The-Homebrew-Cloud/Lime3DS
2
0
Fork 0
mirror of https://github.com/Lime3DS/Lime3DS.git synced 2024-11-14 05:55:13 +01:00
Code Issues Packages Projects Releases Wiki Activity

GPU: Implement blended downscaling for display transfers.

Browse Source
This commit is contained in:
Emmanuel Gil Peyrot 2015-05-22 01:08:30 +02:00
parent 8ee814ec27
commit 641e78bccf
1 changed files with 40 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

67
src/core/hw/gpu.cpp
View File

@ -53,6 +53,29 @@ inline void Read(T &var, const u32 raw_addr) {
var = g_regs[addr / 4]; var = g_regs[addr / 4];
} }
static Math::Vec4<u8> DecodePixel(Regs::PixelFormat input_format, const u8* src_pixel) {
switch (input_format) {
case Regs::PixelFormat::RGBA8:
return Color::DecodeRGBA8(src_pixel);
case Regs::PixelFormat::RGB8:
return Color::DecodeRGB8(src_pixel);
case Regs::PixelFormat::RGB565:
return Color::DecodeRGB565(src_pixel);
case Regs::PixelFormat::RGB5A1:
return Color::DecodeRGB5A1(src_pixel);
case Regs::PixelFormat::RGBA4:
return Color::DecodeRGBA4(src_pixel);
default:
LOG_ERROR(HW_GPU, "Unknown source framebuffer format %x", input_format);
return {0, 0, 0, 0};
}
}
template <typename T> template <typename T>
inline void Write(u32 addr, const T data) { inline void Write(u32 addr, const T data) {
addr -= HW::VADDR_GPU; addr -= HW::VADDR_GPU;
@ -125,6 +148,13 @@ inline void Write(u32 addr, const T data) {
break; break;
} }
if (config.output_tiled &&
(config.scaling == config.ScaleXY || config.scaling == config.ScaleX)) {
LOG_CRITICAL(HW_GPU, "Scaling is only implemented on tiled input");
UNIMPLEMENTED();
break;
}
bool horizontal_scale = config.scaling != config.NoScale; bool horizontal_scale = config.scaling != config.NoScale;
bool vertical_scale = config.scaling == config.ScaleXY; bool vertical_scale = config.scaling == config.ScaleXY;
@ -153,11 +183,9 @@ inline void Write(u32 addr, const T data) {
break; break;
} }
// TODO(Subv): Implement the box filter when scaling is enabled
// right now we're just skipping the extra pixels.
for (u32 y = 0; y < output_height; ++y) { for (u32 y = 0; y < output_height; ++y) {
for (u32 x = 0; x < output_width; ++x) { for (u32 x = 0; x < output_width; ++x) {
Math::Vec4<u8> src_color = { 0, 0, 0, 0 }; Math::Vec4<u8> src_color;
// Calculate the [x,y] position of the input image // Calculate the [x,y] position of the input image
// based on the current output position and the scale // based on the current output position and the scale
@ -193,30 +221,15 @@ inline void Write(u32 addr, const T data) {
} }
const u8* src_pixel = src_pointer + src_offset; const u8* src_pixel = src_pointer + src_offset;
switch (config.input_format) { src_color = DecodePixel(config.input_format, src_pixel);
case Regs::PixelFormat::RGBA8: if (config.scaling == config.ScaleX) {
src_color = Color::DecodeRGBA8(src_pixel); Math::Vec4<u8> pixel = DecodePixel(config.input_format, src_pixel + src_bytes_per_pixel);
break; src_color = ((src_color + pixel) / 2).Cast<u8>();
} else if (config.scaling == config.ScaleXY) {
case Regs::PixelFormat::RGB8: Math::Vec4<u8> pixel1 = DecodePixel(config.input_format, src_pixel + 1 * src_bytes_per_pixel);
src_color = Color::DecodeRGB8(src_pixel); Math::Vec4<u8> pixel2 = DecodePixel(config.input_format, src_pixel + 2 * src_bytes_per_pixel);
break; Math::Vec4<u8> pixel3 = DecodePixel(config.input_format, src_pixel + 3 * src_bytes_per_pixel);
src_color = (((src_color + pixel1) + (pixel2 + pixel3)) / 4).Cast<u8>();
case Regs::PixelFormat::RGB565:
src_color = Color::DecodeRGB565(src_pixel);
break;
case Regs::PixelFormat::RGB5A1:
src_color = Color::DecodeRGB5A1(src_pixel);
break;
case Regs::PixelFormat::RGBA4:
src_color = Color::DecodeRGBA4(src_pixel);
break;
default:
LOG_ERROR(HW_GPU, "Unknown source framebuffer format %x", config.input_format.Value());
break;
} }
u8* dst_pixel = dst_pointer + dst_offset; u8* dst_pixel = dst_pointer + dst_offset;