Compare commits

..

1 Commits

Author SHA1 Message Date
Subv
4cc1e180ec GPU: Implemented the R16 and R16F texture formats. 2018-07-24 13:39:16 -05:00
5 changed files with 48 additions and 18 deletions

View File

@@ -5,7 +5,6 @@
#include <algorithm>
#include <memory>
#include <string>
#include <string_view>
#include <tuple>
#include <utility>
#include <glad/glad.h>
@@ -38,6 +37,11 @@ MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255));
MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
RasterizerOpenGL::RasterizerOpenGL() {
has_ARB_buffer_storage = false;
has_ARB_direct_state_access = false;
has_ARB_separate_shader_objects = false;
has_ARB_vertex_attrib_binding = false;
// Create sampler objects
for (size_t i = 0; i < texture_samplers.size(); ++i) {
texture_samplers[i].Create();
@@ -55,8 +59,7 @@ RasterizerOpenGL::RasterizerOpenGL() {
GLint ext_num;
glGetIntegerv(GL_NUM_EXTENSIONS, &ext_num);
for (GLint i = 0; i < ext_num; i++) {
const std::string_view extension{
reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, i))};
std::string extension{reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, i))};
if (extension == "GL_ARB_buffer_storage") {
has_ARB_buffer_storage = true;
@@ -107,6 +110,8 @@ RasterizerOpenGL::RasterizerOpenGL() {
glBindBufferBase(GL_UNIFORM_BUFFER, index, buffer.handle);
}
accelerate_draw = AccelDraw::Disabled;
glEnable(GL_BLEND);
LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
@@ -689,12 +694,10 @@ u32 RasterizerOpenGL::SetupConstBuffers(Maxwell::ShaderStage stage, GLuint progr
glBindBuffer(GL_UNIFORM_BUFFER, 0);
// Now configure the bindpoint of the buffer inside the shader
const std::string buffer_name = used_buffer.GetName();
const GLuint index =
glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str());
if (index != GL_INVALID_INDEX) {
std::string buffer_name = used_buffer.GetName();
GLuint index = glGetProgramResourceIndex(program, GL_UNIFORM_BLOCK, buffer_name.c_str());
if (index != -1)
glUniformBlockBinding(program, index, buffer_draw_state.bindpoint);
}
}
state.Apply();

View File

@@ -135,10 +135,10 @@ private:
/// Syncs the blend state to match the guest state
void SyncBlendState();
bool has_ARB_buffer_storage = false;
bool has_ARB_direct_state_access = false;
bool has_ARB_separate_shader_objects = false;
bool has_ARB_vertex_attrib_binding = false;
bool has_ARB_buffer_storage;
bool has_ARB_direct_state_access;
bool has_ARB_separate_shader_objects;
bool has_ARB_vertex_attrib_binding;
OpenGLState state;
@@ -167,5 +167,5 @@ private:
void SetupShaders(u8* buffer_ptr, GLintptr buffer_offset);
enum class AccelDraw { Disabled, Arrays, Indexed };
AccelDraw accelerate_draw = AccelDraw::Disabled;
AccelDraw accelerate_draw;
};

View File

@@ -111,6 +111,8 @@ static constexpr std::array<FormatTuple, SurfaceParams::MaxPixelFormat> tex_form
{GL_RGBA32F, GL_RGBA, GL_FLOAT, ComponentType::Float, false}, // RGBA32F
{GL_RG32F, GL_RG, GL_FLOAT, ComponentType::Float, false}, // RG32F
{GL_R32F, GL_RED, GL_FLOAT, ComponentType::Float, false}, // R32F
{GL_R16F, GL_RED, GL_HALF_FLOAT, ComponentType::Float, false}, // R16F
{GL_R16, GL_RED, GL_UNSIGNED_SHORT, ComponentType::UNorm, false}, // R16UNORM
// DepthStencil formats
{GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8, ComponentType::UNorm,
@@ -204,7 +206,8 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<true, PixelFormat::BC7U>, MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
MortonCopy<true, PixelFormat::G8R8>, MortonCopy<true, PixelFormat::BGRA8>,
MortonCopy<true, PixelFormat::RGBA32F>, MortonCopy<true, PixelFormat::RG32F>,
MortonCopy<true, PixelFormat::R32F>, MortonCopy<true, PixelFormat::Z24S8>,
MortonCopy<true, PixelFormat::R32F>, MortonCopy<true, PixelFormat::R16F>,
MortonCopy<true, PixelFormat::R16UNORM>, MortonCopy<true, PixelFormat::Z24S8>,
MortonCopy<true, PixelFormat::S8Z24>, MortonCopy<true, PixelFormat::Z32F>,
MortonCopy<true, PixelFormat::Z16>,
};
@@ -232,6 +235,8 @@ static constexpr std::array<void (*)(u32, u32, u32, u8*, Tegra::GPUVAddr),
MortonCopy<false, PixelFormat::RGBA32F>,
MortonCopy<false, PixelFormat::RG32F>,
MortonCopy<false, PixelFormat::R32F>,
MortonCopy<false, PixelFormat::R16F>,
MortonCopy<false, PixelFormat::R16UNORM>,
MortonCopy<false, PixelFormat::Z24S8>,
MortonCopy<false, PixelFormat::S8Z24>,
MortonCopy<false, PixelFormat::Z32F>,

View File

@@ -41,14 +41,16 @@ struct SurfaceParams {
RGBA32F = 16,
RG32F = 17,
R32F = 18,
R16F = 19,
R16UNORM = 20,
MaxColorFormat,
// DepthStencil formats
Z24S8 = 19,
S8Z24 = 20,
Z32F = 21,
Z16 = 22,
Z24S8 = 21,
S8Z24 = 22,
Z32F = 23,
Z16 = 24,
MaxDepthStencilFormat,
@@ -105,6 +107,8 @@ struct SurfaceParams {
1, // RGBA32F
1, // RG32F
1, // R32F
1, // R16F
1, // R16UNORM
1, // Z24S8
1, // S8Z24
1, // Z32F
@@ -139,6 +143,8 @@ struct SurfaceParams {
128, // RGBA32F
64, // RG32F
32, // R32F
16, // R16F
16, // R16UNORM
32, // Z24S8
32, // S8Z24
32, // Z32F
@@ -226,6 +232,16 @@ struct SurfaceParams {
UNREACHABLE();
case Tegra::Texture::TextureFormat::R32_G32:
return PixelFormat::RG32F;
case Tegra::Texture::TextureFormat::R16:
switch (component_type) {
case Tegra::Texture::ComponentType::FLOAT:
return PixelFormat::R16F;
case Tegra::Texture::ComponentType::UNORM:
return PixelFormat::R16UNORM;
}
LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}",
static_cast<u32>(component_type));
UNREACHABLE();
case Tegra::Texture::TextureFormat::R32:
return PixelFormat::R32F;
case Tegra::Texture::TextureFormat::DXT1:
@@ -290,6 +306,9 @@ struct SurfaceParams {
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;
default:
LOG_CRITICAL(HW_GPU, "Unimplemented format={}", static_cast<u32>(format));
UNREACHABLE();

View File

@@ -66,6 +66,7 @@ u32 BytesPerPixel(TextureFormat format) {
case TextureFormat::A1B5G5R5:
case TextureFormat::B5G6R5:
case TextureFormat::G8R8:
case TextureFormat::R16:
return 2;
case TextureFormat::R8:
return 1;
@@ -123,6 +124,7 @@ std::vector<u8> UnswizzleTexture(VAddr address, TextureFormat format, u32 width,
case TextureFormat::R32_G32_B32_A32:
case TextureFormat::R32_G32:
case TextureFormat::R32:
case TextureFormat::R16:
case TextureFormat::BF10GF11RF11:
case TextureFormat::ASTC_2D_4X4:
CopySwizzledData(width, height, bytes_per_pixel, bytes_per_pixel, data,
@@ -181,6 +183,7 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
case TextureFormat::R32_G32_B32_A32:
case TextureFormat::R32_G32:
case TextureFormat::R32:
case TextureFormat::R16:
// TODO(Subv): For the time being just forward the same data without any decoding.
rgba_data = texture_data;
break;