summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp4
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp47
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.h86
-rw-r--r--src/video_core/textures/decoders.cpp85
-rw-r--r--src/video_core/textures/decoders.h4
-rw-r--r--src/yuzu/debugger/graphics/graphics_surface.cpp6
6 files changed, 41 insertions, 191 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 51e50cbcf..8360feb5d 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -811,9 +811,7 @@ void RasterizerOpenGL::SyncClipCoef() {
void RasterizerOpenGL::SyncCullMode() {
const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
- // TODO(bunnei): Enable the below once more things work - until then, this may hide regressions
- // state.cull.enabled = regs.cull.enabled != 0;
- state.cull.enabled = false;
+ state.cull.enabled = regs.cull.enabled != 0;
if (state.cull.enabled) {
state.cull.front_face = MaxwellToGL::FrontFace(regs.cull.front_face);
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 04ebdbaf9..9fb734b77 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -183,36 +183,49 @@ MathUtil::Rectangle<u32> SurfaceParams::GetRect() const {
return {0, actual_height, width, 0};
}
+/// Returns true if the specified PixelFormat is a BCn format, e.g. DXT or DXN
+static bool IsFormatBCn(PixelFormat format) {
+ switch (format) {
+ case PixelFormat::DXT1:
+ case PixelFormat::DXT23:
+ case PixelFormat::DXT45:
+ case PixelFormat::DXN1:
+ case PixelFormat::DXN2SNORM:
+ case PixelFormat::DXN2UNORM:
+ case PixelFormat::BC7U:
+ return true;
+ }
+ return false;
+}
+
template <bool morton_to_gl, PixelFormat format>
-void MortonCopy(u32 stride, u32 block_height, u32 height, u8* gl_buffer, Tegra::GPUVAddr addr) {
+void MortonCopy(u32 stride, u32 block_height, u32 height, std::vector<u8>& gl_buffer,
+ Tegra::GPUVAddr addr) {
constexpr u32 bytes_per_pixel = SurfaceParams::GetFormatBpp(format) / CHAR_BIT;
constexpr u32 gl_bytes_per_pixel = CachedSurface::GetGLBytesPerPixel(format);
const auto& gpu = Core::System::GetInstance().GPU();
if (morton_to_gl) {
- if (SurfaceParams::GetFormatType(format) == SurfaceType::ColorTexture) {
- auto data = Tegra::Texture::UnswizzleTexture(
- *gpu.memory_manager->GpuToCpuAddress(addr),
- SurfaceParams::TextureFormatFromPixelFormat(format), stride, height, block_height);
- std::memcpy(gl_buffer, data.data(), data.size());
- } else {
- auto data = Tegra::Texture::UnswizzleDepthTexture(
- *gpu.memory_manager->GpuToCpuAddress(addr),
- SurfaceParams::DepthFormatFromPixelFormat(format), stride, height, block_height);
- std::memcpy(gl_buffer, data.data(), data.size());
- }
+ // With the BCn formats (DXT and DXN), each 4x4 tile is swizzled instead of just individual
+ // pixel values.
+ const u32 tile_size{IsFormatBCn(format) ? 4U : 1U};
+ const std::vector<u8> data =
+ Tegra::Texture::UnswizzleTexture(*gpu.memory_manager->GpuToCpuAddress(addr), tile_size,
+ bytes_per_pixel, stride, height, block_height);
+ const size_t size_to_copy{std::min(gl_buffer.size(), data.size())};
+ gl_buffer.assign(data.begin(), data.begin() + size_to_copy);
} else {
// TODO(bunnei): Assumes the default rendering GOB size of 16 (128 lines). We should
// check the configuration for this and perform more generic un/swizzle
LOG_WARNING(Render_OpenGL, "need to use correct swizzle/GOB parameters!");
VideoCore::MortonCopyPixels128(
stride, height, bytes_per_pixel, gl_bytes_per_pixel,
- Memory::GetPointer(*gpu.memory_manager->GpuToCpuAddress(addr)), gl_buffer,
+ Memory::GetPointer(*gpu.memory_manager->GpuToCpuAddress(addr)), gl_buffer.data(),
morton_to_gl);
}
}
-static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
+static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
SurfaceParams::MaxPixelFormat>
morton_to_gl_fns = {
MortonCopy<true, PixelFormat::ABGR8>, MortonCopy<true, PixelFormat::B5G6R5>,
@@ -235,7 +248,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<true, PixelFormat::Z32FS8>,
};
-static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
+static constexpr std::array<void (*)(u32, u32, u32, std::vector<u8>&, Tegra::GPUVAddr),
SurfaceParams::MaxPixelFormat>
gl_to_morton_fns = {
MortonCopy<false, PixelFormat::ABGR8>,
@@ -467,7 +480,7 @@ void CachedSurface::LoadGLBuffer() {
gl_buffer.resize(copy_size);
morton_to_gl_fns[static_cast<size_t>(params.pixel_format)](
- params.width, params.block_height, params.height, gl_buffer.data(), params.addr);
+ params.width, params.block_height, params.height, gl_buffer, params.addr);
} else {
const u8* const texture_src_data_end = texture_src_data + copy_size;
@@ -494,7 +507,7 @@ void CachedSurface::FlushGLBuffer() {
std::memcpy(dst_buffer, gl_buffer.data(), params.size_in_bytes);
} else {
gl_to_morton_fns[static_cast<size_t>(params.pixel_format)](
- params.width, params.block_height, params.height, gl_buffer.data(), params.addr);
+ params.width, params.block_height, params.height, gl_buffer, params.addr);
}
}
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index 8a0ba64e8..829a76dfe 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -348,92 +348,6 @@ struct SurfaceParams {
}
}
- static Tegra::Texture::TextureFormat TextureFormatFromPixelFormat(PixelFormat format) {
- // TODO(Subv): Properly implement this
- switch (format) {
- case PixelFormat::ABGR8:
- case PixelFormat::SRGBA8:
- return Tegra::Texture::TextureFormat::A8R8G8B8;
- case PixelFormat::B5G6R5:
- return Tegra::Texture::TextureFormat::B5G6R5;
- case PixelFormat::A2B10G10R10:
- return Tegra::Texture::TextureFormat::A2B10G10R10;
- case PixelFormat::A1B5G5R5:
- return Tegra::Texture::TextureFormat::A1B5G5R5;
- case PixelFormat::R8:
- return Tegra::Texture::TextureFormat::R8;
- case PixelFormat::G8R8:
- return Tegra::Texture::TextureFormat::G8R8;
- case PixelFormat::RGBA16F:
- return Tegra::Texture::TextureFormat::R16_G16_B16_A16;
- case PixelFormat::R11FG11FB10F:
- return Tegra::Texture::TextureFormat::BF10GF11RF11;
- case PixelFormat::RGBA32UI:
- return Tegra::Texture::TextureFormat::R32_G32_B32_A32;
- case PixelFormat::DXT1:
- return Tegra::Texture::TextureFormat::DXT1;
- case PixelFormat::DXT23:
- return Tegra::Texture::TextureFormat::DXT23;
- case PixelFormat::DXT45:
- return Tegra::Texture::TextureFormat::DXT45;
- case PixelFormat::DXN1:
- return Tegra::Texture::TextureFormat::DXN1;
- case PixelFormat::DXN2UNORM:
- case PixelFormat::DXN2SNORM:
- return Tegra::Texture::TextureFormat::DXN2;
- case PixelFormat::BC7U:
- return Tegra::Texture::TextureFormat::BC7U;
- case PixelFormat::ASTC_2D_4X4:
- return Tegra::Texture::TextureFormat::ASTC_2D_4X4;
- case PixelFormat::BGRA8:
- // TODO(bunnei): This is fine for unswizzling (since we just need the right component
- // sizes), but could be a bug if we used this function in different ways.
- return Tegra::Texture::TextureFormat::A8R8G8B8;
- case PixelFormat::RGBA32F:
- return Tegra::Texture::TextureFormat::R32_G32_B32_A32;
- case PixelFormat::RGB32F:
- return Tegra::Texture::TextureFormat::R32_G32_B32;
- case PixelFormat::RG32F:
- return Tegra::Texture::TextureFormat::R32_G32;
- case PixelFormat::R32F:
- return Tegra::Texture::TextureFormat::R32;
- case PixelFormat::R16F:
- case PixelFormat::R16UNORM:
- return Tegra::Texture::TextureFormat::R16;
- case PixelFormat::Z32F:
- return Tegra::Texture::TextureFormat::ZF32;
- case PixelFormat::Z24S8:
- return Tegra::Texture::TextureFormat::Z24S8;
- case PixelFormat::RG16F:
- case PixelFormat::RG16:
- case PixelFormat::RG16UI:
- case PixelFormat::RG16I:
- case PixelFormat::RG16S:
- return Tegra::Texture::TextureFormat::R16_G16;
- default:
- LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
- UNREACHABLE();
- }
- }
-
- static Tegra::DepthFormat DepthFormatFromPixelFormat(PixelFormat format) {
- switch (format) {
- case PixelFormat::S8Z24:
- return Tegra::DepthFormat::S8_Z24_UNORM;
- case PixelFormat::Z24S8:
- return Tegra::DepthFormat::Z24_S8_UNORM;
- case PixelFormat::Z32F:
- return Tegra::DepthFormat::Z32_FLOAT;
- case PixelFormat::Z16:
- return Tegra::DepthFormat::Z16_UNORM;
- case PixelFormat::Z32FS8:
- return Tegra::DepthFormat::Z32_S8_X24_FLOAT;
- default:
- LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
- UNREACHABLE();
- }
- }
-
static ComponentType ComponentTypeFromTexture(Tegra::Texture::ComponentType type) {
// TODO(Subv): Implement more component types
switch (type) {
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 7ea66584c..70746a34e 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -86,88 +86,11 @@ u32 BytesPerPixel(TextureFormat format) {
}
}
-static u32 DepthBytesPerPixel(DepthFormat format) {
- switch (format) {
- case DepthFormat::Z16_UNORM:
- return 2;
- case DepthFormat::S8_Z24_UNORM:
- case DepthFormat::Z24_S8_UNORM:
- case DepthFormat::Z32_FLOAT:
- return 4;
- case DepthFormat::Z32_S8_X24_FLOAT:
- return 8;
- default:
- UNIMPLEMENTED_MSG("Format not implemented");
- break;
- }
-}
-
-std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width, u32 height,
- u32 block_height) {
- u8* data = Memory::GetPointer(address);
- u32 bytes_per_pixel = BytesPerPixel(format);
-
- std::vector<u8> unswizzled_data(width * height * bytes_per_pixel);
-
- switch (format) {
- case TextureFormat::DXT1:
- case TextureFormat::DXT23:
- case TextureFormat::DXT45:
- case TextureFormat::DXN1:
- case TextureFormat::DXN2:
- case TextureFormat::BC7U:
- // In the DXT and DXN formats, each 4x4 tile is swizzled instead of just individual pixel
- // values.
- CopySwizzledData(width / 4, height / 4, bytes_per_pixel, bytes_per_pixel, data,
- unswizzled_data.data(), true, block_height);
- break;
- case TextureFormat::A8R8G8B8:
- case TextureFormat::A2B10G10R10:
- case TextureFormat::A1B5G5R5:
- case TextureFormat::B5G6R5:
- case TextureFormat::R8:
- case TextureFormat::G8R8:
- case TextureFormat::R16_G16_B16_A16:
- case TextureFormat::R32_G32_B32_A32:
- case TextureFormat::R32_G32:
- case TextureFormat::R32:
- case TextureFormat::R16:
- case TextureFormat::R16_G16:
- case TextureFormat::BF10GF11RF11:
- case TextureFormat::ASTC_2D_4X4:
- case TextureFormat::R32_G32_B32:
- CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
- unswizzled_data.data(), true, block_height);
- break;
- default:
- UNIMPLEMENTED_MSG("Format not implemented");
- break;
- }
-
- return unswizzled_data;
-}
-
-std::vector<u8> UnswizzleDepthTexture(VAddr address, DepthFormat format, u32 width, u32 height,
- u32 block_height) {
- u8* data = Memory::GetPointer(address);
- u32 bytes_per_pixel = DepthBytesPerPixel(format);
-
+std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width,
+ u32 height, u32 block_height) {
std::vector<u8> unswizzled_data(width * height * bytes_per_pixel);
-
- switch (format) {
- case DepthFormat::Z16_UNORM:
- case DepthFormat::S8_Z24_UNORM:
- case DepthFormat::Z24_S8_UNORM:
- case DepthFormat::Z32_FLOAT:
- case DepthFormat::Z32_S8_X24_FLOAT:
- CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
- unswizzled_data.data(), true, block_height);
- break;
- default:
- UNIMPLEMENTED_MSG("Format not implemented");
- break;
- }
-
+ CopySwizzledData(width / tile_size, height / tile_size, bytes_per_pixel, bytes_per_pixel,
+ Memory::GetPointer(address), unswizzled_data.data(), true, block_height);
return unswizzled_data;
}
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index 73a4924d1..1f7b731be 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -13,8 +13,8 @@ namespace Tegra::Texture {
/**
* Unswizzles a swizzled texture without changing its format.
*/
-std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width, u32 height,
- u32 block_height = TICEntry::DefaultBlockHeight);
+std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size, u32 bytes_per_pixel, u32 width,
+ u32 height, u32 block_height = TICEntry::DefaultBlockHeight);
/**
* Unswizzles a swizzled depth texture without changing its format.
diff --git a/src/yuzu/debugger/graphics/graphics_surface.cpp b/src/yuzu/debugger/graphics/graphics_surface.cpp
index 3f7103ab9..e037223c2 100644
--- a/src/yuzu/debugger/graphics/graphics_surface.cpp
+++ b/src/yuzu/debugger/graphics/graphics_surface.cpp
@@ -383,8 +383,10 @@ void GraphicsSurfaceWidget::OnUpdate() {
QImage decoded_image(surface_width, surface_height, QImage::Format_ARGB32);
boost::optional<VAddr> address = gpu.memory_manager->GpuToCpuAddress(surface_address);
- auto unswizzled_data =
- Tegra::Texture::UnswizzleTexture(*address, surface_format, surface_width, surface_height);
+ // TODO(bunnei): Will not work with BCn formats that swizzle 4x4 tiles.
+ // Needs to be fixed if we plan to use this feature more, otherwise we may remove it.
+ auto unswizzled_data = Tegra::Texture::UnswizzleTexture(
+ *address, 1, Tegra::Texture::BytesPerPixel(surface_format), surface_width, surface_height);
auto texture_data = Tegra::Texture::DecodeTexture(unswizzled_data, surface_format,
surface_width, surface_height);
generated by cgit v1.2.3 (git 2.25.1) at 2024-07-15 17:35:21 +0200