Merge pull request #601 from Subv/rgba32_ui

GPU: Implement the RGBA32_UINT rendertarget format.
This commit is contained in:
bunnei 2018-07-01 03:22:38 -04:00 committed by GitHub
commit b611d852db
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 48 additions and 25 deletions

View file

@ -16,6 +16,7 @@ namespace Tegra {
enum class RenderTargetFormat : u32 { enum class RenderTargetFormat : u32 {
NONE = 0x0, NONE = 0x0,
RGBA32_FLOAT = 0xC0, RGBA32_FLOAT = 0xC0,
RGBA32_UINT = 0xC2,
RGBA16_FLOAT = 0xCA, RGBA16_FLOAT = 0xCA,
RGB10_A2_UNORM = 0xD1, RGB10_A2_UNORM = 0xD1,
RGBA8_UNORM = 0xD5, RGBA8_UNORM = 0xD5,

View file

@ -27,6 +27,7 @@ struct FormatTuple {
GLint internal_format; GLint internal_format;
GLenum format; GLenum format;
GLenum type; GLenum type;
ComponentType component_type;
bool compressed; bool compressed;
}; };
@ -65,29 +66,33 @@ struct FormatTuple {
} }
static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_format_tuples = {{
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false}, // ABGR8 {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, ComponentType::UNorm, false}, // ABGR8
{GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false}, // B5G6R5 {GL_RGB, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, ComponentType::UNorm, false}, // B5G6R5
{GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false}, // A2B10G10R10 {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, ComponentType::UNorm,
{GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, false}, // A1B5G5R5 false}, // A2B10G10R10
{GL_R8, GL_RED, GL_UNSIGNED_BYTE, false}, // R8 {GL_RGB5_A1, GL_RGBA, GL_UNSIGNED_SHORT_1_5_5_5_REV, ComponentType::UNorm, false}, // A1B5G5R5
{GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false}, // RGBA16F {GL_R8, GL_RED, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // R8
{GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false}, // R11FG11FB10F {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, ComponentType::Float, false}, // RGBA16F
{GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT1 {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, ComponentType::Float,
{GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT23 false}, // R11FG11FB10F
{GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, true}, // DXT45 {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, ComponentType::UInt, false}, // RGBA32UI
{GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, true}, // DXN1 {GL_COMPRESSED_RGB_S3TC_DXT1_EXT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false}, // ASTC_2D_4X4 true}, // DXT1
{GL_COMPRESSED_RGBA_S3TC_DXT3_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT23
{GL_COMPRESSED_RGBA_S3TC_DXT5_EXT, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm,
true}, // DXT45
{GL_COMPRESSED_RED_RGTC1, GL_RED, GL_UNSIGNED_INT_8_8_8_8, ComponentType::UNorm, true}, // DXN1
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
}}; }};
static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) { static const FormatTuple& GetFormatTuple(PixelFormat pixel_format, ComponentType component_type) {
const SurfaceType type = SurfaceParams::GetFormatType(pixel_format); const SurfaceType type = SurfaceParams::GetFormatType(pixel_format);
if (type == SurfaceType::ColorTexture) { if (type == SurfaceType::ColorTexture) {
ASSERT(static_cast<size_t>(pixel_format) < tex_format_tuples.size()); ASSERT(static_cast<size_t>(pixel_format) < tex_format_tuples.size());
// For now only UNORM components are supported, or either R11FG11FB10F or RGBA16F which auto& format = tex_format_tuples[static_cast<unsigned int>(pixel_format)];
// are type FLOAT ASSERT(component_type == format.component_type);
ASSERT(component_type == ComponentType::UNorm || pixel_format == PixelFormat::RGBA16F || return format;
pixel_format == PixelFormat::R11FG11FB10F);
return tex_format_tuples[static_cast<unsigned int>(pixel_format)];
} else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) { } else if (type == SurfaceType::Depth || type == SurfaceType::DepthStencil) {
// TODO(Subv): Implement depth formats // TODO(Subv): Implement depth formats
ASSERT_MSG(false, "Unimplemented"); ASSERT_MSG(false, "Unimplemented");
@ -166,9 +171,10 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<true, PixelFormat::ABGR8>, MortonCopy<true, PixelFormat::B5G6R5>, MortonCopy<true, PixelFormat::ABGR8>, MortonCopy<true, PixelFormat::B5G6R5>,
MortonCopy<true, PixelFormat::A2B10G10R10>, MortonCopy<true, PixelFormat::A1B5G5R5>, MortonCopy<true, PixelFormat::A2B10G10R10>, MortonCopy<true, PixelFormat::A1B5G5R5>,
MortonCopy<true, PixelFormat::R8>, MortonCopy<true, PixelFormat::RGBA16F>, MortonCopy<true, PixelFormat::R8>, MortonCopy<true, PixelFormat::RGBA16F>,
MortonCopy<true, PixelFormat::R11FG11FB10F>, MortonCopy<true, PixelFormat::DXT1>, MortonCopy<true, PixelFormat::R11FG11FB10F>, MortonCopy<true, PixelFormat::RGBA32UI>,
MortonCopy<true, PixelFormat::DXT23>, MortonCopy<true, PixelFormat::DXT45>, MortonCopy<true, PixelFormat::DXT1>, MortonCopy<true, PixelFormat::DXT23>,
MortonCopy<true, PixelFormat::DXN1>, MortonCopy<true, PixelFormat::ASTC_2D_4X4>, MortonCopy<true, PixelFormat::DXT45>, MortonCopy<true, PixelFormat::DXN1>,
MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
}; };
static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr), static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
@ -181,6 +187,7 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<false, PixelFormat::R8>, MortonCopy<false, PixelFormat::R8>,
MortonCopy<false, PixelFormat::RGBA16F>, MortonCopy<false, PixelFormat::RGBA16F>,
MortonCopy<false, PixelFormat::R11FG11FB10F>, MortonCopy<false, PixelFormat::R11FG11FB10F>,
MortonCopy<false, PixelFormat::RGBA32UI>,
// TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1 formats is not yet supported // TODO(Subv): Swizzling the DXT1/DXT23/DXT45/DXN1 formats is not yet supported
nullptr, nullptr,
nullptr, nullptr,

View file

@ -30,11 +30,12 @@ struct SurfaceParams {
R8 = 4, R8 = 4,
RGBA16F = 5, RGBA16F = 5,
R11FG11FB10F = 6, R11FG11FB10F = 6,
DXT1 = 7, RGBA32UI = 7,
DXT23 = 8, DXT1 = 8,
DXT45 = 9, DXT23 = 9,
DXN1 = 10, // This is also known as BC4 DXT45 = 10,
ASTC_2D_4X4 = 11, DXN1 = 11, // This is also known as BC4
ASTC_2D_4X4 = 12,
Max, Max,
Invalid = 255, Invalid = 255,
@ -77,6 +78,7 @@ struct SurfaceParams {
1, // R8 1, // R8
1, // RGBA16F 1, // RGBA16F
1, // R11FG11FB10F 1, // R11FG11FB10F
1, // RGBA32UI
4, // DXT1 4, // DXT1
4, // DXT23 4, // DXT23
4, // DXT45 4, // DXT45
@ -100,6 +102,7 @@ struct SurfaceParams {
8, // R8 8, // R8
64, // RGBA16F 64, // RGBA16F
32, // R11FG11FB10F 32, // R11FG11FB10F
128, // RGBA32UI
64, // DXT1 64, // DXT1
128, // DXT23 128, // DXT23
128, // DXT45 128, // DXT45
@ -125,6 +128,8 @@ struct SurfaceParams {
return PixelFormat::RGBA16F; return PixelFormat::RGBA16F;
case Tegra::RenderTargetFormat::R11G11B10_FLOAT: case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
return PixelFormat::R11FG11FB10F; return PixelFormat::R11FG11FB10F;
case Tegra::RenderTargetFormat::RGBA32_UINT:
return PixelFormat::RGBA32UI;
default: default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE(); UNREACHABLE();
@ -148,6 +153,8 @@ struct SurfaceParams {
return PixelFormat::RGBA16F; return PixelFormat::RGBA16F;
case Tegra::Texture::TextureFormat::BF10GF11RF11: case Tegra::Texture::TextureFormat::BF10GF11RF11:
return PixelFormat::R11FG11FB10F; return PixelFormat::R11FG11FB10F;
case Tegra::Texture::TextureFormat::R32_G32_B32_A32:
return PixelFormat::RGBA32UI;
case Tegra::Texture::TextureFormat::DXT1: case Tegra::Texture::TextureFormat::DXT1:
return PixelFormat::DXT1; return PixelFormat::DXT1;
case Tegra::Texture::TextureFormat::DXT23: case Tegra::Texture::TextureFormat::DXT23:
@ -181,6 +188,8 @@ struct SurfaceParams {
return Tegra::Texture::TextureFormat::R16_G16_B16_A16; return Tegra::Texture::TextureFormat::R16_G16_B16_A16;
case PixelFormat::R11FG11FB10F: case PixelFormat::R11FG11FB10F:
return Tegra::Texture::TextureFormat::BF10GF11RF11; return Tegra::Texture::TextureFormat::BF10GF11RF11;
case PixelFormat::RGBA32UI:
return Tegra::Texture::TextureFormat::R32_G32_B32_A32;
case PixelFormat::DXT1: case PixelFormat::DXT1:
return Tegra::Texture::TextureFormat::DXT1; return Tegra::Texture::TextureFormat::DXT1;
case PixelFormat::DXT23: case PixelFormat::DXT23:
@ -217,6 +226,8 @@ struct SurfaceParams {
case Tegra::RenderTargetFormat::RGBA16_FLOAT: case Tegra::RenderTargetFormat::RGBA16_FLOAT:
case Tegra::RenderTargetFormat::R11G11B10_FLOAT: case Tegra::RenderTargetFormat::R11G11B10_FLOAT:
return ComponentType::Float; return ComponentType::Float;
case Tegra::RenderTargetFormat::RGBA32_UINT:
return ComponentType::UInt;
default: default:
NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format)); NGLOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE(); UNREACHABLE();

View file

@ -65,6 +65,8 @@ u32 BytesPerPixel(TextureFormat format) {
return 1; return 1;
case TextureFormat::R16_G16_B16_A16: case TextureFormat::R16_G16_B16_A16:
return 8; return 8;
case TextureFormat::R32_G32_B32_A32:
return 16;
default: default:
UNIMPLEMENTED_MSG("Format not implemented"); UNIMPLEMENTED_MSG("Format not implemented");
break; break;
@ -94,6 +96,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
case TextureFormat::B5G6R5: case TextureFormat::B5G6R5:
case TextureFormat::R8: case TextureFormat::R8:
case TextureFormat::R16_G16_B16_A16: case TextureFormat::R16_G16_B16_A16:
case TextureFormat::R32_G32_B32_A32:
case TextureFormat::BF10GF11RF11: case TextureFormat::BF10GF11RF11:
case TextureFormat::ASTC_2D_4X4: case TextureFormat::ASTC_2D_4X4:
CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data, CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
@ -124,6 +127,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
case TextureFormat::B5G6R5: case TextureFormat::B5G6R5:
case TextureFormat::R8: case TextureFormat::R8:
case TextureFormat::BF10GF11RF11: case TextureFormat::BF10GF11RF11:
case TextureFormat::R32_G32_B32_A32:
// TODO(Subv): For the time being just forward the same data without any decoding. // TODO(Subv): For the time being just forward the same data without any decoding.
rgba_data = texture_data; rgba_data = texture_data;
break; break;