2650 #define texture_blocks_switch_untextured(texture_mode) \
2651 texture_blocks_untextured \
2653 #define texture_blocks_switch_textured(texture_mode) \
2654 texture_blocks_##texture_mode \
2656 #define texture_blocks_switch(texturing, texture_mode) \
2657 texture_blocks_switch_##texturing(texture_mode) \
2660 // Shade blocks parametric-variables:
2661 // SHADE TEXTURE_MAP MODULATE_TEXELS dither_mode
2670 // 16 inputs, 8 combinations
2672 #define shade_blocks_switch_unshaded_untextured(modulation, dithering, target) \
2673 shade_blocks_unshaded_untextured_##target \
2675 #define shade_blocks_switch_unshaded_textured_unmodulated(dithering, target) \
2676 shade_blocks_textured_unmodulated_##target \
2678 #define shade_blocks_switch_unshaded_textured_modulated(dithering, target) \
2679 shade_blocks_unshaded_textured_modulated_##dithering##_##target \
2681 #define shade_blocks_switch_unshaded_textured(modulation, dithering, target) \
2682 shade_blocks_switch_unshaded_textured_##modulation(dithering, target) \
2684 #define shade_blocks_switch_unshaded(texturing, modulation, dithering, target) \
2685 shade_blocks_switch_unshaded_##texturing(modulation, dithering, target) \
2687 #define shade_blocks_switch_shaded_untextured(modulation, dithering, target) \
2688 shade_blocks_shaded_untextured \
2690 #define shade_blocks_switch_shaded_textured_unmodulated(dithering, target) \
2691 shade_blocks_textured_unmodulated_##target \
2693 #define shade_blocks_switch_shaded_textured_modulated(dithering, target) \
2694 shade_blocks_shaded_textured_modulated_##dithering##_##target \
2696 #define shade_blocks_switch_shaded_textured(modulation, dithering, target) \
2697 shade_blocks_switch_shaded_textured_##modulation(dithering, target) \
2699 #define shade_blocks_switch_shaded(texturing, modulation, dithering, target) \
2700 shade_blocks_switch_shaded_##texturing(modulation, dithering, target) \
2702 #define shade_blocks_switch_mask_off(shading, texturing, modulation, \
2704 shade_blocks_switch_##shading(texturing, modulation, dithering, direct) \
2706 #define shade_blocks_switch_mask_on(shading, texturing, modulation, \
2708 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2710 #define shade_blocks_switch_blended(shading, texturing, modulation, dithering, \
2712 shade_blocks_switch_##shading(texturing, modulation, dithering, indirect) \
2714 #define shade_blocks_switch_unblended(shading, texturing, modulation, \
2715 dithering, mask_evaluate) \
2716 shade_blocks_switch_mask_##mask_evaluate(shading, texturing, modulation, \
2719 #define shade_blocks_switch(shading, texturing, modulation, dithering, \
2720 blending, mask_evaluate) \
2721 shade_blocks_switch_##blending(shading, texturing, modulation, dithering, \
2725 // Blend blocks parametric-variables:
2726 // TEXTURE_MAP BLEND BM_A BM_B mask_evaluate
2745 // 32 inputs, 18 combinations
2747 #define blend_blocks_switch_unblended(texturing, blend_mode, mask_evaluate) \
2748 blend_blocks_textured_unblended_##mask_evaluate \
2750 #define blend_blocks_switch_blended(texturing, blend_mode, mask_evaluate) \
2751 blend_blocks_##texturing##_##blend_mode##_##mask_evaluate \
2753 #define blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2754 blend_blocks_switch_##blending(texturing, blend_mode, mask_evaluate) \
2757 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2758 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2760 setup_blocks_switch(shading, texturing, texture_mode, dithering, blending, \
2762 texture_blocks_switch(texturing, texture_mode), \
2763 shade_blocks_switch(shading, texturing, modulation, dithering, blending, \
2765 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
2768 #define render_blocks_switch_block_blending(texture_mode, blend_mode, \
2769 mask_evaluate, shading, dithering, texturing, blending) \
2770 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2771 mask_evaluate, shading, dithering, texturing, blending, modulated), \
2772 render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2773 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
2775 #define render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2776 mask_evaluate, shading, dithering, texturing) \
2777 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2778 mask_evaluate, shading, dithering, texturing, unblended), \
2779 render_blocks_switch_block_blending(texture_mode, blend_mode, \
2780 mask_evaluate, shading, dithering, texturing, blended) \
2782 #define render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2783 mask_evaluate, shading, dithering) \
2784 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2785 mask_evaluate, shading, dithering, untextured), \
2786 render_blocks_switch_block_texturing(texture_mode, blend_mode, \
2787 mask_evaluate, shading, dithering, textured) \
2789 #define render_blocks_switch_block_shading(texture_mode, blend_mode, \
2790 mask_evaluate, shading) \
2791 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2792 mask_evaluate, shading, undithered), \
2793 render_blocks_switch_block_dithering(texture_mode, blend_mode, \
2794 mask_evaluate, shading, dithered) \
2796 #define render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
2798 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2800 render_blocks_switch_block_shading(texture_mode, blend_mode, mask_evaluate, \
2803 #define render_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
2804 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, off), \
2805 render_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, on) \
2807 #define render_blocks_switch_block_texture_mode(texture_mode) \
2808 render_blocks_switch_block_blend_mode(texture_mode, average), \
2809 render_blocks_switch_block_blend_mode(texture_mode, add), \
2810 render_blocks_switch_block_blend_mode(texture_mode, subtract), \
2811 render_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
2813 #define render_blocks_switch_block() \
2814 render_blocks_switch_block_texture_mode(4bpp), \
2815 render_blocks_switch_block_texture_mode(8bpp), \
2816 render_blocks_switch_block_texture_mode(16bpp), \
2817 render_blocks_switch_block_texture_mode(16bpp) \
2820 render_block_handler_struct render_triangle_block_handlers[] =
2822 render_blocks_switch_block()
2825 #undef render_blocks_switch_block_modulation
2827 #define render_blocks_switch_block_modulation(texture_mode, blend_mode, \
2828 mask_evaluate, shading, dithering, texturing, blending, modulation) \
2830 "texture mode: " #texture_mode "\n" \
2831 "blend mode: " #blend_mode "\n" \
2832 "mask evaluation: " #mask_evaluate "\n" \
2839 char *render_block_flag_strings[] =
2841 render_blocks_switch_block()
2845 #define triangle_y_direction_up 1
2846 #define triangle_y_direction_flat 2
2847 #define triangle_y_direction_down 0
2849 #define triangle_winding_positive 0
2850 #define triangle_winding_negative 1
2852 #define triangle_set_direction(direction_variable, value) \
2853 u32 direction_variable = (u32)(value) >> 31; \
2855 direction_variable = 2 \
2857 #define triangle_case(direction_a, direction_b, direction_c, winding) \
2858 case (triangle_y_direction_##direction_a | \
2859 (triangle_y_direction_##direction_b << 2) | \
2860 (triangle_y_direction_##direction_c << 4) | \
2861 (triangle_winding_##winding << 6)) \
2863 static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
2864 vertex_struct *vertexes_out[3])
2866 s32 y_top, y_bottom;
2868 u32 triangle_winding = 0;
2870 vertex_struct *a = &(vertexes[0]);
2871 vertex_struct *b = &(vertexes[1]);
2872 vertex_struct *c = &(vertexes[2]);
2874 triangle_area = triangle_signed_area_x2(a->x, a->y, b->x, b->y, c->x, c->y);
2880 if(triangle_area == 0)
2902 if((y_bottom - y_top) >= 512)
2910 if(triangle_area < 0)
2912 triangle_area = -triangle_area;
2913 triangle_winding ^= 1;
2928 if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
2936 if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
2945 psx_gpu->triangle_area = triangle_area;
2946 psx_gpu->triangle_winding = triangle_winding;
2948 vertexes_out[0] = a;
2949 vertexes_out[1] = b;
2950 vertexes_out[2] = c;
2955 static void render_triangle_p(psx_gpu_struct *psx_gpu,
2956 vertex_struct *vertex_ptrs[3], u32 flags)
2958 psx_gpu->num_spans = 0;
2960 vertex_struct *a = vertex_ptrs[0];
2961 vertex_struct *b = vertex_ptrs[1];
2962 vertex_struct *c = vertex_ptrs[2];
2964 s32 y_delta_a = b->y - a->y;
2965 s32 y_delta_b = c->y - b->y;
2966 s32 y_delta_c = c->y - a->y;
2968 triangle_set_direction(y_direction_a, y_delta_a);
2969 triangle_set_direction(y_direction_b, y_delta_b);
2970 triangle_set_direction(y_direction_c, y_delta_c);
2972 compute_all_gradients(psx_gpu, a, b, c);
2974 switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
2975 (psx_gpu->triangle_winding << 6))
2977 triangle_case(up, up, up, negative):
2978 triangle_case(up, up, flat, negative):
2979 triangle_case(up, up, down, negative):
2980 setup_spans_up_right(psx_gpu, a, b, c);
2983 triangle_case(flat, up, up, negative):
2984 triangle_case(flat, up, flat, negative):
2985 triangle_case(flat, up, down, negative):
2986 setup_spans_up_a(psx_gpu, a, b, c);
2989 triangle_case(down, up, up, negative):
2990 setup_spans_up_down(psx_gpu, a, c, b);
2993 triangle_case(down, up, flat, negative):
2994 setup_spans_down_a(psx_gpu, a, c, b);
2997 triangle_case(down, up, down, negative):
2998 setup_spans_down_right(psx_gpu, a, c, b);
3001 triangle_case(down, flat, up, negative):
3002 triangle_case(down, flat, flat, negative):
3003 triangle_case(down, flat, down, negative):
3004 setup_spans_down_b(psx_gpu, a, b, c);
3007 triangle_case(down, down, up, negative):
3008 triangle_case(down, down, flat, negative):
3009 triangle_case(down, down, down, negative):
3010 setup_spans_down_left(psx_gpu, a, b, c);
3013 triangle_case(up, up, up, positive):
3014 triangle_case(up, up, flat, positive):
3015 triangle_case(up, up, down, positive):
3016 setup_spans_up_left(psx_gpu, a, b, c);
3019 triangle_case(up, flat, up, positive):
3020 triangle_case(up, flat, flat, positive):
3021 triangle_case(up, flat, down, positive):
3022 setup_spans_up_b(psx_gpu, a, b, c);
3025 triangle_case(up, down, up, positive):
3026 setup_spans_up_right(psx_gpu, a, c, b);
3029 triangle_case(up, down, flat, positive):
3030 setup_spans_up_a(psx_gpu, a, c, b);
3033 triangle_case(up, down, down, positive):
3034 setup_spans_up_down(psx_gpu, a, b, c);
3037 triangle_case(flat, down, up, positive):
3038 triangle_case(flat, down, flat, positive):
3039 triangle_case(flat, down, down, positive):
3040 setup_spans_down_a(psx_gpu, a, b, c);
3043 triangle_case(down, down, up, positive):
3044 triangle_case(down, down, flat, positive):
3045 triangle_case(down, down, down, positive):
3046 setup_spans_down_right(psx_gpu, a, b, c);
3051 spans += psx_gpu->num_spans;
3054 if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
3058 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
3060 for(i = 0; i < psx_gpu->num_spans; i++)
3062 if((psx_gpu->span_edge_data[i].y & 1) == 0)
3063 psx_gpu->span_edge_data[i].num_blocks = 0;
3068 for(i = 0; i < psx_gpu->num_spans; i++)
3070 if(psx_gpu->span_edge_data[i].y & 1)
3071 psx_gpu->span_edge_data[i].num_blocks = 0;
3075 assert(psx_gpu->span_edge_data[0].y < 1024u);
3077 u32 render_state = flags &
3078 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
3079 RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
3080 render_state |= psx_gpu->render_state_base;
3082 if((psx_gpu->render_state != render_state) ||
3083 (psx_gpu->primitive_type != PRIMITIVE_TYPE_TRIANGLE))
3085 psx_gpu->render_state = render_state;
3086 flush_render_block_buffer(psx_gpu);
3092 psx_gpu->primitive_type = PRIMITIVE_TYPE_TRIANGLE;
3094 psx_gpu->render_block_handler =
3095 &(render_triangle_block_handlers[render_state]);
3096 ((setup_blocks_function_type *)psx_gpu->render_block_handler->setup_blocks)
3100 void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
3103 vertex_struct *vertex_ptrs[3];
3104 if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
3105 render_triangle_p(psx_gpu, vertex_ptrs, flags);
3108 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3110 void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu)
3112 block_struct *block = psx_gpu->blocks;
3113 u32 num_blocks = psx_gpu->num_blocks;
3116 vec_8x8u texel_indexes;
3118 u16 *clut_ptr = psx_gpu->clut_ptr;
3123 texel_indexes = block->r;
3125 for(i = 0; i < 8; i++)
3127 texels.e[i] = clut_ptr[texel_indexes.e[i]];
3130 block->texels = texels;
3140 #define setup_sprite_tiled_initialize_4bpp_clut() \
3141 u16 *clut_ptr = psx_gpu->clut_ptr; \
3142 vec_8x16u clut_a, clut_b; \
3143 vec_16x8u clut_low, clut_high; \
3145 load_8x16b(clut_a, clut_ptr); \
3146 load_8x16b(clut_b, clut_ptr + 8); \
3147 unzip_16x8b(clut_low, clut_high, clut_a, clut_b) \
3149 #define setup_sprite_tiled_initialize_4bpp() \
3150 setup_sprite_tiled_initialize_4bpp_clut(); \
3152 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_4bpp_mask) \
3153 update_texture_4bpp_cache(psx_gpu) \
3155 #define setup_sprite_tiled_initialize_8bpp() \
3156 if(psx_gpu->current_texture_mask & psx_gpu->dirty_textures_8bpp_mask) \
3157 update_texture_8bpp_cache(psx_gpu) \
3160 #define setup_sprite_tile_fetch_texel_block_8bpp(offset) \
3161 texture_block_ptr = (u8 *)psx_gpu->texture_page_ptr + \
3162 ((texture_offset + offset) & texture_mask); \
3164 load_64b(texels, texture_block_ptr) \
3167 #define setup_sprite_tile_add_blocks(tile_num_blocks) \
3168 num_blocks += tile_num_blocks; \
3169 sprite_blocks += tile_num_blocks; \
3171 if(num_blocks > MAX_BLOCKS) \
3173 flush_render_block_buffer(psx_gpu); \
3174 num_blocks = tile_num_blocks; \
3175 block = psx_gpu->blocks; \
3178 #define setup_sprite_tile_full_4bpp(edge) \
3180 vec_8x8u texels_low, texels_high; \
3182 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3184 while(sub_tile_height) \
3186 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3187 tbl_16(texels_low, texels, clut_low); \
3188 tbl_16(texels_high, texels, clut_high); \
3189 zip_8x16b(pixels, texels_low, texels_high); \
3191 block->texels = pixels; \
3192 block->draw_mask_bits = left_mask_bits; \
3193 block->fb_ptr = fb_ptr; \
3196 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3197 tbl_16(texels_low, texels, clut_low); \
3198 tbl_16(texels_high, texels, clut_high); \
3199 zip_8x16b(pixels, texels_low, texels_high); \
3201 block->texels = pixels; \
3202 block->draw_mask_bits = right_mask_bits; \
3203 block->fb_ptr = fb_ptr + 8; \
3207 texture_offset += 0x10; \
3208 sub_tile_height--; \
3210 texture_offset += 0xF00; \
3211 psx_gpu->num_blocks = num_blocks; \
3214 #define setup_sprite_tile_half_4bpp(edge) \
3216 vec_8x8u texels_low, texels_high; \
3218 setup_sprite_tile_add_blocks(sub_tile_height); \
3220 while(sub_tile_height) \
3222 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3223 tbl_16(texels_low, texels, clut_low); \
3224 tbl_16(texels_high, texels, clut_high); \
3225 zip_8x16b(pixels, texels_low, texels_high); \
3227 block->texels = pixels; \
3228 block->draw_mask_bits = edge##_mask_bits; \
3229 block->fb_ptr = fb_ptr; \
3233 texture_offset += 0x10; \
3234 sub_tile_height--; \
3236 texture_offset += 0xF00; \
3237 psx_gpu->num_blocks = num_blocks; \
3241 #define setup_sprite_tile_full_8bpp(edge) \
3243 setup_sprite_tile_add_blocks(sub_tile_height * 2); \
3245 while(sub_tile_height) \
3247 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3248 block->r = texels; \
3249 block->draw_mask_bits = left_mask_bits; \
3250 block->fb_ptr = fb_ptr; \
3253 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3254 block->r = texels; \
3255 block->draw_mask_bits = right_mask_bits; \
3256 block->fb_ptr = fb_ptr + 8; \
3260 texture_offset += 0x10; \
3261 sub_tile_height--; \
3263 texture_offset += 0xF00; \
3264 psx_gpu->num_blocks = num_blocks; \
3267 #define setup_sprite_tile_half_8bpp(edge) \
3269 setup_sprite_tile_add_blocks(sub_tile_height); \
3271 while(sub_tile_height) \
3273 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3274 block->r = texels; \
3275 block->draw_mask_bits = edge##_mask_bits; \
3276 block->fb_ptr = fb_ptr; \
3280 texture_offset += 0x10; \
3281 sub_tile_height--; \
3283 texture_offset += 0xF00; \
3284 psx_gpu->num_blocks = num_blocks; \
3288 #define setup_sprite_tile_column_edge_pre_adjust_half_right() \
3289 texture_offset = texture_offset_base + 8; \
3292 #define setup_sprite_tile_column_edge_pre_adjust_half_left() \
3293 texture_offset = texture_offset_base \
3295 #define setup_sprite_tile_column_edge_pre_adjust_half(edge) \
3296 setup_sprite_tile_column_edge_pre_adjust_half_##edge() \
3298 #define setup_sprite_tile_column_edge_pre_adjust_full(edge) \
3299 texture_offset = texture_offset_base \
3301 #define setup_sprite_tile_column_edge_post_adjust_half_right() \
3304 #define setup_sprite_tile_column_edge_post_adjust_half_left() \
3306 #define setup_sprite_tile_column_edge_post_adjust_half(edge) \
3307 setup_sprite_tile_column_edge_post_adjust_half_##edge() \
3309 #define setup_sprite_tile_column_edge_post_adjust_full(edge) \
3312 #define setup_sprite_tile_column_height_single(edge_mode, edge, texture_mode, \
3316 sub_tile_height = column_data; \
3317 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3318 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3319 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3322 #define setup_sprite_tile_column_height_multi(edge_mode, edge, texture_mode, \
3326 u32 tiles_remaining = column_data >> 16; \
3327 sub_tile_height = column_data & 0xFF; \
3328 setup_sprite_tile_column_edge_pre_adjust_##edge_mode##x4mode(edge); \
3329 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3330 tiles_remaining -= 1; \
3332 while(tiles_remaining) \
3334 sub_tile_height = 16; \
3335 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3336 tiles_remaining--; \
3339 sub_tile_height = (column_data >> 8) & 0xFF; \
3340 setup_sprite_tile_##edge_mode##_##texture_mode##x4mode(edge); \
3341 setup_sprite_tile_column_edge_post_adjust_##edge_mode##x4mode(edge); \
3345 #define setup_sprite_column_data_single() \
3346 column_data = height \
3348 #define setup_sprite_column_data_multi() \
3349 column_data = 16 - offset_v; \
3350 column_data |= ((height_rounded & 0xF) + 1) << 8; \
3351 column_data |= (tile_height - 1) << 16 \
3354 #define RIGHT_MASK_BIT_SHIFT 8
3355 #define RIGHT_MASK_BIT_SHIFT_4x 16
3357 #define setup_sprite_tile_column_width_single(texture_mode, multi_height, \
3358 edge_mode, edge, x4mode) \
3360 setup_sprite_column_data_##multi_height(); \
3361 left_mask_bits = left_block_mask | right_block_mask; \
3362 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3364 setup_sprite_tile_column_height_##multi_height(edge_mode, edge, \
3365 texture_mode, x4mode); \
3368 #define setup_sprite_tiled_advance_column() \
3369 texture_offset_base += 0x100; \
3370 if((texture_offset_base & 0xF00) == 0) \
3371 texture_offset_base -= (0x100 + 0xF00) \
3373 #define FB_PTR_MULTIPLIER 1
3374 #define FB_PTR_MULTIPLIER_4x 2
3376 #define setup_sprite_tile_column_width_multi(texture_mode, multi_height, \
3377 left_mode, right_mode, x4mode) \
3379 setup_sprite_column_data_##multi_height(); \
3380 s32 fb_ptr_advance_column = (16 - (1024 * height)) \
3381 * FB_PTR_MULTIPLIER##x4mode; \
3384 left_mask_bits = left_block_mask; \
3385 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3387 setup_sprite_tile_column_height_##multi_height(left_mode, right, \
3388 texture_mode, x4mode); \
3389 fb_ptr += fb_ptr_advance_column; \
3391 left_mask_bits = 0x00; \
3392 right_mask_bits = 0x00; \
3396 setup_sprite_tiled_advance_column(); \
3397 setup_sprite_tile_column_height_##multi_height(full, none, \
3398 texture_mode, x4mode); \
3399 fb_ptr += fb_ptr_advance_column; \
3403 left_mask_bits = right_block_mask; \
3404 right_mask_bits = left_mask_bits >> RIGHT_MASK_BIT_SHIFT##x4mode; \
3406 setup_sprite_tiled_advance_column(); \
3407 setup_sprite_tile_column_height_##multi_height(right_mode, left, \
3408 texture_mode, x4mode); \
3413 #define setup_sprite_tiled_initialize_4bpp_4x() \
3414 setup_sprite_tiled_initialize_4bpp_clut() \
3416 #define setup_sprite_tiled_initialize_8bpp_4x() \
3419 #define setup_sprite_tile_full_4bpp_4x(edge) \
3421 vec_8x8u texels_low, texels_high; \
3422 vec_8x16u pixels, pixels_wide; \
3423 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3424 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3425 u32 left_mask_bits_b = left_mask_bits >> 8; \
3426 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3427 u32 right_mask_bits_b = right_mask_bits >> 8; \
3429 while(sub_tile_height) \
3431 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3432 tbl_16(texels_low, texels, clut_low); \
3433 tbl_16(texels_high, texels, clut_high); \
3434 zip_8x16b(pixels, texels_low, texels_high); \
3436 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3437 block->texels = pixels_wide; \
3438 block->draw_mask_bits = left_mask_bits_a; \
3439 block->fb_ptr = fb_ptr; \
3442 block->texels = pixels_wide; \
3443 block->draw_mask_bits = left_mask_bits_a; \
3444 block->fb_ptr = fb_ptr + 1024; \
3447 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3448 block->texels = pixels_wide; \
3449 block->draw_mask_bits = left_mask_bits_b; \
3450 block->fb_ptr = fb_ptr + 8; \
3453 block->texels = pixels_wide; \
3454 block->draw_mask_bits = left_mask_bits_b; \
3455 block->fb_ptr = fb_ptr + 1024 + 8; \
3458 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3459 tbl_16(texels_low, texels, clut_low); \
3460 tbl_16(texels_high, texels, clut_high); \
3461 zip_8x16b(pixels, texels_low, texels_high); \
3463 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3464 block->texels = pixels_wide; \
3465 block->draw_mask_bits = right_mask_bits_a; \
3466 block->fb_ptr = fb_ptr + 16; \
3469 block->texels = pixels_wide; \
3470 block->draw_mask_bits = right_mask_bits_a; \
3471 block->fb_ptr = fb_ptr + 1024 + 16; \
3474 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3475 block->texels = pixels_wide; \
3476 block->draw_mask_bits = right_mask_bits_b; \
3477 block->fb_ptr = fb_ptr + 24; \
3480 block->texels = pixels_wide; \
3481 block->draw_mask_bits = right_mask_bits_b; \
3482 block->fb_ptr = fb_ptr + 1024 + 24; \
3486 texture_offset += 0x10; \
3487 sub_tile_height--; \
3489 texture_offset += 0xF00; \
3490 psx_gpu->num_blocks = num_blocks; \
3493 #define setup_sprite_tile_half_4bpp_4x(edge) \
3495 vec_8x8u texels_low, texels_high; \
3496 vec_8x16u pixels, pixels_wide; \
3497 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3498 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3499 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3501 while(sub_tile_height) \
3503 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3504 tbl_16(texels_low, texels, clut_low); \
3505 tbl_16(texels_high, texels, clut_high); \
3506 zip_8x16b(pixels, texels_low, texels_high); \
3508 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.low, pixels.low); \
3509 block->texels = pixels_wide; \
3510 block->draw_mask_bits = edge##_mask_bits_a; \
3511 block->fb_ptr = fb_ptr; \
3514 block->texels = pixels_wide; \
3515 block->draw_mask_bits = edge##_mask_bits_a; \
3516 block->fb_ptr = fb_ptr + 1024; \
3519 zip_4x32b(vector_cast(vec_4x32u, pixels_wide), pixels.high, pixels.high); \
3520 block->texels = pixels_wide; \
3521 block->draw_mask_bits = edge##_mask_bits_b; \
3522 block->fb_ptr = fb_ptr + 8; \
3525 block->texels = pixels_wide; \
3526 block->draw_mask_bits = edge##_mask_bits_b; \
3527 block->fb_ptr = fb_ptr + 1024 + 8; \
3531 texture_offset += 0x10; \
3532 sub_tile_height--; \
3534 texture_offset += 0xF00; \
3535 psx_gpu->num_blocks = num_blocks; \
3539 #define setup_sprite_tile_full_8bpp_4x(edge) \
3541 setup_sprite_tile_add_blocks(sub_tile_height * 2 * 4); \
3542 vec_16x8u texels_wide; \
3543 u32 left_mask_bits_a = left_mask_bits & 0xFF; \
3544 u32 left_mask_bits_b = left_mask_bits >> 8; \
3545 u32 right_mask_bits_a = right_mask_bits & 0xFF; \
3546 u32 right_mask_bits_b = right_mask_bits >> 8; \
3548 while(sub_tile_height) \
3550 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3551 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3552 block->r = texels_wide.low; \
3553 block->draw_mask_bits = left_mask_bits_a; \
3554 block->fb_ptr = fb_ptr; \
3557 block->r = texels_wide.low; \
3558 block->draw_mask_bits = left_mask_bits_a; \
3559 block->fb_ptr = fb_ptr + 1024; \
3562 block->r = texels_wide.high; \
3563 block->draw_mask_bits = left_mask_bits_b; \
3564 block->fb_ptr = fb_ptr + 8; \
3567 block->r = texels_wide.high; \
3568 block->draw_mask_bits = left_mask_bits_b; \
3569 block->fb_ptr = fb_ptr + 1024 + 8; \
3572 setup_sprite_tile_fetch_texel_block_8bpp(8); \
3573 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3574 block->r = texels_wide.low; \
3575 block->draw_mask_bits = right_mask_bits_a; \
3576 block->fb_ptr = fb_ptr + 16; \
3579 block->r = texels_wide.low; \
3580 block->draw_mask_bits = right_mask_bits_a; \
3581 block->fb_ptr = fb_ptr + 1024 + 16; \
3584 block->r = texels_wide.high; \
3585 block->draw_mask_bits = right_mask_bits_b; \
3586 block->fb_ptr = fb_ptr + 24; \
3589 block->r = texels_wide.high; \
3590 block->draw_mask_bits = right_mask_bits_b; \
3591 block->fb_ptr = fb_ptr + 24 + 1024; \
3595 texture_offset += 0x10; \
3596 sub_tile_height--; \
3598 texture_offset += 0xF00; \
3599 psx_gpu->num_blocks = num_blocks; \
3602 #define setup_sprite_tile_half_8bpp_4x(edge) \
3604 setup_sprite_tile_add_blocks(sub_tile_height * 4); \
3605 vec_16x8u texels_wide; \
3606 u32 edge##_mask_bits_a = edge##_mask_bits & 0xFF; \
3607 u32 edge##_mask_bits_b = edge##_mask_bits >> 8; \
3609 while(sub_tile_height) \
3611 setup_sprite_tile_fetch_texel_block_8bpp(0); \
3612 zip_8x16b(vector_cast(vec_8x16u, texels_wide), texels, texels); \
3613 block->r = texels_wide.low; \
3614 block->draw_mask_bits = edge##_mask_bits_a; \
3615 block->fb_ptr = fb_ptr; \
3618 block->r = texels_wide.low; \
3619 block->draw_mask_bits = edge##_mask_bits_a; \
3620 block->fb_ptr = fb_ptr + 1024; \
3623 block->r = texels_wide.high; \
3624 block->draw_mask_bits = edge##_mask_bits_b; \
3625 block->fb_ptr = fb_ptr + 8; \
3628 block->r = texels_wide.high; \
3629 block->draw_mask_bits = edge##_mask_bits_b; \
3630 block->fb_ptr = fb_ptr + 8 + 1024; \
3634 texture_offset += 0x10; \
3635 sub_tile_height--; \
3637 texture_offset += 0xF00; \
3638 psx_gpu->num_blocks = num_blocks; \
3642 #define setup_sprite_tile_column_edge_pre_adjust_half_right_4x() \
3643 texture_offset = texture_offset_base + 8; \
3646 #define setup_sprite_tile_column_edge_pre_adjust_half_left_4x() \
3647 texture_offset = texture_offset_base \
3649 #define setup_sprite_tile_column_edge_pre_adjust_half_4x(edge) \
3650 setup_sprite_tile_column_edge_pre_adjust_half_##edge##_4x() \
3652 #define setup_sprite_tile_column_edge_pre_adjust_full_4x(edge) \
3653 texture_offset = texture_offset_base \
3655 #define setup_sprite_tile_column_edge_post_adjust_half_right_4x() \
3658 #define setup_sprite_tile_column_edge_post_adjust_half_left_4x() \
3660 #define setup_sprite_tile_column_edge_post_adjust_half_4x(edge) \
3661 setup_sprite_tile_column_edge_post_adjust_half_##edge##_4x() \
3663 #define setup_sprite_tile_column_edge_post_adjust_full_4x(edge) \
3666 #define setup_sprite_offset_u_adjust() \
3668 #define setup_sprite_comapre_left_block_mask() \
3669 ((left_block_mask & 0xFF) == 0xFF) \
3671 #define setup_sprite_comapre_right_block_mask() \
3672 (((right_block_mask >> 8) & 0xFF) == 0xFF) \
3675 #define setup_sprite_offset_u_adjust_4x() \
3677 offset_u_right = offset_u_right * 2 + 1 \
3679 #define setup_sprite_comapre_left_block_mask_4x() \
3680 ((left_block_mask & 0xFFFF) == 0xFFFF) \
3682 #define setup_sprite_comapre_right_block_mask_4x() \
3683 (((right_block_mask >> 16) & 0xFFFF) == 0xFFFF) \
3686 #define setup_sprite_tiled_builder(texture_mode, x4mode) \
3687 void setup_sprite_##texture_mode##x4mode(psx_gpu_struct *psx_gpu, s32 x, s32 y,\
3688 s32 u, s32 v, s32 width, s32 height, u32 color) \
3690 s32 offset_u = u & 0xF; \
3691 s32 offset_v = v & 0xF; \
3693 s32 width_rounded = offset_u + width + 15; \
3694 s32 height_rounded = offset_v + height + 15; \
3695 s32 tile_height = height_rounded / 16; \
3696 s32 tile_width = width_rounded / 16; \
3697 u32 offset_u_right = width_rounded & 0xF; \
3699 setup_sprite_offset_u_adjust##x4mode(); \
3701 u32 left_block_mask = ~(0xFFFFFFFF << offset_u); \
3702 u32 right_block_mask = 0xFFFFFFFE << offset_u_right; \
3704 u32 left_mask_bits; \
3705 u32 right_mask_bits; \
3707 u32 sub_tile_height; \
3710 u32 texture_mask = (psx_gpu->texture_mask_width & 0xF) | \
3711 ((psx_gpu->texture_mask_height & 0xF) << 4) | \
3712 ((psx_gpu->texture_mask_width >> 4) << 8) | \
3713 ((psx_gpu->texture_mask_height >> 4) << 12); \
3714 u32 texture_offset = ((v & 0xF) << 4) | ((u & 0xF0) << 4) | \
3715 ((v & 0xF0) << 8); \
3716 u32 texture_offset_base = texture_offset; \
3719 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (x - offset_u); \
3720 u32 num_blocks = psx_gpu->num_blocks; \
3721 block_struct *block = psx_gpu->blocks + num_blocks; \
3723 u8 *texture_block_ptr; \
3726 setup_sprite_tiled_initialize_##texture_mode##x4mode(); \
3728 control_mask = tile_width == 1; \
3729 control_mask |= (tile_height == 1) << 1; \
3730 control_mask |= setup_sprite_comapre_left_block_mask##x4mode() << 2; \
3731 control_mask |= setup_sprite_comapre_right_block_mask##x4mode() << 3; \
3733 sprites_##texture_mode++; \
3735 switch(control_mask) \
3739 setup_sprite_tile_column_width_multi(texture_mode, multi, full, full, \
3744 setup_sprite_tile_column_width_single(texture_mode, multi, full, none, \
3749 setup_sprite_tile_column_width_multi(texture_mode, single, full, full, \
3754 setup_sprite_tile_column_width_single(texture_mode, single, full, none, \
3759 setup_sprite_tile_column_width_multi(texture_mode, multi, half, full, \
3764 setup_sprite_tile_column_width_single(texture_mode, multi, half, right, \
3769 setup_sprite_tile_column_width_multi(texture_mode, single, half, full, \
3774 setup_sprite_tile_column_width_single(texture_mode, single, half, right, \
3779 setup_sprite_tile_column_width_multi(texture_mode, multi, full, half, \
3784 setup_sprite_tile_column_width_single(texture_mode, multi, half, left, \
3789 setup_sprite_tile_column_width_multi(texture_mode, single, full, half, \
3794 setup_sprite_tile_column_width_single(texture_mode, single, half, left, \
3799 setup_sprite_tile_column_width_multi(texture_mode, multi, half, half, \
3804 setup_sprite_tile_column_width_multi(texture_mode, single, half, half, \
3811 setup_sprite_tiled_builder(4bpp,);
3812 setup_sprite_tiled_builder(8bpp,);
3814 setup_sprite_tiled_builder(4bpp,_4x);
3815 setup_sprite_tiled_builder(8bpp,_4x);
3818 #if !defined(NEON_BUILD) || defined(SIMD_BUILD)
3820 void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3821 s32 v, s32 width, s32 height, u32 color)
3823 u32 left_offset = u & 0x7;
3824 u32 width_rounded = width + left_offset + 7;
3826 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + (s32)(x - left_offset);
3827 u32 right_width = width_rounded & 0x7;
3828 u32 block_width = width_rounded / 8;
3829 u32 fb_ptr_pitch = (1024 + 8) - (block_width * 8);
3831 u32 left_mask_bits = ~(0xFF << left_offset);
3832 u32 right_mask_bits = 0xFE << right_width;
3834 u32 texture_offset_base = u + (v * 1024);
3836 psx_gpu->texture_mask_width | (psx_gpu->texture_mask_height * 1024);
3838 u32 blocks_remaining;
3839 u32 num_blocks = psx_gpu->num_blocks;
3840 block_struct *block = psx_gpu->blocks + num_blocks;
3842 u16 *texture_page_ptr = psx_gpu->texture_page_ptr;
3843 u16 *texture_block_ptr;
3845 texture_offset_base &= ~0x7;
3847 stats_add(sprites_16bpp, 1);
3849 if(block_width == 1)
3851 u32 mask_bits = left_mask_bits | right_mask_bits;
3858 if(num_blocks > MAX_BLOCKS)
3860 flush_render_block_buffer(psx_gpu);
3862 block = psx_gpu->blocks;
3866 texture_page_ptr + (texture_offset_base & texture_mask);
3868 block->texels = *(vec_8x16u *)texture_block_ptr;
3869 block->draw_mask_bits = mask_bits;
3870 block->fb_ptr = fb_ptr;
3874 texture_offset_base += 1024;
3878 psx_gpu->num_blocks = num_blocks;
3887 blocks_remaining = block_width - 2;
3888 num_blocks += block_width;
3889 sprite_blocks += block_width;
3891 if(num_blocks > MAX_BLOCKS)
3893 flush_render_block_buffer(psx_gpu);
3894 num_blocks = block_width;
3895 block = psx_gpu->blocks;
3898 texture_offset = texture_offset_base;
3899 texture_offset_base += 1024;
3901 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3902 block->texels = *(vec_8x16u *)texture_block_ptr;
3904 block->draw_mask_bits = left_mask_bits;
3905 block->fb_ptr = fb_ptr;
3907 texture_offset += 8;
3911 while(blocks_remaining)
3913 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3914 block->texels = *(vec_8x16u *)texture_block_ptr;
3916 block->draw_mask_bits = 0;
3917 block->fb_ptr = fb_ptr;
3919 texture_offset += 8;
3926 texture_block_ptr = texture_page_ptr + (texture_offset & texture_mask);
3927 block->texels = *(vec_8x16u *)texture_block_ptr;
3929 block->draw_mask_bits = right_mask_bits;
3930 block->fb_ptr = fb_ptr;
3932 fb_ptr += fb_ptr_pitch;
3936 psx_gpu->num_blocks = num_blocks;
3945 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
3946 s32 v, s32 width, s32 height, u32 color)
3948 u32 right_width = ((width - 1) & 0x7) + 1;
3949 u32 right_mask_bits = (0xFF << right_width);
3950 u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x;
3951 u32 block_width = (width + 7) / 8;
3952 u32 fb_ptr_pitch = 1024 - ((block_width - 1) * 8);
3953 u32 blocks_remaining;
3954 u32 num_blocks = psx_gpu->num_blocks;
3955 block_struct *block = psx_gpu->blocks + num_blocks;
3957 u32 color_r = color & 0xFF;
3958 u32 color_g = (color >> 8) & 0xFF;
3959 u32 color_b = (color >> 16) & 0xFF;
3961 vec_8x16u right_mask;
3962 vec_8x16u test_mask = psx_gpu->test_mask;
3963 vec_8x16u zero_mask;
3965 sprites_untextured++;
3967 color = (color_r >> 3) | ((color_g >> 3) << 5) | ((color_b >> 3) << 10);
3969 dup_8x16b(colors, color);
3970 dup_8x16b(zero_mask, 0x00);
3971 dup_8x16b(right_mask, right_mask_bits);
3972 tst_8x16b(right_mask, right_mask, test_mask);
3976 blocks_remaining = block_width - 1;
3977 num_blocks += block_width;
3980 sprite_blocks += block_width;
3983 if(num_blocks > MAX_BLOCKS)
3985 flush_render_block_buffer(psx_gpu);
3986 num_blocks = block_width;
3987 block = psx_gpu->blocks;
3990 while(blocks_remaining)
3992 block->pixels = colors;
3993 block->draw_mask = zero_mask;
3994 block->fb_ptr = fb_ptr;
4001 block->pixels = colors;
4002 block->draw_mask = right_mask;
4003 block->fb_ptr = fb_ptr;
4006 fb_ptr += fb_ptr_pitch;
4009 psx_gpu->num_blocks = num_blocks;
4015 static void __attribute__((noinline))
4016 setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4017 s32 v, s32 width, s32 height, u32 color)
4019 u32 r = color & 0xFF;
4020 u32 g = (color >> 8) & 0xFF;
4021 u32 b = (color >> 16) & 0xFF;
4022 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4024 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4026 u16 *vram_ptr16 = psx_gpu->vram_out_ptr + x + (y * 1024);
4031 if(psx_gpu->num_blocks)
4033 flush_render_block_buffer(psx_gpu);
4040 vram_ptr = (void *)vram_ptr16;
4041 if((uintptr_t)vram_ptr16 & 2)
4043 *vram_ptr16 = color_32bpp;
4044 vram_ptr = (void *)(vram_ptr16 + 1);
4048 while(num_width >= 4 * 2)
4050 vram_ptr[0] = color_32bpp;
4051 vram_ptr[1] = color_32bpp;
4052 vram_ptr[2] = color_32bpp;
4053 vram_ptr[3] = color_32bpp;
4059 while(num_width >= 2)
4061 *vram_ptr++ = color_32bpp;
4067 *(u16 *)vram_ptr = color_32bpp;
4075 void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4076 s32 v, s32 width, s32 height, u32 color);
4078 void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u,
4079 s32 v, s32 width, s32 height, u32 color)
4081 if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE |
4082 RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 &&
4083 (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0)
4085 setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color);
4091 s32 w1 = width > 512 ? 512 : width;
4092 setup_sprite_untextured_512(psx_gpu, x, y, 0, 0, w1, height, color);
4099 #define setup_sprite_blocks_switch_textured(texture_mode) \
4100 setup_sprite_##texture_mode \
4102 #define setup_sprite_blocks_switch_untextured(texture_mode) \
4103 setup_sprite_untextured \
4105 #define setup_sprite_blocks_switch(texturing, texture_mode) \
4106 setup_sprite_blocks_switch_##texturing(texture_mode) \
4109 #define texture_sprite_blocks_switch_4bpp() \
4110 texture_blocks_untextured \
4112 #define texture_sprite_blocks_switch_8bpp() \
4113 texture_sprite_blocks_8bpp \
4115 #define texture_sprite_blocks_switch_16bpp() \
4116 texture_blocks_untextured \
4118 #define texture_sprite_blocks_switch_untextured(texture_mode) \
4119 texture_blocks_untextured \
4121 #define texture_sprite_blocks_switch_textured(texture_mode) \
4122 texture_sprite_blocks_switch_##texture_mode() \
4124 #define render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4125 mask_evaluate, shading, dithering, texturing, blending, modulation) \
4127 setup_sprite_blocks_switch(texturing, texture_mode), \
4128 texture_sprite_blocks_switch_##texturing(texture_mode), \
4129 shade_blocks_switch(unshaded, texturing, modulation, undithered, blending, \
4131 blend_blocks_switch(texturing, blending, blend_mode, mask_evaluate) \
4134 #define render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4135 mask_evaluate, shading, dithering, texturing, blending) \
4136 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4137 mask_evaluate, shading, dithering, texturing, blending, modulated), \
4138 render_sprite_blocks_switch_block_modulation(texture_mode, blend_mode, \
4139 mask_evaluate, shading, dithering, texturing, blending, unmodulated) \
4141 #define render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4142 mask_evaluate, shading, dithering, texturing) \
4143 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4144 mask_evaluate, shading, dithering, texturing, unblended), \
4145 render_sprite_blocks_switch_block_blending(texture_mode, blend_mode, \
4146 mask_evaluate, shading, dithering, texturing, blended) \
4148 #define render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4149 mask_evaluate, shading, dithering) \
4150 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4151 mask_evaluate, shading, dithering, untextured), \
4152 render_sprite_blocks_switch_block_texturing(texture_mode, blend_mode, \
4153 mask_evaluate, shading, dithering, textured) \
4155 #define render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4156 mask_evaluate, shading) \
4157 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4158 mask_evaluate, shading, undithered), \
4159 render_sprite_blocks_switch_block_dithering(texture_mode, blend_mode, \
4160 mask_evaluate, shading, dithered) \
4162 #define render_sprite_blocks_switch_block_mask_evaluate(texture_mode, \
4163 blend_mode, mask_evaluate) \
4164 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4165 mask_evaluate, unshaded), \
4166 render_sprite_blocks_switch_block_shading(texture_mode, blend_mode, \
4167 mask_evaluate, shaded) \
4169 #define render_sprite_blocks_switch_block_blend_mode(texture_mode, blend_mode) \
4170 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4172 render_sprite_blocks_switch_block_mask_evaluate(texture_mode, blend_mode, \
4175 #define render_sprite_blocks_switch_block_texture_mode(texture_mode) \
4176 render_sprite_blocks_switch_block_blend_mode(texture_mode, average), \
4177 render_sprite_blocks_switch_block_blend_mode(texture_mode, add), \
4178 render_sprite_blocks_switch_block_blend_mode(texture_mode, subtract), \
4179 render_sprite_blocks_switch_block_blend_mode(texture_mode, add_fourth) \
4181 #define render_sprite_blocks_switch_block() \
4182 render_sprite_blocks_switch_block_texture_mode(4bpp), \
4183 render_sprite_blocks_switch_block_texture_mode(8bpp), \
4184 render_sprite_blocks_switch_block_texture_mode(16bpp), \
4185 render_sprite_blocks_switch_block_texture_mode(16bpp) \
4188 render_block_handler_struct render_sprite_block_handlers[] =
4190 render_sprite_blocks_switch_block()
4194 void render_sprite(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 u, u32 v,
4195 s32 width, s32 height, u32 flags, u32 color)
4197 s32 x_right = x + width - 1;
4198 s32 y_bottom = y + height - 1;
4204 if(invalidate_texture_cache_region_viewport(psx_gpu, x, y, x_right,
4210 if(x < psx_gpu->viewport_start_x)
4212 u32 clip = psx_gpu->viewport_start_x - x;
4218 if(y < psx_gpu->viewport_start_y)
4220 s32 clip = psx_gpu->viewport_start_y - y;
4226 if(x_right > psx_gpu->viewport_end_x)
4227 width -= x_right - psx_gpu->viewport_end_x;
4229 if(y_bottom > psx_gpu->viewport_end_y)
4230 height -= y_bottom - psx_gpu->viewport_end_y;
4232 if((width <= 0) || (height <= 0))
4236 span_pixels += width * height;
4240 u32 render_state = flags &
4241 (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
4242 RENDER_FLAGS_TEXTURE_MAP);
4244 (psx_gpu->render_state_base & ~RENDER_STATE_DITHER);
4246 if((psx_gpu->render_state != render_state) ||
4247 (psx_gpu->primitive_type != PRIMITIVE_TYPE_SPRITE))
4249 psx_gpu->render_state = render_state;
4250 flush_render_block_buffer(psx_gpu);
4256 psx_gpu->primitive_type = PRIMITIVE_TYPE_SPRITE;
4260 if(psx_gpu->triangle_color != color)
4262 flush_render_block_buffer(psx_gpu);
4263 psx_gpu->triangle_color = color;
4266 if(color == 0x808080)
4267 render_state |= RENDER_FLAGS_MODULATE_TEXELS;
4269 render_block_handler_struct *render_block_handler =
4270 &(render_sprite_block_handlers[render_state]);
4271 psx_gpu->render_block_handler = render_block_handler;
4273 ((setup_sprite_function_type *)render_block_handler->setup_blocks)
4274 (psx_gpu, x, y, u, v, width, height, color);
4277 #define draw_pixel_line_mask_evaluate_yes() \
4278 if((*vram_ptr & 0x8000) == 0) \
4280 #define draw_pixel_line_mask_evaluate_no() \
4283 #define draw_pixel_line_shaded() \
4285 color_r = fixed_to_int(current_r); \
4286 color_g = fixed_to_int(current_g); \
4287 color_b = fixed_to_int(current_b); \
4289 current_r += gradient_r; \
4290 current_g += gradient_g; \
4291 current_b += gradient_b; \
4294 #define draw_pixel_line_unshaded() \
4296 color_r = color & 0xFF; \
4297 color_g = (color >> 8) & 0xFF; \
4298 color_b = (color >> 16) & 0xFF; \
4302 #define draw_pixel_line_dithered(_x, _y) \
4304 u32 dither_xor = _x ^ _y; \
4305 s32 dither_offset = (dither_xor >> 1) & 0x1; \
4306 dither_offset |= (_y & 0x1) << 1; \
4307 dither_offset |= (dither_xor & 0x1) << 2; \
4308 dither_offset -= 4; \
4310 color_r += dither_offset; \
4311 color_g += dither_offset; \
4312 color_b += dither_offset; \
4333 #define draw_pixel_line_undithered(_x, _y) \
4336 #define draw_pixel_line_average() \
4337 color_r = (color_r + fb_r) / 2; \
4338 color_g = (color_g + fb_g) / 2; \
4339 color_b = (color_b + fb_b) / 2 \
4341 #define draw_pixel_line_add() \
4356 #define draw_pixel_line_subtract() \
4357 color_r = fb_r - color_r; \
4358 color_g = fb_g - color_g; \
4359 color_b = fb_b - color_b; \
4370 #define draw_pixel_line_add_fourth() \
4371 color_r = fb_r + (color_r / 4); \
4372 color_g = fb_g + (color_g / 4); \
4373 color_b = fb_b + (color_b / 4); \
4385 #define draw_pixel_line_blended(blend_mode) \
4386 s32 fb_pixel = *vram_ptr; \
4387 s32 fb_r = fb_pixel & 0x1F; \
4388 s32 fb_g = (fb_pixel >> 5) & 0x1F; \
4389 s32 fb_b = (fb_pixel >> 10) & 0x1F; \
4391 draw_pixel_line_##blend_mode() \
4393 #define draw_pixel_line_unblended(blend_mode) \
4396 #define draw_pixel_line(_x, _y, shading, blending, dithering, mask_evaluate, \
4398 if((_x >= psx_gpu->viewport_start_x) && (_y >= psx_gpu->viewport_start_y) && \
4399 (_x <= psx_gpu->viewport_end_x) && (_y <= psx_gpu->viewport_end_y)) \
4401 draw_pixel_line_mask_evaluate_##mask_evaluate() \
4403 draw_pixel_line_##shading(); \
4404 draw_pixel_line_##dithering(_x, _y); \
4410 draw_pixel_line_##blending(blend_mode); \
4412 *vram_ptr = color_r | (color_g << 5) | (color_b << 10) | \
4413 psx_gpu->mask_msb; \
4417 #define update_increment(value) \
4420 #define update_decrement(value) \
4423 #define update_vram_row_increment(value) \
4426 #define update_vram_row_decrement(value) \
4429 #define compare_increment(a, b) \
4432 #define compare_decrement(a, b) \
4435 #define set_line_gradients(minor) \
4437 s32 gradient_divisor = delta_##minor; \
4438 if(gradient_divisor != 0) \
4440 gradient_r = int_to_fixed(vertex_b->r - vertex_a->r) / gradient_divisor; \
4441 gradient_g = int_to_fixed(vertex_b->g - vertex_a->g) / gradient_divisor; \
4442 gradient_b = int_to_fixed(vertex_b->b - vertex_a->b) / gradient_divisor; \
4450 current_r = fixed_center(vertex_a->r); \
4451 current_g = fixed_center(vertex_a->g); \
4452 current_b = fixed_center(vertex_a->b); \
4455 #define draw_line_span_horizontal(direction, shading, blending, dithering, \
4456 mask_evaluate, blend_mode) \
4459 error_step = delta_y * 2; \
4460 error_wrap = delta_x * 2; \
4464 set_line_gradients(x); \
4466 for(current_x = x_a; current_x <= x_b; current_x++) \
4468 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4469 mask_evaluate, blend_mode); \
4470 error += error_step; \
4473 if(error >= error_wrap) \
4475 update_##direction(current_y); \
4476 update_vram_row_##direction(); \
4477 error -= error_wrap; \
4482 #define draw_line_span_vertical(direction, shading, blending, dithering, \
4483 mask_evaluate, blend_mode) \
4486 error_step = delta_x * 2; \
4487 error_wrap = delta_y * 2; \
4491 set_line_gradients(y); \
4493 for(current_y = y_a; compare_##direction(current_y, y_b); \
4494 update_##direction(current_y)) \
4496 draw_pixel_line(current_x, current_y, shading, blending, dithering, \
4497 mask_evaluate, blend_mode); \
4498 error += error_step; \
4499 update_vram_row_##direction(); \
4501 if(error > error_wrap) \
4505 error -= error_wrap; \
4511 #define render_line_body(shading, blending, dithering, mask_evaluate, \
4517 if(delta_x > delta_y) \
4519 draw_line_span_horizontal(decrement, shading, blending, dithering, \
4520 mask_evaluate, blend_mode); \
4524 draw_line_span_vertical(decrement, shading, blending, dithering, \
4525 mask_evaluate, blend_mode); \
4530 if(delta_x > delta_y) \
4532 draw_line_span_horizontal(increment, shading, blending, dithering, \
4533 mask_evaluate, blend_mode); \
4537 draw_line_span_vertical(increment, shading, blending, dithering, \
4538 mask_evaluate, blend_mode); \
4543 void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags,
4544 u32 color, int double_resolution)
4546 s32 color_r, color_g, color_b;
4547 u32 triangle_winding = 0;
4549 fixed_type gradient_r = 0;
4550 fixed_type gradient_g = 0;
4551 fixed_type gradient_b = 0;
4552 fixed_type current_r = 0;
4553 fixed_type current_g = 0;
4554 fixed_type current_b = 0;
4559 s32 delta_x, delta_y;
4570 flush_render_block_buffer(psx_gpu);
4571 psx_gpu->primitive_type = PRIMITIVE_TYPE_LINE;
4573 vertex_struct *vertex_a = &(vertexes[0]);
4574 vertex_struct *vertex_b = &(vertexes[1]);
4582 if(vertex_a->x >= vertex_b->x)
4584 vertex_swap(vertex_a, vertex_b);
4593 delta_x = x_b - x_a;
4594 delta_y = y_b - y_a;
4596 if(delta_x >= 1024 || delta_y >= 512 || delta_y <= -512)
4599 if(double_resolution)
4609 flags &= ~RENDER_FLAGS_TEXTURE_MAP;
4611 vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
4615 if(flags & RENDER_FLAGS_SHADE)
4616 control_mask |= 0x1;
4618 if(flags & RENDER_FLAGS_BLEND)
4620 control_mask |= 0x2;
4621 control_mask |= ((psx_gpu->render_state_base >> 6) & 0x3) << 4;
4624 if(psx_gpu->render_state_base & RENDER_STATE_DITHER)
4625 control_mask |= 0x4;
4627 if(psx_gpu->render_state_base & RENDER_STATE_MASK_EVALUATE)
4628 control_mask |= 0x8;
4630 switch(control_mask)
4633 render_line_body(unshaded, unblended, undithered, no, none);
4637 render_line_body(shaded, unblended, undithered, no, none);
4641 render_line_body(unshaded, blended, undithered, no, average);
4645 render_line_body(shaded, blended, undithered, no, average);
4649 render_line_body(unshaded, unblended, dithered, no, none);
4653 render_line_body(shaded, unblended, dithered, no, none);
4657 render_line_body(unshaded, blended, dithered, no, average);
4661 render_line_body(shaded, blended, dithered, no, average);
4665 render_line_body(unshaded, unblended, undithered, yes, none);
4669 render_line_body(shaded, unblended, undithered, yes, none);
4673 render_line_body(unshaded, blended, undithered, yes, average);
4677 render_line_body(shaded, blended, undithered, yes, average);
4681 render_line_body(unshaded, unblended, dithered, yes, none);
4685 render_line_body(shaded, unblended, dithered, yes, none);
4689 render_line_body(unshaded, blended, dithered, yes, average);
4693 render_line_body(shaded, blended, dithered, yes, average);
4697 render_line_body(unshaded, blended, undithered, no, add);
4701 render_line_body(shaded, blended, undithered, no, add);
4705 render_line_body(unshaded, blended, dithered, no, add);
4709 render_line_body(shaded, blended, dithered, no, add);
4713 render_line_body(unshaded, blended, undithered, yes, add);
4717 render_line_body(shaded, blended, undithered, yes, add);
4721 render_line_body(unshaded, blended, dithered, yes, add);
4725 render_line_body(shaded, blended, dithered, yes, add);
4729 render_line_body(unshaded, blended, undithered, no, subtract);
4733 render_line_body(shaded, blended, undithered, no, subtract);
4737 render_line_body(unshaded, blended, dithered, no, subtract);
4741 render_line_body(shaded, blended, dithered, no, subtract);
4745 render_line_body(unshaded, blended, undithered, yes, subtract);
4749 render_line_body(shaded, blended, undithered, yes, subtract);
4753 render_line_body(unshaded, blended, dithered, yes, subtract);
4757 render_line_body(shaded, blended, dithered, yes, subtract);
4761 render_line_body(unshaded, blended, undithered, no, add_fourth);
4765 render_line_body(shaded, blended, undithered, no, add_fourth);
4769 render_line_body(unshaded, blended, dithered, no, add_fourth);
4773 render_line_body(shaded, blended, dithered, no, add_fourth);
4777 render_line_body(unshaded, blended, undithered, yes, add_fourth);
4781 render_line_body(shaded, blended, undithered, yes, add_fourth);
4785 render_line_body(unshaded, blended, dithered, yes, add_fourth);
4789 render_line_body(shaded, blended, dithered, yes, add_fourth);
4795 void render_block_fill(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4796 u32 width, u32 height)
4798 if((width == 0) || (height == 0))
4801 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4803 u32 r = color & 0xFF;
4804 u32 g = (color >> 8) & 0xFF;
4805 u32 b = (color >> 16) & 0xFF;
4806 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4808 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4810 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4812 u32 pitch = 512 - (width / 2);
4815 if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
4820 if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
4829 vram_ptr[0] = color_32bpp;
4830 vram_ptr[1] = color_32bpp;
4831 vram_ptr[2] = color_32bpp;
4832 vram_ptr[3] = color_32bpp;
4833 vram_ptr[4] = color_32bpp;
4834 vram_ptr[5] = color_32bpp;
4835 vram_ptr[6] = color_32bpp;
4836 vram_ptr[7] = color_32bpp;
4847 void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
4848 u32 width, u32 height)
4850 if((width == 0) || (height == 0))
4856 u32 r = color & 0xFF;
4857 u32 g = (color >> 8) & 0xFF;
4858 u32 b = (color >> 16) & 0xFF;
4859 u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
4861 u32 color_32bpp = color_16bpp | (color_16bpp << 16);
4863 u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
4865 u32 pitch = 1024 / 2 - (width / 2);
4873 vram_ptr[0] = color_32bpp;
4874 vram_ptr[1] = color_32bpp;
4875 vram_ptr[2] = color_32bpp;
4876 vram_ptr[3] = color_32bpp;
4877 vram_ptr[4] = color_32bpp;
4878 vram_ptr[5] = color_32bpp;
4879 vram_ptr[6] = color_32bpp;
4880 vram_ptr[7] = color_32bpp;
4892 void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
4893 u32 width, u32 height, u32 pitch)
4895 u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
4897 u32 mask_msb = psx_gpu->mask_msb;
4899 if((width == 0) || (height == 0))
4902 flush_render_block_buffer(psx_gpu);
4903 invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
4905 for(draw_y = 0; draw_y < height; draw_y++)
4907 for(draw_x = 0; draw_x < width; draw_x++)
4909 vram_ptr[draw_x] = source[draw_x] | mask_msb;
4917 void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y,
4918 u32 dest_x, u32 dest_y, u32 width, u32 height)
4920 render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024),
4921 dest_x, dest_y, width, height, 1024);
4925 void initialize_reciprocal_table(void)
4928 u32 height_normalized;
4929 u32 height_reciprocal;
4932 for(height = 1; height < sizeof(reciprocal_table)
4933 / sizeof(reciprocal_table[0]); height++)
4935 shift = __builtin_clz(height);
4936 height_normalized = height << shift;
4937 height_reciprocal = ((1ULL << 51) + (height_normalized - 1)) /
4940 shift = 32 - (51 - shift);
4942 reciprocal_table[height] = (height_reciprocal << 10) | shift;
4947 #define dither_table_row(a, b, c, d) \
4948 ((a & 0xFF) | ((b & 0xFF) << 8) | ((c & 0xFF) << 16) | ((d & 0xFF) << 24)) \
4950 void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram)
4952 vec_8x16u test_mask =
4953 { { { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80 } } };
4955 psx_gpu->test_mask = test_mask;
4957 psx_gpu->dirty_textures_4bpp_mask = 0xFFFFFFFF;
4958 psx_gpu->dirty_textures_8bpp_mask = 0xFFFFFFFF;
4959 psx_gpu->dirty_textures_8bpp_alternate_mask = 0xFFFFFFFF;
4960 psx_gpu->viewport_mask = 0;
4961 psx_gpu->current_texture_page = 0;
4962 psx_gpu->current_texture_mask = 0;
4963 psx_gpu->last_8bpp_texture_page = 0;
4965 psx_gpu->clut_settings = 0;
4966 psx_gpu->texture_settings = 0;
4967 psx_gpu->render_state = 0;
4968 psx_gpu->render_state_base = 0;
4969 psx_gpu->num_blocks = 0;
4970 psx_gpu->uvrgb_phase = 0x8000;
4972 psx_gpu->vram_ptr = vram;
4973 psx_gpu->vram_out_ptr = vram;
4975 psx_gpu->texture_page_base = psx_gpu->vram_ptr;
4976 psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
4977 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
4979 psx_gpu->viewport_start_x = psx_gpu->viewport_start_y = 0;
4980 psx_gpu->viewport_end_x = psx_gpu->viewport_end_y = 0;
4981 psx_gpu->mask_msb = 0;
4983 psx_gpu->texture_window_x = 0;
4984 psx_gpu->texture_window_y = 0;
4985 psx_gpu->texture_mask_width = 0xFF;
4986 psx_gpu->texture_mask_height = 0xFF;
4988 psx_gpu->render_mode = 0;
4990 memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
4992 initialize_reciprocal_table();
4993 psx_gpu->reciprocal_table_ptr = reciprocal_table;
5000 // (minus ones(4) * 4)
5002 // d0: (1 3 5 7): x1 ^ y1
5003 // d1: (2 3 6 7): y0
5004 // d2: (4 5 6 7): x0 ^ y0
5006 psx_gpu->dither_table[0] = dither_table_row(-4, 0, -3, 1);
5007 psx_gpu->dither_table[1] = dither_table_row(2, -2, 3, -1);
5008 psx_gpu->dither_table[2] = dither_table_row(-3, 1, -4, 0);
5009 psx_gpu->dither_table[3] = dither_table_row(3, -1, 2, -2);
5011 psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN;
5013 psx_gpu->saved_hres = 256;
5019 gettimeofday(&tv, NULL);
5021 return (tv.tv_sec * 1000000ULL) + tv.tv_usec;
5024 #if 0 //def NEON_BUILD
5029 __asm__ volatile("mrc p15, 0, %0, c9, c13, 0" : "=r"(counter));
5034 void init_counter(void)
5037 asm volatile("mrc p15, 0, %0, c9, c12, 0" : "=r"(value));
5038 value |= 5; // master enable, ccnt reset
5039 value &= ~8; // ccnt divider 0
5040 asm volatile("mcr p15, 0, %0, c9, c12, 0" :: "r"(value));
5041 // enable cycle counter
5042 asm volatile("mcr p15, 0, %0, c9, c12, 1" :: "r"(1 << 31));
5045 void triangle_benchmark(psx_gpu_struct *psx_gpu)
5052 const u32 iterations = 500000;
5054 psx_gpu->num_blocks = 64;
5055 psx_gpu->clut_ptr = psx_gpu->vram_ptr;
5057 for(i = 0; i < 64; i++)
5059 memset(&(psx_gpu->blocks[i].r), 0, 16);
5064 ticks = get_counter();
5066 for(i = 0; i < iterations; i++)
5068 texture_sprite_blocks_8bpp(psx_gpu);
5071 ticks_elapsed = get_counter() - ticks;
5073 printf("benchmark: %lf cycles\n", (double)ticks_elapsed / (iterations * 64));
5078 #include "psx_gpu_4x.c"
5080 // vim:ts=2:sw=2:expandtab