X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=plugins%2Fgpu_neon%2Fpsx_gpu%2Fpsx_gpu.c;h=62080f3f8c835d20ec195cdc8e15fef1d2847b65;hb=1587b08ecb80dd257ac255f76e716a6053c2785b;hp=1d513d8be19d4260784f37a4958eef7b9aebd5af;hpb=df740cdce57508b93e158391e2331eebacd24cb4;p=pcsx_rearmed.git diff --git a/plugins/gpu_neon/psx_gpu/psx_gpu.c b/plugins/gpu_neon/psx_gpu/psx_gpu.c index 1d513d8b..62080f3f 100644 --- a/plugins/gpu_neon/psx_gpu/psx_gpu.c +++ b/plugins/gpu_neon/psx_gpu/psx_gpu.c @@ -14,13 +14,24 @@ #include #include +#include +#include #include +#include #include "common.h" #ifndef NEON_BUILD #include "vector_ops.h" #endif #include "psx_gpu_simd.h" +#include "psx_gpu_offsets.h" + +#if 0 +void dump_r_d(const char *name, void *dump); +void dump_r_q(const char *name, void *dump); +#define dumprd(n) dump_r_d(#n, n.e) +#define dumprq(n) dump_r_q(#n, n.e) +#endif u32 span_pixels = 0; u32 span_pixel_blocks = 0; @@ -251,8 +262,8 @@ u32 invalidate_texture_cache_region_viewport(psx_gpu_struct *psx_gpu, u32 x1, return mask; } -void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1, - u32 x2, u32 y2) +static void update_texture_cache_region_(psx_gpu_struct *psx_gpu, + u32 x1, u32 y1, u32 x2, u32 y2) { u32 mask = texture_region_mask(x1, y1, x2, y2); u32 texture_page; @@ -304,6 +315,22 @@ void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1, } } +void update_texture_cache_region(psx_gpu_struct *psx_gpu, u32 x1, u32 y1, + u32 x2, u32 y2) +{ + s32 w = x2 - x1; + do + { + x2 = x1 + w; + if (x2 > 1023) + x2 = 1023; + update_texture_cache_region_(psx_gpu, x1, y1, x2, y2); + w -= x2 - x1; + x1 = 0; + } + while (unlikely(w > 0)); +} + #ifndef NEON_BUILD void update_texture_4bpp_cache(psx_gpu_struct *psx_gpu) @@ -764,31 +791,30 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, printf("mismatch on %s %s: %x vs %x\n", #_a, #_b, _a, _b) \ -#ifndef NDEBUG -#define setup_spans_debug_check(span_edge_data_element) \ -{ \ - u32 _num_spans = &span_edge_data_element - psx_gpu->span_edge_data; \ - if (_num_spans > MAX_SPANS) \ - *(int *)0 = 1; \ - if (_num_spans < psx_gpu->num_spans) \ - { \ - if(span_edge_data_element.num_blocks > MAX_BLOCKS_PER_ROW) \ - *(int *)0 = 1; \ - if(span_edge_data_element.y > 2048) \ - *(int *)0 = 1; \ - } \ -} \ - +#if !defined(NEON_BUILD) && !defined(NDEBUG) +static void setup_spans_debug_check(psx_gpu_struct *psx_gpu, + edge_data_struct *span_edge_data_element) +{ + u32 _num_spans = span_edge_data_element - psx_gpu->span_edge_data; + if (_num_spans > MAX_SPANS) + *(volatile int *)0 = 1; + if (_num_spans < psx_gpu->num_spans) + { + if(span_edge_data_element->num_blocks > MAX_BLOCKS_PER_ROW) + *(volatile int *)0 = 2; + if(span_edge_data_element->y >= 2048) + *(volatile int *)0 = 3; + } +} #else -#define setup_spans_debug_check(span_edge_data_element) \ - +#define setup_spans_debug_check(psx_gpu, span_edge_data_element) #endif #define setup_spans_prologue_alternate_yes() \ vec_2x64s alternate_x; \ vec_2x64s alternate_dx_dy; \ vec_4x32s alternate_x_32; \ - vec_2x32s alternate_x_16; \ + vec_4x16u alternate_x_16; \ \ vec_4x16u alternate_select; \ vec_4x16s y_mid_point; \ @@ -848,6 +874,7 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, span_b_offset = psx_gpu->span_b_offset; \ \ vec_8x16u c_0x0001; \ + vec_4x16u c_max_blocks_per_row; \ \ dup_8x16b(c_0x0001, 0x0001); \ dup_8x16b(left_edge, psx_gpu->viewport_start_x); \ @@ -856,6 +883,7 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, dup_4x16b(c_0x04, 0x04); \ dup_4x16b(c_0x07, 0x07); \ dup_4x16b(c_0xFFFE, 0xFFFE); \ + dup_4x16b(c_max_blocks_per_row, MAX_BLOCKS_PER_ROW); \ #define compute_edge_delta_x2() \ @@ -1079,6 +1107,7 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, and_4x16b(span_shift, left_right_x_16.high, c_0x07); \ shl_variable_4x16b(span_shift, c_0xFFFE, span_shift); \ shr_4x16b(left_right_x_16.high, left_right_x_16.high, 3); \ + min_4x16b(left_right_x_16.high, left_right_x_16.high, c_max_blocks_per_row); \ \ u32 i; \ for(i = 0; i < 4; i++) \ @@ -1087,7 +1116,7 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, span_edge_data[i].num_blocks = left_right_x_16.high.e[i]; \ span_edge_data[i].right_mask = span_shift.e[i]; \ span_edge_data[i].y = y_x4.e[i]; \ - setup_spans_debug_check(span_edge_data[i]); \ + setup_spans_debug_check(psx_gpu, &span_edge_data[i]); \ } \ \ span_edge_data += 4; \ @@ -1117,7 +1146,9 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, \ setup_spans_prologue_b(); \ \ - if(height > 0) \ + if (height > 512) \ + height = 512; \ + if (height > 0) \ { \ y_x4.e[0] = y_a; \ y_x4.e[1] = y_a + 1; \ @@ -1165,7 +1196,9 @@ void compute_all_gradients(psx_gpu_struct *psx_gpu, vertex_struct *a, \ setup_spans_prologue_b(); \ \ - if(height > 0) \ + if (height > 512) \ + height = 512; \ + if (height > 0) \ { \ y_x4.e[0] = y_a; \ y_x4.e[1] = y_a - 1; \ @@ -1355,7 +1388,9 @@ void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a, setup_spans_prologue_b(); - if(height_minor_a > 0) + if (height_minor_a > 512) + height_minor_a = 512; + if (height_minor_a > 0) { y_x4.e[0] = y_a; y_x4.e[1] = y_a - 1; @@ -1397,7 +1432,9 @@ void setup_spans_up_down(psx_gpu_struct *psx_gpu, vertex_struct *v_a, setup_spans_clip(increment, no); } - if(height_minor_b > 0) + if (height_minor_b > 512) + height_minor_b = 512; + if (height_minor_b > 0) { y_x4.e[0] = y_a; y_x4.e[1] = y_a + 1; @@ -3037,6 +3074,7 @@ static void render_triangle_p(psx_gpu_struct *psx_gpu, } } } + assert(psx_gpu->span_edge_data[0].y < 1024u); u32 render_state = flags & (RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND | @@ -3906,17 +3944,9 @@ void setup_sprite_16bpp(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, #ifndef NEON_BUILD -void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, +void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, s32 v, s32 width, s32 height, u32 color) { - if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE | - RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 && - (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0) - { - setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color); - return; - } - u32 right_width = ((width - 1) & 0x7) + 1; u32 right_mask_bits = (0xFF << right_width); u16 *fb_ptr = psx_gpu->vram_out_ptr + (y * 1024) + x; @@ -3984,8 +4014,9 @@ void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, #endif -void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, - s32 u, s32 v, s32 width, s32 height, u32 color) +static void __attribute__((noinline)) +setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, + s32 v, s32 width, s32 height, u32 color) { u32 r = color & 0xFF; u32 g = (color >> 8) & 0xFF; @@ -3999,7 +4030,7 @@ void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, u32 num_width; - if(psx_gpu->num_blocks > MAX_BLOCKS) + if(psx_gpu->num_blocks) { flush_render_block_buffer(psx_gpu); } @@ -4009,7 +4040,7 @@ void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, num_width = width; vram_ptr = (void *)vram_ptr16; - if((long)vram_ptr16 & 2) + if((uintptr_t)vram_ptr16 & 2) { *vram_ptr16 = color_32bpp; vram_ptr = (void *)(vram_ptr16 + 1); @@ -4043,6 +4074,29 @@ void setup_sprite_untextured_simple(psx_gpu_struct *psx_gpu, s32 x, s32 y, } } +void setup_sprite_untextured_512(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, + s32 v, s32 width, s32 height, u32 color); + +void setup_sprite_untextured(psx_gpu_struct *psx_gpu, s32 x, s32 y, s32 u, + s32 v, s32 width, s32 height, u32 color) +{ + if((psx_gpu->render_state & (RENDER_STATE_MASK_EVALUATE | + RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND)) == 0 && + (psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) == 0) + { + setup_sprite_untextured_simple(psx_gpu, x, y, u, v, width, height, color); + return; + } + + while (width > 0) + { + s32 w1 = width > 512 ? 512 : width; + setup_sprite_untextured_512(psx_gpu, x, y, 0, 0, w1, height, color); + x += 512; + width -= 512; + } +} + #define setup_sprite_blocks_switch_textured(texture_mode) \ setup_sprite_##texture_mode \ @@ -4530,6 +4584,7 @@ void render_line(psx_gpu_struct *psx_gpu, vertex_struct *vertexes, u32 flags, if(vertex_a->x >= vertex_b->x) { vertex_swap(vertex_a, vertex_b); + (void)triangle_winding; } x_a = vertex_a->x; @@ -4836,6 +4891,7 @@ void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y, } } +#ifndef PCSX void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y, u32 width, u32 height, u32 pitch) { @@ -4867,7 +4923,7 @@ void render_block_move(psx_gpu_struct *psx_gpu, u32 source_x, u32 source_y, render_block_copy(psx_gpu, psx_gpu->vram_ptr + source_x + (source_y * 1024), dest_x, dest_y, width, height, 1024); } - +#endif void initialize_reciprocal_table(void) { @@ -4923,6 +4979,8 @@ void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram) psx_gpu->texture_page_ptr = psx_gpu->vram_ptr; psx_gpu->clut_ptr = psx_gpu->vram_ptr; + psx_gpu->viewport_start_x = psx_gpu->viewport_start_y = 0; + psx_gpu->viewport_end_x = psx_gpu->viewport_end_y = 0; psx_gpu->mask_msb = 0; psx_gpu->texture_window_x = 0; @@ -4955,7 +5013,10 @@ void initialize_psx_gpu(psx_gpu_struct *psx_gpu, u16 *vram) psx_gpu->primitive_type = PRIMITIVE_TYPE_UNKNOWN; - psx_gpu->enhancement_x_threshold = 256; + psx_gpu->saved_hres = 256; + + // check some offset + psx_gpu->reserved_a[(offsetof(psx_gpu_struct, blocks) == psx_gpu_blocks_offset) - 1] = 0; } u64 get_us(void) @@ -5021,3 +5082,5 @@ void triangle_benchmark(psx_gpu_struct *psx_gpu) #endif #include "psx_gpu_4x.c" + +// vim:ts=2:sw=2:expandtab