u32 texture_cache_loads = 0;
u32 false_modulated_blocks = 0;
-u32 reciprocal_table[512];
+/* double size for enhancement */
+u32 reciprocal_table[512 * 2];
typedef s32 fixed_type;
return mask;
}
+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;
+ u8 *texture_page_ptr;
+ u16 *vram_ptr;
+ u32 texel_block;
+ u32 sub_x, sub_y;
+
+ psx_gpu->dirty_textures_8bpp_mask |= mask;
+ psx_gpu->dirty_textures_8bpp_alternate_mask |= mask;
+
+ if ((psx_gpu->dirty_textures_4bpp_mask & mask) == 0 &&
+ (x1 & 3) == 0 && (y1 & 15) == 0 && x2 - x1 < 4 && y2 - y1 < 16)
+ {
+ texture_page = ((x1 / 64) & 15) + (y1 / 256) * 16;
+ texture_page_ptr = psx_gpu->texture_4bpp_cache[texture_page];
+ texture_page_ptr += (x1 / 4 & 15) * 16*16 + (y1 / 16 & 15) * 16*16*16;
+ vram_ptr = psx_gpu->vram_ptr + x1 + y1 * 1024;
+ sub_x = 4;
+ sub_y = 16;
+
+ while(sub_y)
+ {
+ while(sub_x)
+ {
+ texel_block = *vram_ptr;
+
+ texture_page_ptr[0] = texel_block & 0xF;
+ texture_page_ptr[1] = (texel_block >> 4) & 0xF;
+ texture_page_ptr[2] = (texel_block >> 8) & 0xF;
+ texture_page_ptr[3] = texel_block >> 12;
+
+ vram_ptr++;
+ texture_page_ptr += 4;
+
+ sub_x--;
+ }
+
+ vram_ptr -= 4;
+ sub_x = 4;
+
+ sub_y--;
+ vram_ptr += 1024;
+ }
+ }
+ else
+ {
+ psx_gpu->dirty_textures_4bpp_mask |= mask;
+ }
+}
void update_texture_8bpp_cache_slice(psx_gpu_struct *psx_gpu,
u32 texture_page);
void flush_render_block_buffer(psx_gpu_struct *psx_gpu)
{
+ if((psx_gpu->render_mode & RENDER_INTERLACE_ENABLED) &&
+ (psx_gpu->primitive_type == PRIMITIVE_TYPE_SPRITE))
+ {
+ u32 num_blocks_dest = 0;
+ block_struct *block_src = psx_gpu->blocks;
+ block_struct *block_dest = psx_gpu->blocks;
+
+ u16 *vram_ptr = psx_gpu->vram_ptr;
+ u32 i;
+
+ if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
+ {
+ for(i = 0; i < psx_gpu->num_blocks; i++)
+ {
+ u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
+ if(fb_offset & (1 << 11))
+ {
+ *block_dest = *block_src;
+ num_blocks_dest++;
+ block_dest++;
+ }
+ block_src++;
+ }
+ }
+ else
+ {
+ for(i = 0; i < psx_gpu->num_blocks; i++)
+ {
+ u32 fb_offset = (u32)((u8 *)block_src->fb_ptr - (u8 *)vram_ptr);
+ if((fb_offset & (1 << 11)) == 0)
+ {
+ *block_dest = *block_src;
+ num_blocks_dest++;
+ block_dest++;
+ }
+ block_src++;
+ }
+ }
+
+ psx_gpu->num_blocks = num_blocks_dest;
+ }
+
if(psx_gpu->num_blocks)
{
render_block_handler_struct *render_block_handler =
\
dup_2x32b(edge_shifts, edge_shift); \
sub_2x32b(heights_b, heights, c_0x01); \
- shr_2x32b(height_reciprocals, edge_shifts, 12); \
+ shr_2x32b(height_reciprocals, edge_shifts, 10); \
\
mla_2x32b(heights_b, x_starts, heights); \
bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
sub_2x32b(widths, x_ends, x_starts); \
width_alt = x_c - start_c; \
\
- shr_2x32b(height_reciprocals, edge_shifts, 12); \
- height_reciprocal_alt = edge_shift_alt >> 12; \
+ shr_2x32b(height_reciprocals, edge_shifts, 10); \
+ height_reciprocal_alt = edge_shift_alt >> 10; \
\
bic_immediate_4x16b(vector_cast(vec_4x16u, edge_shifts), 0xE0); \
edge_shift_alt &= 0x1F; \
if(span_num_blocks) \
{ \
y = span_edge_data->y; \
- fb_ptr = psx_gpu->vram_ptr + span_edge_data->left_x + (y * 1024); \
+ fb_ptr = psx_gpu->vram_out_ptr + span_edge_data->left_x + (y * 1024); \
\
setup_blocks_span_initialize_##shading##_##texturing(); \
setup_blocks_span_initialize_##dithering(texturing); \
(triangle_y_direction_##direction_c << 4) | \
(triangle_winding_##winding << 6)) \
-void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
- u32 flags)
+static int prepare_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
+ vertex_struct *vertexes_out[3])
{
s32 y_top, y_bottom;
s32 triangle_area;
#ifdef PROFILE
trivial_rejects++;
#endif
- return;
+ return 0;
}
if(b->y < a->y)
#ifdef PROFILE
trivial_rejects++;
#endif
- return;
+ return 0;
}
if(triangle_area < 0)
vertex_swap(a, b);
}
- if((c->x - a->x) >= 1024)
+ if((c->x - psx_gpu->offset_x) >= 1024 || (c->x - a->x) >= 1024)
{
#ifdef PROFILE
trivial_rejects++;
#endif
- return;
+ return 0;
}
if(invalidate_texture_cache_region_viewport(psx_gpu, a->x, y_top, c->x,
#ifdef PROFILE
trivial_rejects++;
#endif
- return;
+ return 0;
}
- psx_gpu->num_spans = 0;
psx_gpu->triangle_area = triangle_area;
psx_gpu->triangle_winding = triangle_winding;
+ vertexes_out[0] = a;
+ vertexes_out[1] = b;
+ vertexes_out[2] = c;
+
+ return 1;
+}
+
+static void render_triangle_p(psx_gpu_struct *psx_gpu,
+ vertex_struct *vertex_ptrs[3], u32 flags)
+{
+ psx_gpu->num_spans = 0;
+
+ vertex_struct *a = vertex_ptrs[0];
+ vertex_struct *b = vertex_ptrs[1];
+ vertex_struct *c = vertex_ptrs[2];
+
s32 y_delta_a = b->y - a->y;
s32 y_delta_b = c->y - b->y;
s32 y_delta_c = c->y - a->y;
compute_all_gradients(psx_gpu, a, b, c);
switch(y_direction_a | (y_direction_b << 2) | (y_direction_c << 4) |
- (triangle_winding << 6))
+ (psx_gpu->triangle_winding << 6))
{
triangle_case(up, up, up, negative):
triangle_case(up, up, flat, negative):
spans += psx_gpu->num_spans;
#endif
+ if(unlikely(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED))
+ {
+ u32 i;
+
+ if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
+ {
+ for(i = 0; i < psx_gpu->num_spans; i++)
+ {
+ if((psx_gpu->span_edge_data[i].y & 1) == 0)
+ psx_gpu->span_edge_data[i].num_blocks = 0;
+ }
+ }
+ else
+ {
+ for(i = 0; i < psx_gpu->num_spans; i++)
+ {
+ if(psx_gpu->span_edge_data[i].y & 1)
+ psx_gpu->span_edge_data[i].num_blocks = 0;
+ }
+ }
+ }
+ if(psx_gpu->render_mode & RENDER_DOUBLE_MODE)
+ {
+ u32 i;
+ for(i = 0; i < psx_gpu->num_spans; i++)
+ {
+ psx_gpu->span_edge_data[i].y *= 2;
+ }
+ }
+
u32 render_state = flags &
(RENDER_FLAGS_MODULATE_TEXELS | RENDER_FLAGS_BLEND |
RENDER_FLAGS_TEXTURE_MAP | RENDER_FLAGS_SHADE);
(psx_gpu);
}
+void render_triangle(psx_gpu_struct *psx_gpu, vertex_struct *vertexes,
+ u32 flags)
+{
+ vertex_struct *vertex_ptrs[3];
+ if (prepare_triangle(psx_gpu, vertexes, vertex_ptrs))
+ render_triangle_p(psx_gpu, vertex_ptrs, flags);
+}
+
void texture_sprite_blocks_8bpp(psx_gpu_struct *psx_gpu);
flags &= ~RENDER_FLAGS_TEXTURE_MAP;
- vram_ptr = psx_gpu->vram_ptr + (y_a * 1024) + x_a;
+ vram_ptr = psx_gpu->vram_out_ptr + (y_a * 1024) + x_a;
control_mask = 0x0;
if((width == 0) || (height == 0))
return;
- flush_render_block_buffer(psx_gpu);
invalidate_texture_cache_region(psx_gpu, x, y, x + width - 1, y + height - 1);
-#ifndef NEON_BUILD
u32 r = color & 0xFF;
u32 g = (color >> 8) & 0xFF;
u32 b = (color >> 16) & 0xFF;
- u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10);
+ u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
+ psx_gpu->mask_msb;
+ u32 color_32bpp = color_16bpp | (color_16bpp << 16);
- u16 *vram_ptr = psx_gpu->vram_ptr + x + (y * 1024);
- u32 draw_x, draw_y;
+ u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 1024));
- for(draw_y = 0; draw_y < height; draw_y++)
+ u32 pitch = 512 - (width / 2);
+ u32 num_width;
+
+ if(psx_gpu->render_mode & RENDER_INTERLACE_ENABLED)
{
- for(draw_x = 0; draw_x < width; draw_x++)
+ pitch += 512;
+ height /= 2;
+
+ if(psx_gpu->render_mode & RENDER_INTERLACE_ODD)
+ vram_ptr += 512;
+ }
+
+ while(height)
+ {
+ num_width = width;
+ while(num_width)
{
- vram_ptr[draw_x] = color_16bpp;
+ vram_ptr[0] = color_32bpp;
+ vram_ptr[1] = color_32bpp;
+ vram_ptr[2] = color_32bpp;
+ vram_ptr[3] = color_32bpp;
+ vram_ptr[4] = color_32bpp;
+ vram_ptr[5] = color_32bpp;
+ vram_ptr[6] = color_32bpp;
+ vram_ptr[7] = color_32bpp;
+
+ vram_ptr += 8;
+ num_width -= 16;
}
- vram_ptr += 1024;
+ vram_ptr += pitch;
+ height--;
}
-#else
- void render_block_fill_body(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
- u32 width, u32 height);
+}
- render_block_fill_body(psx_gpu, color, x, y, width, height);
-#endif
+void render_block_fill_enh(psx_gpu_struct *psx_gpu, u32 color, u32 x, u32 y,
+ u32 width, u32 height)
+{
+ if((width == 0) || (height == 0))
+ return;
+
+ u32 r = color & 0xFF;
+ u32 g = (color >> 8) & 0xFF;
+ u32 b = (color >> 16) & 0xFF;
+ u32 color_16bpp = (r >> 3) | ((g >> 3) << 5) | ((b >> 3) << 10) |
+ psx_gpu->mask_msb;
+ u32 color_32bpp = color_16bpp | (color_16bpp << 16);
+
+ u32 *vram_ptr = (u32 *)(psx_gpu->vram_out_ptr + x + (y * 2048));
+
+ u32 pitch = 2048 / 2 - (width / 2);
+ u32 num_width;
+
+ while(height)
+ {
+ num_width = width;
+ while(num_width)
+ {
+ vram_ptr[0] = color_32bpp;
+ vram_ptr[1] = color_32bpp;
+ vram_ptr[2] = color_32bpp;
+ vram_ptr[3] = color_32bpp;
+ vram_ptr[4] = color_32bpp;
+ vram_ptr[5] = color_32bpp;
+ vram_ptr[6] = color_32bpp;
+ vram_ptr[7] = color_32bpp;
+
+ vram_ptr += 8;
+ num_width -= 16;
+ }
+
+ vram_ptr += pitch;
+ height--;
+ }
}
void render_block_copy(psx_gpu_struct *psx_gpu, u16 *source, u32 x, u32 y,
u32 height_reciprocal;
s32 shift;
- for(height = 1; height < 512; height++)
+ for(height = 1; height < sizeof(reciprocal_table)
+ / sizeof(reciprocal_table[0]); height++)
{
shift = __builtin_clz(height);
height_normalized = height << shift;
- height_reciprocal = ((1ULL << 50) + (height_normalized - 1)) /
+ height_reciprocal = ((1ULL << 52) + (height_normalized - 1)) /
height_normalized;
- shift = 32 - (50 - shift);
+ shift = 32 - (52 - shift);
- reciprocal_table[height] = (height_reciprocal << 12) | shift;
+ reciprocal_table[height] = (height_reciprocal << 10) | shift;
}
}
psx_gpu->num_blocks = 0;
psx_gpu->vram_ptr = vram;
+ psx_gpu->vram_out_ptr = vram;
psx_gpu->texture_page_base = psx_gpu->vram_ptr;
psx_gpu->texture_page_ptr = psx_gpu->vram_ptr;
psx_gpu->texture_mask_width = 0xFF;
psx_gpu->texture_mask_height = 0xFF;
+ psx_gpu->render_mode = 0;
+
memset(psx_gpu->vram_ptr, 0, sizeof(u16) * 1024 * 512);
initialize_reciprocal_table();