Mirrors/citra-nightly
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master' into feature/savestates-2

Browse source
This commit is contained in:
Hamish Milne 2020-04-12 11:24:06 +01:00
commit 828f88d20a
76 changed files with 3507 additions and 1369 deletions
Download patch file Download diff file Expand all files Collapse all files

13
src/video_core/CMakeLists.txt
View file

@ -23,6 +23,8 @@ add_library(video_core STATIC
regs_texturing.h
renderer_base.cpp
renderer_base.h
renderer_opengl/frame_dumper_opengl.cpp
renderer_opengl/frame_dumper_opengl.h
renderer_opengl/gl_rasterizer.cpp
renderer_opengl/gl_rasterizer.h
renderer_opengl/gl_rasterizer_cache.cpp
@ -43,6 +45,8 @@ add_library(video_core STATIC
renderer_opengl/gl_state.h
renderer_opengl/gl_stream_buffer.cpp
renderer_opengl/gl_stream_buffer.h
renderer_opengl/gl_surface_params.cpp
renderer_opengl/gl_surface_params.h
renderer_opengl/gl_vars.cpp
renderer_opengl/gl_vars.h
renderer_opengl/pica_to_gl.h
@ -54,11 +58,14 @@ add_library(video_core STATIC
renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h
renderer_opengl/texture_filters/bicubic/bicubic.cpp
renderer_opengl/texture_filters/bicubic/bicubic.h
renderer_opengl/texture_filters/texture_filter_interface.h
renderer_opengl/texture_filters/texture_filter_manager.cpp
renderer_opengl/texture_filters/texture_filter_manager.h
renderer_opengl/texture_filters/texture_filter_base.h
renderer_opengl/texture_filters/texture_filterer.cpp
renderer_opengl/texture_filters/texture_filterer.h
renderer_opengl/texture_filters/xbrz/xbrz_freescale.cpp
renderer_opengl/texture_filters/xbrz/xbrz_freescale.h
#temporary, move these back in alphabetical order before merging
renderer_opengl/gl_format_reinterpreter.cpp
renderer_opengl/gl_format_reinterpreter.h
shader/debug_data.h
shader/shader.cpp
shader/shader.h

98
src/video_core/renderer_opengl/frame_dumper_opengl.cpp Normal file
View file

@ -0,0 +1,98 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <glad/glad.h>
#include "core/frontend/emu_window.h"
#include "core/frontend/scope_acquire_context.h"
#include "video_core/renderer_opengl/frame_dumper_opengl.h"
#include "video_core/renderer_opengl/renderer_opengl.h"
namespace OpenGL {
FrameDumperOpenGL::FrameDumperOpenGL(VideoDumper::Backend& video_dumper_,
Frontend::EmuWindow& emu_window)
: video_dumper(video_dumper_), context(emu_window.CreateSharedContext()) {}
FrameDumperOpenGL::~FrameDumperOpenGL() {
if (present_thread.joinable())
present_thread.join();
}
bool FrameDumperOpenGL::IsDumping() const {
return video_dumper.IsDumping();
}
Layout::FramebufferLayout FrameDumperOpenGL::GetLayout() const {
return video_dumper.GetLayout();
}
void FrameDumperOpenGL::StartDumping() {
if (present_thread.joinable())
present_thread.join();
present_thread = std::thread(&FrameDumperOpenGL::PresentLoop, this);
}
void FrameDumperOpenGL::StopDumping() {
stop_requested.store(true, std::memory_order_relaxed);
}
void FrameDumperOpenGL::PresentLoop() {
Frontend::ScopeAcquireContext scope{*context};
InitializeOpenGLObjects();
const auto& layout = GetLayout();
while (!stop_requested.exchange(false)) {
auto frame = mailbox->TryGetPresentFrame(200);
if (!frame) {
continue;
}
if (frame->color_reloaded) {
LOG_DEBUG(Render_OpenGL, "Reloading present frame");
mailbox->ReloadPresentFrame(frame, layout.width, layout.height);
}
glWaitSync(frame->render_fence, 0, GL_TIMEOUT_IGNORED);
glBindFramebuffer(GL_READ_FRAMEBUFFER, frame->present.handle);
glBindBuffer(GL_PIXEL_PACK_BUFFER, pbos[current_pbo].handle);
glReadPixels(0, 0, layout.width, layout.height, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, 0);
// Insert fence for the main thread to block on
frame->present_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
glFlush();
// Bind the previous PBO and read the pixels
glBindBuffer(GL_PIXEL_PACK_BUFFER, pbos[next_pbo].handle);
GLubyte* pixels = static_cast<GLubyte*>(glMapBuffer(GL_PIXEL_PACK_BUFFER, GL_READ_ONLY));
VideoDumper::VideoFrame frame_data{layout.width, layout.height, pixels};
video_dumper.AddVideoFrame(std::move(frame_data));
glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
current_pbo = (current_pbo + 1) % 2;
next_pbo = (current_pbo + 1) % 2;
}
CleanupOpenGLObjects();
}
void FrameDumperOpenGL::InitializeOpenGLObjects() {
const auto& layout = GetLayout();
for (auto& buffer : pbos) {
buffer.Create();
glBindBuffer(GL_PIXEL_PACK_BUFFER, buffer.handle);
glBufferData(GL_PIXEL_PACK_BUFFER, layout.width * layout.height * 4, nullptr,
GL_STREAM_READ);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
}
}
void FrameDumperOpenGL::CleanupOpenGLObjects() {
for (auto& buffer : pbos) {
buffer.Release();
}
}
} // namespace OpenGL

57
src/video_core/renderer_opengl/frame_dumper_opengl.h Normal file
View file

@ -0,0 +1,57 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <atomic>
#include <memory>
#include <thread>
#include "core/dumping/backend.h"
#include "core/frontend/framebuffer_layout.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
namespace Frontend {
class EmuWindow;
class GraphicsContext;
class TextureMailbox;
} // namespace Frontend
namespace OpenGL {
class RendererOpenGL;
/**
* This is the 'presentation' part in frame dumping.
* Processes frames/textures sent to its mailbox, downloads the pixels and sends the data
* to the video encoding backend.
*/
class FrameDumperOpenGL {
public:
explicit FrameDumperOpenGL(VideoDumper::Backend& video_dumper, Frontend::EmuWindow& emu_window);
~FrameDumperOpenGL();
bool IsDumping() const;
Layout::FramebufferLayout GetLayout() const;
void StartDumping();
void StopDumping();
std::unique_ptr<Frontend::TextureMailbox> mailbox;
private:
void InitializeOpenGLObjects();
void CleanupOpenGLObjects();
void PresentLoop();
VideoDumper::Backend& video_dumper;
std::unique_ptr<Frontend::GraphicsContext> context;
std::thread present_thread;
std::atomic_bool stop_requested{false};
// PBOs used to dump frames faster
std::array<OGLBuffer, 2> pbos;
GLuint current_pbo = 1;
GLuint next_pbo = 0;
};
} // namespace OpenGL

238
src/video_core/renderer_opengl/gl_format_reinterpreter.cpp Normal file
View file

@ -0,0 +1,238 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/assert.h"
#include "common/scope_exit.h"
#include "video_core/renderer_opengl/gl_format_reinterpreter.h"
#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/gl_vars.h"
#include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
namespace OpenGL {
using PixelFormat = SurfaceParams::PixelFormat;
class RGBA4toRGB5A1 final : public FormatReinterpreterBase {
public:
RGBA4toRGB5A1() {
constexpr std::string_view vs_source = R"(
out vec2 dst_coord;
uniform mediump ivec2 dst_size;
const vec2 vertices[4] =
vec2[4](vec2(-1.0, -1.0), vec2(1.0, -1.0), vec2(-1.0, 1.0), vec2(1.0, 1.0));
void main() {
gl_Position = vec4(vertices[gl_VertexID], 0.0, 1.0);
dst_coord = (vertices[gl_VertexID] / 2.0 + 0.5) * vec2(dst_size);
}
)";
constexpr std::string_view fs_source = R"(
in mediump vec2 dst_coord;
out lowp vec4 frag_color;
uniform lowp sampler2D source;
uniform mediump ivec2 dst_size;
uniform mediump ivec2 src_size;
uniform mediump ivec2 src_offset;
void main() {
mediump ivec2 tex_coord;
if (src_size == dst_size) {
tex_coord = ivec2(dst_coord);
} else {
highp int tex_index = int(dst_coord.y) * dst_size.x + int(dst_coord.x);
mediump int y = tex_index / src_size.x;
tex_coord = ivec2(tex_index - y * src_size.x, y);
}
tex_coord -= src_offset;
lowp ivec4 rgba4 = ivec4(texelFetch(source, tex_coord, 0) * (exp2(4.0) - 1.0));
lowp ivec3 rgb5 =
((rgba4.rgb << ivec3(1, 2, 3)) | (rgba4.gba >> ivec3(3, 2, 1))) & 0x1F;
frag_color = vec4(vec3(rgb5) / (exp2(5.0) - 1.0), rgba4.a & 0x01);
}
)";
program.Create(vs_source.data(), fs_source.data());
dst_size_loc = glGetUniformLocation(program.handle, "dst_size");
src_size_loc = glGetUniformLocation(program.handle, "src_size");
src_offset_loc = glGetUniformLocation(program.handle, "src_offset");
vao.Create();
}
void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint read_fb_handle,
GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
GLuint draw_fb_handle) override {
OpenGLState prev_state = OpenGLState::GetCurState();
SCOPE_EXIT({ prev_state.Apply(); });
OpenGLState state;
state.texture_units[0].texture_2d = src_tex;
state.draw.draw_framebuffer = draw_fb_handle;
state.draw.shader_program = program.handle;
state.draw.vertex_array = vao.handle;
state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
static_cast<GLsizei>(dst_rect.GetWidth()),
static_cast<GLsizei>(dst_rect.GetHeight())};
state.Apply();
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
0);
glUniform2i(dst_size_loc, dst_rect.GetWidth(), dst_rect.GetHeight());
glUniform2i(src_size_loc, src_rect.GetWidth(), src_rect.GetHeight());
glUniform2i(src_offset_loc, src_rect.left, src_rect.bottom);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
private:
OGLProgram program;
GLint dst_size_loc{-1}, src_size_loc{-1}, src_offset_loc{-1};
OGLVertexArray vao;
};
class PixelBufferD24S8toABGR final : public FormatReinterpreterBase {
public:
PixelBufferD24S8toABGR() {
attributeless_vao.Create();
d24s8_abgr_buffer.Create();
d24s8_abgr_buffer_size = 0;
constexpr std::string_view vs_source = R"(
const vec2 vertices[4] = vec2[4](vec2(-1.0, -1.0), vec2(1.0, -1.0),
vec2(-1.0, 1.0), vec2(1.0, 1.0));
void main() {
gl_Position = vec4(vertices[gl_VertexID], 0.0, 1.0);
}
)";
std::string fs_source = GLES ? fragment_shader_precision_OES : "";
fs_source += R"(
uniform samplerBuffer tbo;
uniform vec2 tbo_size;
uniform vec4 viewport;
out vec4 color;
void main() {
vec2 tbo_coord = (gl_FragCoord.xy - viewport.xy) * tbo_size / viewport.zw;
int tbo_offset = int(tbo_coord.y) * int(tbo_size.x) + int(tbo_coord.x);
color = texelFetch(tbo, tbo_offset).rabg;
}
)";
d24s8_abgr_shader.Create(vs_source.data(), fs_source.c_str());
OpenGLState state = OpenGLState::GetCurState();
GLuint old_program = state.draw.shader_program;
state.draw.shader_program = d24s8_abgr_shader.handle;
state.Apply();
GLint tbo_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo");
ASSERT(tbo_u_id != -1);
glUniform1i(tbo_u_id, 0);
state.draw.shader_program = old_program;
state.Apply();
d24s8_abgr_tbo_size_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo_size");
ASSERT(d24s8_abgr_tbo_size_u_id != -1);
d24s8_abgr_viewport_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "viewport");
ASSERT(d24s8_abgr_viewport_u_id != -1);
}
~PixelBufferD24S8toABGR() {}
void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint read_fb_handle,
GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
GLuint draw_fb_handle) override {
OpenGLState prev_state = OpenGLState::GetCurState();
SCOPE_EXIT({ prev_state.Apply(); });
OpenGLState state;
state.draw.read_framebuffer = read_fb_handle;
state.draw.draw_framebuffer = draw_fb_handle;
state.Apply();
glBindBuffer(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer.handle);
GLsizeiptr target_pbo_size =
static_cast<GLsizeiptr>(src_rect.GetWidth()) * src_rect.GetHeight() * 4;
if (target_pbo_size > d24s8_abgr_buffer_size) {
d24s8_abgr_buffer_size = target_pbo_size * 2;
glBufferData(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer_size, nullptr, GL_STREAM_COPY);
}
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
src_tex, 0);
glReadPixels(static_cast<GLint>(src_rect.left), static_cast<GLint>(src_rect.bottom),
static_cast<GLsizei>(src_rect.GetWidth()),
static_cast<GLsizei>(src_rect.GetHeight()), GL_DEPTH_STENCIL,
GL_UNSIGNED_INT_24_8, 0);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
// PBO now contains src_tex in RABG format
state.draw.shader_program = d24s8_abgr_shader.handle;
state.draw.vertex_array = attributeless_vao.handle;
state.viewport.x = static_cast<GLint>(dst_rect.left);
state.viewport.y = static_cast<GLint>(dst_rect.bottom);
state.viewport.width = static_cast<GLsizei>(dst_rect.GetWidth());
state.viewport.height = static_cast<GLsizei>(dst_rect.GetHeight());
state.Apply();
OGLTexture tbo;
tbo.Create();
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_BUFFER, tbo.handle);
glTexBuffer(GL_TEXTURE_BUFFER, GL_RGBA8, d24s8_abgr_buffer.handle);
glUniform2f(d24s8_abgr_tbo_size_u_id, static_cast<GLfloat>(src_rect.GetWidth()),
static_cast<GLfloat>(src_rect.GetHeight()));
glUniform4f(d24s8_abgr_viewport_u_id, static_cast<GLfloat>(state.viewport.x),
static_cast<GLfloat>(state.viewport.y),
static_cast<GLfloat>(state.viewport.width),
static_cast<GLfloat>(state.viewport.height));
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex,
0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
0);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_BUFFER, 0);
}
private:
OGLVertexArray attributeless_vao;
OGLBuffer d24s8_abgr_buffer;
GLsizeiptr d24s8_abgr_buffer_size;
OGLProgram d24s8_abgr_shader;
GLint d24s8_abgr_tbo_size_u_id;
GLint d24s8_abgr_viewport_u_id;
};
FormatReinterpreterOpenGL::FormatReinterpreterOpenGL() {
reinterpreters.emplace(PixelFormatPair{PixelFormat::RGBA8, PixelFormat::D24S8},
std::make_unique<PixelBufferD24S8toABGR>());
reinterpreters.emplace(PixelFormatPair{PixelFormat::RGB5A1, PixelFormat::RGBA4},
std::make_unique<RGBA4toRGB5A1>());
}
FormatReinterpreterOpenGL::~FormatReinterpreterOpenGL() = default;
std::pair<FormatReinterpreterOpenGL::ReinterpreterMap::iterator,
FormatReinterpreterOpenGL::ReinterpreterMap::iterator>
FormatReinterpreterOpenGL::GetPossibleReinterpretations(PixelFormat dst_format) {
return reinterpreters.equal_range(dst_format);
}
} // namespace OpenGL

62
src/video_core/renderer_opengl/gl_format_reinterpreter.h Normal file
View file

@ -0,0 +1,62 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <map>
#include <type_traits>
#include <glad/glad.h>
#include "common/common_types.h"
#include "common/math_util.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_surface_params.h"
namespace OpenGL {
class RasterizerCacheOpenGL;
struct PixelFormatPair {
const SurfaceParams::PixelFormat dst_format, src_format;
struct less {
using is_transparent = void;
constexpr bool operator()(OpenGL::PixelFormatPair lhs, OpenGL::PixelFormatPair rhs) const {
return std::tie(lhs.dst_format, lhs.src_format) <
std::tie(rhs.dst_format, rhs.src_format);
}
constexpr bool operator()(OpenGL::SurfaceParams::PixelFormat lhs,
OpenGL::PixelFormatPair rhs) const {
return lhs < rhs.dst_format;
}
constexpr bool operator()(OpenGL::PixelFormatPair lhs,
OpenGL::SurfaceParams::PixelFormat rhs) const {
return lhs.dst_format < rhs;
}
};
};
class FormatReinterpreterBase {
public:
virtual ~FormatReinterpreterBase() = default;
virtual void Reinterpret(GLuint src_tex, const Common::Rectangle<u32>& src_rect,
GLuint read_fb_handle, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint draw_fb_handle) = 0;
};
class FormatReinterpreterOpenGL : NonCopyable {
using ReinterpreterMap =
std::map<PixelFormatPair, std::unique_ptr<FormatReinterpreterBase>, PixelFormatPair::less>;
public:
explicit FormatReinterpreterOpenGL();
~FormatReinterpreterOpenGL();
std::pair<ReinterpreterMap::iterator, ReinterpreterMap::iterator> GetPossibleReinterpretations(
SurfaceParams::PixelFormat dst_format);
private:
ReinterpreterMap reinterpreters;
};
} // namespace OpenGL

474
src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
View file

@ -5,6 +5,7 @@
#include <algorithm>
#include <array>
#include <atomic>
#include <bitset>
#include <cstring>
#include <iterator>
#include <memory>
@ -31,10 +32,11 @@
#include "core/settings.h"
#include "video_core/pica_state.h"
#include "video_core/renderer_base.h"
#include "video_core/renderer_opengl/gl_format_reinterpreter.h"
#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/gl_vars.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_manager.h"
#include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
#include "video_core/utils.h"
#include "video_core/video_core.h"
@ -493,125 +495,6 @@ static bool FillSurface(const Surface& surface, const u8* fill_data,
return true;
}
SurfaceParams SurfaceParams::FromInterval(SurfaceInterval interval) const {
SurfaceParams params = *this;
const u32 tiled_size = is_tiled ? 8 : 1;
const u32 stride_tiled_bytes = BytesInPixels(stride * tiled_size);
PAddr aligned_start =
addr + Common::AlignDown(boost::icl::first(interval) - addr, stride_tiled_bytes);
PAddr aligned_end =
addr + Common::AlignUp(boost::icl::last_next(interval) - addr, stride_tiled_bytes);
if (aligned_end - aligned_start > stride_tiled_bytes) {
params.addr = aligned_start;
params.height = (aligned_end - aligned_start) / BytesInPixels(stride);
} else {
// 1 row
ASSERT(aligned_end - aligned_start == stride_tiled_bytes);
const u32 tiled_alignment = BytesInPixels(is_tiled ? 8 * 8 : 1);
aligned_start =
addr + Common::AlignDown(boost::icl::first(interval) - addr, tiled_alignment);
aligned_end =
addr + Common::AlignUp(boost::icl::last_next(interval) - addr, tiled_alignment);
params.addr = aligned_start;
params.width = PixelsInBytes(aligned_end - aligned_start) / tiled_size;
params.stride = params.width;
params.height = tiled_size;
}
params.UpdateParams();
return params;
}
SurfaceInterval SurfaceParams::GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const {
if (unscaled_rect.GetHeight() == 0 || unscaled_rect.GetWidth() == 0) {
return {};
}
if (is_tiled) {
unscaled_rect.left = Common::AlignDown(unscaled_rect.left, 8) * 8;
unscaled_rect.bottom = Common::AlignDown(unscaled_rect.bottom, 8) / 8;
unscaled_rect.right = Common::AlignUp(unscaled_rect.right, 8) * 8;
unscaled_rect.top = Common::AlignUp(unscaled_rect.top, 8) / 8;
}
const u32 stride_tiled = !is_tiled ? stride : stride * 8;
const u32 pixel_offset =
stride_tiled * (!is_tiled ? unscaled_rect.bottom : (height / 8) - unscaled_rect.top) +
unscaled_rect.left;
const u32 pixels = (unscaled_rect.GetHeight() - 1) * stride_tiled + unscaled_rect.GetWidth();
return {addr + BytesInPixels(pixel_offset), addr + BytesInPixels(pixel_offset + pixels)};
}
Common::Rectangle<u32> SurfaceParams::GetSubRect(const SurfaceParams& sub_surface) const {
const u32 begin_pixel_index = PixelsInBytes(sub_surface.addr - addr);
if (is_tiled) {
const int x0 = (begin_pixel_index % (stride * 8)) / 8;
const int y0 = (begin_pixel_index / (stride * 8)) * 8;
// Top to bottom
return Common::Rectangle<u32>(x0, height - y0, x0 + sub_surface.width,
height - (y0 + sub_surface.height));
}
const int x0 = begin_pixel_index % stride;
const int y0 = begin_pixel_index / stride;
// Bottom to top
return Common::Rectangle<u32>(x0, y0 + sub_surface.height, x0 + sub_surface.width, y0);
}
Common::Rectangle<u32> SurfaceParams::GetScaledSubRect(const SurfaceParams& sub_surface) const {
auto rect = GetSubRect(sub_surface);
rect.left = rect.left * res_scale;
rect.right = rect.right * res_scale;
rect.top = rect.top * res_scale;
rect.bottom = rect.bottom * res_scale;
return rect;
}
bool SurfaceParams::ExactMatch(const SurfaceParams& other_surface) const {
return std::tie(other_surface.addr, other_surface.width, other_surface.height,
other_surface.stride, other_surface.pixel_format, other_surface.is_tiled) ==
std::tie(addr, width, height, stride, pixel_format, is_tiled) &&
pixel_format != PixelFormat::Invalid;
}
bool SurfaceParams::CanSubRect(const SurfaceParams& sub_surface) const {
return sub_surface.addr >= addr && sub_surface.end <= end &&
sub_surface.pixel_format == pixel_format && pixel_format != PixelFormat::Invalid &&
sub_surface.is_tiled == is_tiled &&
(sub_surface.addr - addr) % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
(sub_surface.stride == stride || sub_surface.height <= (is_tiled ? 8u : 1u)) &&
GetSubRect(sub_surface).right <= stride;
}
bool SurfaceParams::CanExpand(const SurfaceParams& expanded_surface) const {
return pixel_format != PixelFormat::Invalid && pixel_format == expanded_surface.pixel_format &&
addr <= expanded_surface.end && expanded_surface.addr <= end &&
is_tiled == expanded_surface.is_tiled && stride == expanded_surface.stride &&
(std::max(expanded_surface.addr, addr) - std::min(expanded_surface.addr, addr)) %
BytesInPixels(stride * (is_tiled ? 8 : 1)) ==
0;
}
bool SurfaceParams::CanTexCopy(const SurfaceParams& texcopy_params) const {
if (pixel_format == PixelFormat::Invalid || addr > texcopy_params.addr ||
end < texcopy_params.end) {
return false;
}
if (texcopy_params.width != texcopy_params.stride) {
const u32 tile_stride = BytesInPixels(stride * (is_tiled ? 8 : 1));
return (texcopy_params.addr - addr) % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
texcopy_params.width % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
(texcopy_params.height == 1 || texcopy_params.stride == tile_stride) &&
((texcopy_params.addr - addr) % tile_stride) + texcopy_params.width <= tile_stride;
}
return FromInterval(texcopy_params.GetInterval()).GetInterval() == texcopy_params.GetInterval();
}
bool CachedSurface::CanFill(const SurfaceParams& dest_surface,
SurfaceInterval fill_interval) const {
if (type == SurfaceType::Fill && IsRegionValid(fill_interval) &&
@ -653,47 +536,6 @@ bool CachedSurface::CanCopy(const SurfaceParams& dest_surface,
return false;
}
SurfaceInterval SurfaceParams::GetCopyableInterval(const Surface& src_surface) const {
SurfaceInterval result{};
const auto valid_regions =
SurfaceRegions(GetInterval() & src_surface->GetInterval()) - src_surface->invalid_regions;
for (auto& valid_interval : valid_regions) {
const SurfaceInterval aligned_interval{
addr + Common::AlignUp(boost::icl::first(valid_interval) - addr,
BytesInPixels(is_tiled ? 8 * 8 : 1)),
addr + Common::AlignDown(boost::icl::last_next(valid_interval) - addr,
BytesInPixels(is_tiled ? 8 * 8 : 1))};
if (BytesInPixels(is_tiled ? 8 * 8 : 1) > boost::icl::length(valid_interval) ||
boost::icl::length(aligned_interval) == 0) {
continue;
}
// Get the rectangle within aligned_interval
const u32 stride_bytes = BytesInPixels(stride) * (is_tiled ? 8 : 1);
SurfaceInterval rect_interval{
addr + Common::AlignUp(boost::icl::first(aligned_interval) - addr, stride_bytes),
addr + Common::AlignDown(boost::icl::last_next(aligned_interval) - addr, stride_bytes),
};
if (boost::icl::first(rect_interval) > boost::icl::last_next(rect_interval)) {
// 1 row
rect_interval = aligned_interval;
} else if (boost::icl::length(rect_interval) == 0) {
// 2 rows that do not make a rectangle, return the larger one
const SurfaceInterval row1{boost::icl::first(aligned_interval),
boost::icl::first(rect_interval)};
const SurfaceInterval row2{boost::icl::first(rect_interval),
boost::icl::last_next(aligned_interval)};
rect_interval = (boost::icl::length(row1) > boost::icl::length(row2)) ? row1 : row2;
}
if (boost::icl::length(rect_interval) > boost::icl::length(result)) {
result = rect_interval;
}
}
return result;
}
MICROPROFILE_DEFINE(OpenGL_CopySurface, "OpenGL", "CopySurface", MP_RGB(128, 192, 64));
void RasterizerCacheOpenGL::CopySurface(const Surface& src_surface, const Surface& dst_surface,
SurfaceInterval copy_interval) {
@ -884,6 +726,16 @@ bool CachedSurface::LoadCustomTexture(u64 tex_hash, Core::CustomTexInfo& tex_inf
}
void CachedSurface::DumpTexture(GLuint target_tex, u64 tex_hash) {
// Make sure the texture size is a power of 2
// If not, the surface is actually a framebuffer
std::bitset<32> width_bits(width);
std::bitset<32> height_bits(height);
if (width_bits.count() != 1 || height_bits.count() != 1) {
LOG_WARNING(Render_OpenGL, "Not dumping {:016X} because size isn't a power of 2 ({}x{})",
tex_hash, width, height);
return;
}
// Dump texture to RGBA8 and encode as PNG
const auto& image_interface = Core::System::GetInstance().GetImageInterface();
auto& custom_tex_cache = Core::System::GetInstance().CustomTexCache();
@ -945,10 +797,6 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
if (Settings::values.custom_textures)
is_custom = LoadCustomTexture(tex_hash, custom_tex_info);
TextureFilterInterface* const texture_filter =
is_custom ? nullptr : TextureFilterManager::GetInstance().GetTextureFilter();
const u16 default_scale = texture_filter ? texture_filter->scale_factor : 1;
// Load data from memory to the surface
GLint x0 = static_cast<GLint>(rect.left);
GLint y0 = static_cast<GLint>(rect.bottom);
@ -960,7 +808,7 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
// If not 1x scale, create 1x texture that we will blit from to replace texture subrect in
// surface
OGLTexture unscaled_tex;
if (res_scale != default_scale) {
if (res_scale != 1) {
x0 = 0;
y0 = 0;
@ -969,8 +817,7 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
AllocateSurfaceTexture(unscaled_tex.handle, GetFormatTuple(PixelFormat::RGBA8),
custom_tex_info.width, custom_tex_info.height);
} else {
AllocateSurfaceTexture(unscaled_tex.handle, tuple, rect.GetWidth() * default_scale,
rect.GetHeight() * default_scale);
AllocateSurfaceTexture(unscaled_tex.handle, tuple, rect.GetWidth(), rect.GetHeight());
}
target_tex = unscaled_tex.handle;
}
@ -996,16 +843,6 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
glActiveTexture(GL_TEXTURE0);
glTexSubImage2D(GL_TEXTURE_2D, 0, x0, y0, custom_tex_info.width, custom_tex_info.height,
GL_RGBA, GL_UNSIGNED_BYTE, custom_tex_info.tex.data());
} else if (texture_filter) {
if (res_scale == default_scale) {
AllocateSurfaceTexture(texture.handle, GetFormatTuple(pixel_format),
rect.GetWidth() * default_scale,
rect.GetHeight() * default_scale);
cur_state.texture_units[0].texture_2d = texture.handle;
cur_state.Apply();
}
texture_filter->scale(*this, {(u32)x0, (u32)y0, rect.GetWidth(), rect.GetHeight()},
buffer_offset);
} else {
glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(stride));
@ -1016,13 +853,13 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
}
glPixelStorei(GL_UNPACK_ROW_LENGTH, 0);
if (Settings::values.dump_textures && !is_custom && !texture_filter)
if (Settings::values.dump_textures && !is_custom)
DumpTexture(target_tex, tex_hash);
cur_state.texture_units[0].texture_2d = old_tex;
cur_state.Apply();
if (res_scale != default_scale) {
if (res_scale != 1) {
auto scaled_rect = rect;
scaled_rect.left *= res_scale;
scaled_rect.top *= res_scale;
@ -1031,8 +868,11 @@ void CachedSurface::UploadGLTexture(Common::Rectangle<u32> rect, GLuint read_fb_
auto from_rect =
is_custom ? Common::Rectangle<u32>{0, custom_tex_info.height, custom_tex_info.width, 0}
: Common::Rectangle<u32>{0, rect.GetHeight(), rect.GetWidth(), 0};
BlitTextures(unscaled_tex.handle, from_rect, texture.handle, scaled_rect, type,
read_fb_handle, draw_fb_handle);
if (!owner.texture_filterer->Filter(unscaled_tex.handle, from_rect, texture.handle,
scaled_rect, type, read_fb_handle, draw_fb_handle)) {
BlitTextures(unscaled_tex.handle, from_rect, texture.handle, scaled_rect, type,
read_fb_handle, draw_fb_handle);
}
}
InvalidateAllWatcher();
@ -1221,53 +1061,13 @@ Surface FindMatch(const SurfaceCache& surface_cache, const SurfaceParams& params
}
RasterizerCacheOpenGL::RasterizerCacheOpenGL() {
resolution_scale_factor = VideoCore::GetResolutionScaleFactor();
texture_filterer = std::make_unique<TextureFilterer>(Settings::values.texture_filter_name,
resolution_scale_factor);
format_reinterpreter = std::make_unique<FormatReinterpreterOpenGL>();
read_framebuffer.Create();
draw_framebuffer.Create();
attributeless_vao.Create();
d24s8_abgr_buffer.Create();
d24s8_abgr_buffer_size = 0;
std::string vs_source = R"(
const vec2 vertices[4] = vec2[4](vec2(-1.0, -1.0), vec2(1.0, -1.0), vec2(-1.0, 1.0), vec2(1.0, 1.0));
void main() {
gl_Position = vec4(vertices[gl_VertexID], 0.0, 1.0);
}
)";
std::string fs_source = GLES ? fragment_shader_precision_OES : "";
fs_source += R"(
uniform samplerBuffer tbo;
uniform vec2 tbo_size;
uniform vec4 viewport;
out vec4 color;
void main() {
vec2 tbo_coord = (gl_FragCoord.xy - viewport.xy) * tbo_size / viewport.zw;
int tbo_offset = int(tbo_coord.y) * int(tbo_size.x) + int(tbo_coord.x);
color = texelFetch(tbo, tbo_offset).rabg;
}
)";
d24s8_abgr_shader.Create(vs_source.c_str(), fs_source.c_str());
OpenGLState state = OpenGLState::GetCurState();
GLuint old_program = state.draw.shader_program;
state.draw.shader_program = d24s8_abgr_shader.handle;
state.Apply();
GLint tbo_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo");
ASSERT(tbo_u_id != -1);
glUniform1i(tbo_u_id, 0);
state.draw.shader_program = old_program;
state.Apply();
d24s8_abgr_tbo_size_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "tbo_size");
ASSERT(d24s8_abgr_tbo_size_u_id != -1);
d24s8_abgr_viewport_u_id = glGetUniformLocation(d24s8_abgr_shader.handle, "viewport");
ASSERT(d24s8_abgr_viewport_u_id != -1);
}
RasterizerCacheOpenGL::~RasterizerCacheOpenGL() {
@ -1291,64 +1091,6 @@ bool RasterizerCacheOpenGL::BlitSurfaces(const Surface& src_surface,
draw_framebuffer.handle);
}
void RasterizerCacheOpenGL::ConvertD24S8toABGR(GLuint src_tex,
const Common::Rectangle<u32>& src_rect,
GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect) {
OpenGLState prev_state = OpenGLState::GetCurState();
SCOPE_EXIT({ prev_state.Apply(); });
OpenGLState state;
state.draw.read_framebuffer = read_framebuffer.handle;
state.draw.draw_framebuffer = draw_framebuffer.handle;
state.Apply();
glBindBuffer(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer.handle);
GLsizeiptr target_pbo_size = src_rect.GetWidth() * src_rect.GetHeight() * 4;
if (target_pbo_size > d24s8_abgr_buffer_size) {
d24s8_abgr_buffer_size = target_pbo_size * 2;
glBufferData(GL_PIXEL_PACK_BUFFER, d24s8_abgr_buffer_size, nullptr, GL_STREAM_COPY);
}
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, src_tex,
0);
glReadPixels(static_cast<GLint>(src_rect.left), static_cast<GLint>(src_rect.bottom),
static_cast<GLsizei>(src_rect.GetWidth()),
static_cast<GLsizei>(src_rect.GetHeight()), GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8,
0);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
// PBO now contains src_tex in RABG format
state.draw.shader_program = d24s8_abgr_shader.handle;
state.draw.vertex_array = attributeless_vao.handle;
state.viewport.x = static_cast<GLint>(dst_rect.left);
state.viewport.y = static_cast<GLint>(dst_rect.bottom);
state.viewport.width = static_cast<GLsizei>(dst_rect.GetWidth());
state.viewport.height = static_cast<GLsizei>(dst_rect.GetHeight());
state.Apply();
OGLTexture tbo;
tbo.Create();
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_BUFFER, tbo.handle);
glTexBuffer(GL_TEXTURE_BUFFER, GL_RGBA8, d24s8_abgr_buffer.handle);
glUniform2f(d24s8_abgr_tbo_size_u_id, static_cast<GLfloat>(src_rect.GetWidth()),
static_cast<GLfloat>(src_rect.GetHeight()));
glUniform4f(d24s8_abgr_viewport_u_id, static_cast<GLfloat>(state.viewport.x),
static_cast<GLfloat>(state.viewport.y), static_cast<GLfloat>(state.viewport.width),
static_cast<GLfloat>(state.viewport.height));
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_BUFFER, 0);
}
Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, ScaleMatch match_res_scale,
bool load_if_create) {
if (params.addr == 0 || params.height * params.width == 0) {
@ -1495,11 +1237,7 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Pica::Texture::TextureInf
params.height = info.height;
params.is_tiled = true;
params.pixel_format = SurfaceParams::PixelFormatFromTextureFormat(info.format);
TextureFilterInterface* filter{};
params.res_scale = (filter = TextureFilterManager::GetInstance().GetTextureFilter())
? filter->scale_factor
: 1;
params.res_scale = texture_filterer->IsNull() ? 1 : resolution_scale_factor;
params.UpdateParams();
u32 min_width = info.width >> max_level;
@ -1552,7 +1290,7 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Pica::Texture::TextureInf
glTexImage2D(GL_TEXTURE_2D, level, format_tuple.internal_format, width >> level,
height >> level, 0, format_tuple.format, format_tuple.type, nullptr);
}
if (surface->is_custom) {
if (surface->is_custom || !texture_filterer->IsNull()) {
// TODO: proper mipmap support for custom textures
glGenerateMipmap(GL_TEXTURE_2D);
}
@ -1588,7 +1326,7 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Pica::Texture::TextureInf
}
state.ResetTexture(level_surface->texture.handle);
state.Apply();
if (!surface->is_custom) {
if (!surface->is_custom && texture_filterer->IsNull()) {
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
level_surface->texture.handle, 0);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT,
@ -1712,10 +1450,9 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces(
const auto& config = regs.framebuffer.framebuffer;
// update resolution_scale_factor and reset cache if changed
static u16 resolution_scale_factor = VideoCore::GetResolutionScaleFactor();
if (resolution_scale_factor != VideoCore::GetResolutionScaleFactor() ||
TextureFilterManager::GetInstance().IsUpdated()) {
TextureFilterManager::GetInstance().Reset();
if ((resolution_scale_factor != VideoCore::GetResolutionScaleFactor()) |
(VideoCore::g_texture_filter_update_requested.exchange(false) &&
texture_filterer->Reset(Settings::values.texture_filter_name, resolution_scale_factor))) {
resolution_scale_factor = VideoCore::GetResolutionScaleFactor();
FlushAll();
while (!surface_cache.empty())
@ -1800,7 +1537,7 @@ SurfaceSurfaceRect_Tuple RasterizerCacheOpenGL::GetFramebufferSurfaces(
}
Surface RasterizerCacheOpenGL::GetFillSurface(const GPU::Regs::MemoryFillConfig& config) {
Surface new_surface = std::make_shared<CachedSurface>();
Surface new_surface = std::make_shared<CachedSurface>(*this);
new_surface->addr = config.GetStartAddress();
new_surface->end = config.GetEndAddress();
@ -1881,9 +1618,15 @@ void RasterizerCacheOpenGL::ValidateSurface(const Surface& surface, PAddr addr,
return;
}
auto validate_regions = surface->invalid_regions & validate_interval;
auto notify_validated = [&](SurfaceInterval interval) {
surface->invalid_regions.erase(interval);
validate_regions.erase(interval);
};
while (true) {
const auto it = surface->invalid_regions.find(validate_interval);
if (it == surface->invalid_regions.end())
const auto it = validate_regions.begin();
if (it == validate_regions.end())
break;
const auto interval = *it & validate_interval;
@ -1895,27 +1638,27 @@ void RasterizerCacheOpenGL::ValidateSurface(const Surface& surface, PAddr addr,
if (copy_surface != nullptr) {
SurfaceInterval copy_interval = params.GetCopyableInterval(copy_surface);
CopySurface(copy_surface, surface, copy_interval);
surface->invalid_regions.erase(copy_interval);
notify_validated(copy_interval);
continue;
}
// D24S8 to RGBA8
if (surface->pixel_format == PixelFormat::RGBA8) {
params.pixel_format = PixelFormat::D24S8;
Surface reinterpret_surface =
FindMatch<MatchFlags::Copy>(surface_cache, params, ScaleMatch::Ignore, interval);
if (reinterpret_surface != nullptr) {
ASSERT(reinterpret_surface->pixel_format == PixelFormat::D24S8);
SurfaceInterval convert_interval = params.GetCopyableInterval(reinterpret_surface);
SurfaceParams convert_params = surface->FromInterval(convert_interval);
auto src_rect = reinterpret_surface->GetScaledSubRect(convert_params);
auto dest_rect = surface->GetScaledSubRect(convert_params);
ConvertD24S8toABGR(reinterpret_surface->texture.handle, src_rect,
surface->texture.handle, dest_rect);
surface->invalid_regions.erase(convert_interval);
// Try to find surface in cache with different format
// that can can be reinterpreted to the requested format.
if (ValidateByReinterpretation(surface, params, interval)) {
notify_validated(interval);
continue;
}
// Could not find a matching reinterpreter, check if we need to implement a
// reinterpreter
if (NoUnimplementedReinterpretations(surface, params, interval) &&
!IntervalHasInvalidPixelFormat(params, interval)) {
// No surfaces were found in the cache that had a matching bit-width.
// If the region was created entirely on the GPU,
// assume it was a developer mistake and skip flushing.
if (boost::icl::contains(dirty_regions, interval)) {
LOG_DEBUG(Render_OpenGL, "Region created fully on GPU and reinterpretation is "
"invalid. Skipping validation");
validate_regions.erase(interval);
continue;
}
}
@ -1925,10 +1668,103 @@ void RasterizerCacheOpenGL::ValidateSurface(const Surface& surface, PAddr addr,
surface->LoadGLBuffer(params.addr, params.end);
surface->UploadGLTexture(surface->GetSubRect(params), read_framebuffer.handle,
draw_framebuffer.handle);
surface->invalid_regions.erase(params.GetInterval());
notify_validated(params.GetInterval());
}
}
bool RasterizerCacheOpenGL::NoUnimplementedReinterpretations(const Surface& surface,
SurfaceParams& params,
const SurfaceInterval& interval) {
static constexpr std::array<PixelFormat, 17> all_formats{
PixelFormat::RGBA8, PixelFormat::RGB8, PixelFormat::RGB5A1, PixelFormat::RGB565,
PixelFormat::RGBA4, PixelFormat::IA8, PixelFormat::RG8, PixelFormat::I8,
PixelFormat::A8, PixelFormat::IA4, PixelFormat::I4, PixelFormat::A4,
PixelFormat::ETC1, PixelFormat::ETC1A4, PixelFormat::D16, PixelFormat::D24,
PixelFormat::D24S8,
};
bool implemented = true;
for (PixelFormat format : all_formats) {
if (SurfaceParams::GetFormatBpp(format) == surface->GetFormatBpp()) {
params.pixel_format = format;
// This could potentially be expensive,
// although experimentally it hasn't been too bad
Surface test_surface =
FindMatch<MatchFlags::Copy>(surface_cache, params, ScaleMatch::Ignore, interval);
if (test_surface != nullptr) {
LOG_WARNING(Render_OpenGL, "Missing pixel_format reinterpreter: {} -> {}",
SurfaceParams::PixelFormatAsString(format),
SurfaceParams::PixelFormatAsString(surface->pixel_format));
implemented = false;
}
}
}
return implemented;
}
bool RasterizerCacheOpenGL::IntervalHasInvalidPixelFormat(SurfaceParams& params,
const SurfaceInterval& interval) {
params.pixel_format = PixelFormat::Invalid;
for (const auto& set : RangeFromInterval(surface_cache, interval))
for (const auto& surface : set.second)
if (surface->pixel_format == PixelFormat::Invalid) {
LOG_WARNING(Render_OpenGL, "Surface found with invalid pixel format");
return true;
}
return false;
}
bool RasterizerCacheOpenGL::ValidateByReinterpretation(const Surface& surface,
SurfaceParams& params,
const SurfaceInterval& interval) {
auto [cvt_begin, cvt_end] =
format_reinterpreter->GetPossibleReinterpretations(surface->pixel_format);
for (auto reinterpreter = cvt_begin; reinterpreter != cvt_end; ++reinterpreter) {
PixelFormat format = reinterpreter->first.src_format;
params.pixel_format = format;
Surface reinterpret_surface =
FindMatch<MatchFlags::Copy>(surface_cache, params, ScaleMatch::Ignore, interval);
if (reinterpret_surface != nullptr) {
SurfaceInterval reinterpret_interval = params.GetCopyableInterval(reinterpret_surface);
SurfaceParams reinterpret_params = surface->FromInterval(reinterpret_interval);
auto src_rect = reinterpret_surface->GetScaledSubRect(reinterpret_params);
auto dest_rect = surface->GetScaledSubRect(reinterpret_params);
if (!texture_filterer->IsNull() && reinterpret_surface->res_scale == 1 &&
surface->res_scale == resolution_scale_factor) {
// The destination surface is either a framebuffer, or a filtered texture.
OGLTexture tmp_tex;
tmp_tex.Create();
// Create an intermediate surface to convert to before blitting to the
// destination.
Common::Rectangle<u32> tmp_rect{0, dest_rect.GetHeight() / resolution_scale_factor,
dest_rect.GetWidth() / resolution_scale_factor, 0};
AllocateSurfaceTexture(tmp_tex.handle,
GetFormatTuple(reinterpreter->first.dst_format),
tmp_rect.right, tmp_rect.top);
reinterpreter->second->Reinterpret(reinterpret_surface->texture.handle, src_rect,
read_framebuffer.handle, tmp_tex.handle,
tmp_rect, draw_framebuffer.handle);
SurfaceParams::SurfaceType type =
SurfaceParams::GetFormatType(reinterpreter->first.dst_format);
if (!texture_filterer->Filter(tmp_tex.handle, tmp_rect, surface->texture.handle,
dest_rect, type, read_framebuffer.handle,
draw_framebuffer.handle)) {
BlitTextures(tmp_tex.handle, tmp_rect, surface->texture.handle, dest_rect, type,
read_framebuffer.handle, draw_framebuffer.handle);
}
} else {
reinterpreter->second->Reinterpret(reinterpret_surface->texture.handle, src_rect,
read_framebuffer.handle, surface->texture.handle,
dest_rect, draw_framebuffer.handle);
}
return true;
}
}
return false;
}
void RasterizerCacheOpenGL::ClearAll(bool flush) {
const auto flush_interval = PageMap::interval_type::right_open(0x0, 0xFFFFFFFF);
// Force flush all surfaces from the cache
@ -2053,7 +1889,7 @@ void RasterizerCacheOpenGL::InvalidateRegion(PAddr addr, u32 size, const Surface
}
Surface RasterizerCacheOpenGL::CreateSurface(const SurfaceParams& params) {
Surface surface = std::make_shared<CachedSurface>();
Surface surface = std::make_shared<CachedSurface>(*this);
static_cast<SurfaceParams&>(*surface) = params;
surface->texture.Create();

246
src/video_core/renderer_opengl/gl_rasterizer_cache.h
View file

@ -26,14 +26,16 @@
#include "common/common_types.h"
#include "common/math_util.h"
#include "core/custom_tex_cache.h"
#include "core/hw/gpu.h"
#include "video_core/regs_framebuffer.h"
#include "video_core/regs_texturing.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_surface_params.h"
#include "video_core/texture/texture_decode.h"
namespace OpenGL {
class RasterizerCacheOpenGL;
class TextureFilterer;
class FormatReinterpreterOpenGL;
struct TextureCubeConfig {
PAddr px;
PAddr nx;
@ -76,11 +78,8 @@ struct hash<OpenGL::TextureCubeConfig> {
namespace OpenGL {
struct CachedSurface;
using Surface = std::shared_ptr<CachedSurface>;
using SurfaceSet = std::set<Surface>;
using SurfaceInterval = boost::icl::right_open_interval<PAddr>;
using SurfaceRegions = boost::icl::interval_set<PAddr, std::less, SurfaceInterval>;
using SurfaceMap =
boost::icl::interval_map<PAddr, Surface, boost::icl::partial_absorber, std::less,
@ -104,212 +103,6 @@ enum class ScaleMatch {
Ignore // accept every scaled res
};
struct SurfaceParams {
private:
static constexpr std::array<unsigned int, 18> BPP_TABLE = {
32, // RGBA8
24, // RGB8
16, // RGB5A1
16, // RGB565
16, // RGBA4
16, // IA8
16, // RG8
8, // I8
8, // A8
8, // IA4
4, // I4
4, // A4
4, // ETC1
8, // ETC1A4
16, // D16
0,
24, // D24
32, // D24S8
};
public:
enum class PixelFormat {
// First 5 formats are shared between textures and color buffers
RGBA8 = 0,
RGB8 = 1,
RGB5A1 = 2,
RGB565 = 3,
RGBA4 = 4,
// Texture-only formats
IA8 = 5,
RG8 = 6,
I8 = 7,
A8 = 8,
IA4 = 9,
I4 = 10,
A4 = 11,
ETC1 = 12,
ETC1A4 = 13,
// Depth buffer-only formats
D16 = 14,
// gap
D24 = 16,
D24S8 = 17,
Invalid = 255,
};
enum class SurfaceType {
Color = 0,
Texture = 1,
Depth = 2,
DepthStencil = 3,
Fill = 4,
Invalid = 5
};
static constexpr unsigned int GetFormatBpp(PixelFormat format) {
const auto format_idx = static_cast<std::size_t>(format);
DEBUG_ASSERT_MSG(format_idx < BPP_TABLE.size(), "Invalid pixel format {}", format_idx);
return BPP_TABLE[format_idx];
}
unsigned int GetFormatBpp() const {
return GetFormatBpp(pixel_format);
}
static PixelFormat PixelFormatFromTextureFormat(Pica::TexturingRegs::TextureFormat format) {
return ((unsigned int)format < 14) ? (PixelFormat)format : PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromColorFormat(Pica::FramebufferRegs::ColorFormat format) {
return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromDepthFormat(Pica::FramebufferRegs::DepthFormat format) {
return ((unsigned int)format < 4) ? (PixelFormat)((unsigned int)format + 14)
: PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromGPUPixelFormat(GPU::Regs::PixelFormat format) {
switch (format) {
// RGB565 and RGB5A1 are switched in PixelFormat compared to ColorFormat
case GPU::Regs::PixelFormat::RGB565:
return PixelFormat::RGB565;
case GPU::Regs::PixelFormat::RGB5A1:
return PixelFormat::RGB5A1;
default:
return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
}
}
static bool CheckFormatsBlittable(PixelFormat pixel_format_a, PixelFormat pixel_format_b) {
SurfaceType a_type = GetFormatType(pixel_format_a);
SurfaceType b_type = GetFormatType(pixel_format_b);
if ((a_type == SurfaceType::Color || a_type == SurfaceType::Texture) &&
(b_type == SurfaceType::Color || b_type == SurfaceType::Texture)) {
return true;
}
if (a_type == SurfaceType::Depth && b_type == SurfaceType::Depth) {
return true;
}
if (a_type == SurfaceType::DepthStencil && b_type == SurfaceType::DepthStencil) {
return true;
}
return false;
}
static constexpr SurfaceType GetFormatType(PixelFormat pixel_format) {
if ((unsigned int)pixel_format < 5) {
return SurfaceType::Color;
}
if ((unsigned int)pixel_format < 14) {
return SurfaceType::Texture;
}
if (pixel_format == PixelFormat::D16 || pixel_format == PixelFormat::D24) {
return SurfaceType::Depth;
}
if (pixel_format == PixelFormat::D24S8) {
return SurfaceType::DepthStencil;
}
return SurfaceType::Invalid;
}
/// Update the params "size", "end" and "type" from the already set "addr", "width", "height"
/// and "pixel_format"
void UpdateParams() {
if (stride == 0) {
stride = width;
}
type = GetFormatType(pixel_format);
size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
: BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
end = addr + size;
}
SurfaceInterval GetInterval() const {
return SurfaceInterval(addr, end);
}
// Returns the outer rectangle containing "interval"
SurfaceParams FromInterval(SurfaceInterval interval) const;
SurfaceInterval GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const;
// Returns the region of the biggest valid rectange within interval
SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
u32 GetScaledWidth() const {
return width * res_scale;
}
u32 GetScaledHeight() const {
return height * res_scale;
}
Common::Rectangle<u32> GetRect() const {
return {0, height, width, 0};
}
Common::Rectangle<u32> GetScaledRect() const {
return {0, GetScaledHeight(), GetScaledWidth(), 0};
}
u32 PixelsInBytes(u32 size) const {
return size * CHAR_BIT / GetFormatBpp(pixel_format);
}
u32 BytesInPixels(u32 pixels) const {
return pixels * GetFormatBpp(pixel_format) / CHAR_BIT;
}
bool ExactMatch(const SurfaceParams& other_surface) const;
bool CanSubRect(const SurfaceParams& sub_surface) const;
bool CanExpand(const SurfaceParams& expanded_surface) const;
bool CanTexCopy(const SurfaceParams& texcopy_params) const;
Common::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
Common::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
PAddr addr = 0;
PAddr end = 0;
u32 size = 0;
u32 width = 0;
u32 height = 0;
u32 stride = 0;
u16 res_scale = 1;
bool is_tiled = false;
PixelFormat pixel_format = PixelFormat::Invalid;
SurfaceType type = SurfaceType::Invalid;
};
/**
* A watcher that notifies whether a cached surface has been changed. This is useful for caching
* surface collection objects, including texture cube and mipmap.
@ -345,6 +138,8 @@ private:
};
struct CachedSurface : SurfaceParams, std::enable_shared_from_this<CachedSurface> {
CachedSurface(RasterizerCacheOpenGL& owner) : owner{owner} {}
bool CanFill(const SurfaceParams& dest_surface, SurfaceInterval fill_interval) const;
bool CanCopy(const SurfaceParams& dest_surface, SurfaceInterval copy_interval) const;
@ -422,6 +217,7 @@ struct CachedSurface : SurfaceParams, std::enable_shared_from_this<CachedSurface
}
private:
RasterizerCacheOpenGL& owner;
std::list<std::weak_ptr<SurfaceWatcher>> watchers;
};
@ -445,9 +241,6 @@ public:
bool BlitSurfaces(const Surface& src_surface, const Common::Rectangle<u32>& src_rect,
const Surface& dst_surface, const Common::Rectangle<u32>& dst_rect);
void ConvertD24S8toABGR(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect);
/// Copy one surface's region to another
void CopySurface(const Surface& src_surface, const Surface& dst_surface,
SurfaceInterval copy_interval);
@ -496,6 +289,18 @@ private:
/// Update surface's texture for given region when necessary
void ValidateSurface(const Surface& surface, PAddr addr, u32 size);
// Returns false if there is a surface in the cache at the interval with the same bit-width,
bool NoUnimplementedReinterpretations(const OpenGL::Surface& surface,
OpenGL::SurfaceParams& params,
const OpenGL::SurfaceInterval& interval);
// Return true if a surface with an invalid pixel format exists at the interval
bool IntervalHasInvalidPixelFormat(SurfaceParams& params, const SurfaceInterval& interval);
// Attempt to find a reinterpretable surface in the cache and use it to copy for validation
bool ValidateByReinterpretation(const Surface& surface, SurfaceParams& params,
const SurfaceInterval& interval);
/// Create a new surface
Surface CreateSurface(const SurfaceParams& params);
@ -516,14 +321,13 @@ private:
OGLFramebuffer read_framebuffer;
OGLFramebuffer draw_framebuffer;
OGLVertexArray attributeless_vao;
OGLBuffer d24s8_abgr_buffer;
GLsizeiptr d24s8_abgr_buffer_size;
OGLProgram d24s8_abgr_shader;
GLint d24s8_abgr_tbo_size_u_id;
GLint d24s8_abgr_viewport_u_id;
u16 resolution_scale_factor;
std::unordered_map<TextureCubeConfig, CachedTextureCube> texture_cube_cache;
public:
std::unique_ptr<TextureFilterer> texture_filterer;
std::unique_ptr<FormatReinterpreterOpenGL> format_reinterpreter;
};
struct FormatTuple {

14
src/video_core/renderer_opengl/gl_state.h
View file

@ -38,13 +38,6 @@ constexpr GLuint ShadowTexturePZ = 5;
constexpr GLuint ShadowTextureNZ = 6;
} // namespace ImageUnits
struct Viewport {
GLint x;
GLint y;
GLsizei width;
GLsizei height;
};
class OpenGLState {
public:
struct {
@ -142,7 +135,12 @@ public:
GLsizei height;
} scissor;
Viewport viewport;
struct {
GLint x;
GLint y;
GLsizei width;
GLsizei height;
} viewport;
std::array<bool, 2> clip_distance; // GL_CLIP_DISTANCE

171
src/video_core/renderer_opengl/gl_surface_params.cpp Normal file
View file

@ -0,0 +1,171 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/alignment.h"
#include "video_core/renderer_opengl/gl_rasterizer_cache.h"
#include "video_core/renderer_opengl/gl_surface_params.h"
namespace OpenGL {
SurfaceParams SurfaceParams::FromInterval(SurfaceInterval interval) const {
SurfaceParams params = *this;
const u32 tiled_size = is_tiled ? 8 : 1;
const u32 stride_tiled_bytes = BytesInPixels(stride * tiled_size);
PAddr aligned_start =
addr + Common::AlignDown(boost::icl::first(interval) - addr, stride_tiled_bytes);
PAddr aligned_end =
addr + Common::AlignUp(boost::icl::last_next(interval) - addr, stride_tiled_bytes);
if (aligned_end - aligned_start > stride_tiled_bytes) {
params.addr = aligned_start;
params.height = (aligned_end - aligned_start) / BytesInPixels(stride);
} else {
// 1 row
ASSERT(aligned_end - aligned_start == stride_tiled_bytes);
const u32 tiled_alignment = BytesInPixels(is_tiled ? 8 * 8 : 1);
aligned_start =
addr + Common::AlignDown(boost::icl::first(interval) - addr, tiled_alignment);
aligned_end =
addr + Common::AlignUp(boost::icl::last_next(interval) - addr, tiled_alignment);
params.addr = aligned_start;
params.width = PixelsInBytes(aligned_end - aligned_start) / tiled_size;
params.stride = params.width;
params.height = tiled_size;
}
params.UpdateParams();
return params;
}
SurfaceInterval SurfaceParams::GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const {
if (unscaled_rect.GetHeight() == 0 || unscaled_rect.GetWidth() == 0) {
return {};
}
if (is_tiled) {
unscaled_rect.left = Common::AlignDown(unscaled_rect.left, 8) * 8;
unscaled_rect.bottom = Common::AlignDown(unscaled_rect.bottom, 8) / 8;
unscaled_rect.right = Common::AlignUp(unscaled_rect.right, 8) * 8;
unscaled_rect.top = Common::AlignUp(unscaled_rect.top, 8) / 8;
}
const u32 stride_tiled = !is_tiled ? stride : stride * 8;
const u32 pixel_offset =
stride_tiled * (!is_tiled ? unscaled_rect.bottom : (height / 8) - unscaled_rect.top) +
unscaled_rect.left;
const u32 pixels = (unscaled_rect.GetHeight() - 1) * stride_tiled + unscaled_rect.GetWidth();
return {addr + BytesInPixels(pixel_offset), addr + BytesInPixels(pixel_offset + pixels)};
}
SurfaceInterval SurfaceParams::GetCopyableInterval(const Surface& src_surface) const {
SurfaceInterval result{};
const auto valid_regions =
SurfaceRegions(GetInterval() & src_surface->GetInterval()) - src_surface->invalid_regions;
for (auto& valid_interval : valid_regions) {
const SurfaceInterval aligned_interval{
addr + Common::AlignUp(boost::icl::first(valid_interval) - addr,
BytesInPixels(is_tiled ? 8 * 8 : 1)),
addr + Common::AlignDown(boost::icl::last_next(valid_interval) - addr,
BytesInPixels(is_tiled ? 8 * 8 : 1))};
if (BytesInPixels(is_tiled ? 8 * 8 : 1) > boost::icl::length(valid_interval) ||
boost::icl::length(aligned_interval) == 0) {
continue;
}
// Get the rectangle within aligned_interval
const u32 stride_bytes = BytesInPixels(stride) * (is_tiled ? 8 : 1);
SurfaceInterval rect_interval{
addr + Common::AlignUp(boost::icl::first(aligned_interval) - addr, stride_bytes),
addr + Common::AlignDown(boost::icl::last_next(aligned_interval) - addr, stride_bytes),
};
if (boost::icl::first(rect_interval) > boost::icl::last_next(rect_interval)) {
// 1 row
rect_interval = aligned_interval;
} else if (boost::icl::length(rect_interval) == 0) {
// 2 rows that do not make a rectangle, return the larger one
const SurfaceInterval row1{boost::icl::first(aligned_interval),
boost::icl::first(rect_interval)};
const SurfaceInterval row2{boost::icl::first(rect_interval),
boost::icl::last_next(aligned_interval)};
rect_interval = (boost::icl::length(row1) > boost::icl::length(row2)) ? row1 : row2;
}
if (boost::icl::length(rect_interval) > boost::icl::length(result)) {
result = rect_interval;
}
}
return result;
}
Common::Rectangle<u32> SurfaceParams::GetSubRect(const SurfaceParams& sub_surface) const {
const u32 begin_pixel_index = PixelsInBytes(sub_surface.addr - addr);
if (is_tiled) {
const int x0 = (begin_pixel_index % (stride * 8)) / 8;
const int y0 = (begin_pixel_index / (stride * 8)) * 8;
// Top to bottom
return Common::Rectangle<u32>(x0, height - y0, x0 + sub_surface.width,
height - (y0 + sub_surface.height));
}
const int x0 = begin_pixel_index % stride;
const int y0 = begin_pixel_index / stride;
// Bottom to top
return Common::Rectangle<u32>(x0, y0 + sub_surface.height, x0 + sub_surface.width, y0);
}
Common::Rectangle<u32> SurfaceParams::GetScaledSubRect(const SurfaceParams& sub_surface) const {
auto rect = GetSubRect(sub_surface);
rect.left = rect.left * res_scale;
rect.right = rect.right * res_scale;
rect.top = rect.top * res_scale;
rect.bottom = rect.bottom * res_scale;
return rect;
}
bool SurfaceParams::ExactMatch(const SurfaceParams& other_surface) const {
return std::tie(other_surface.addr, other_surface.width, other_surface.height,
other_surface.stride, other_surface.pixel_format, other_surface.is_tiled) ==
std::tie(addr, width, height, stride, pixel_format, is_tiled) &&
pixel_format != PixelFormat::Invalid;
}
bool SurfaceParams::CanSubRect(const SurfaceParams& sub_surface) const {
return sub_surface.addr >= addr && sub_surface.end <= end &&
sub_surface.pixel_format == pixel_format && pixel_format != PixelFormat::Invalid &&
sub_surface.is_tiled == is_tiled &&
(sub_surface.addr - addr) % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
(sub_surface.stride == stride || sub_surface.height <= (is_tiled ? 8u : 1u)) &&
GetSubRect(sub_surface).right <= stride;
}
bool SurfaceParams::CanExpand(const SurfaceParams& expanded_surface) const {
return pixel_format != PixelFormat::Invalid && pixel_format == expanded_surface.pixel_format &&
addr <= expanded_surface.end && expanded_surface.addr <= end &&
is_tiled == expanded_surface.is_tiled && stride == expanded_surface.stride &&
(std::max(expanded_surface.addr, addr) - std::min(expanded_surface.addr, addr)) %
BytesInPixels(stride * (is_tiled ? 8 : 1)) ==
0;
}
bool SurfaceParams::CanTexCopy(const SurfaceParams& texcopy_params) const {
if (pixel_format == PixelFormat::Invalid || addr > texcopy_params.addr ||
end < texcopy_params.end) {
return false;
}
if (texcopy_params.width != texcopy_params.stride) {
const u32 tile_stride = BytesInPixels(stride * (is_tiled ? 8 : 1));
return (texcopy_params.addr - addr) % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
texcopy_params.width % BytesInPixels(is_tiled ? 64 : 1) == 0 &&
(texcopy_params.height == 1 || texcopy_params.stride == tile_stride) &&
((texcopy_params.addr - addr) % tile_stride) + texcopy_params.width <= tile_stride;
}
return FromInterval(texcopy_params.GetInterval()).GetInterval() == texcopy_params.GetInterval();
}
} // namespace OpenGL

270
src/video_core/renderer_opengl/gl_surface_params.h Normal file
View file

@ -0,0 +1,270 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <array>
#include <climits>
#include <boost/icl/interval.hpp>
#include "common/assert.h"
#include "common/math_util.h"
#include "core/hw/gpu.h"
#include "video_core/regs_framebuffer.h"
#include "video_core/regs_texturing.h"
namespace OpenGL {
struct CachedSurface;
using Surface = std::shared_ptr<CachedSurface>;
using SurfaceInterval = boost::icl::right_open_interval<PAddr>;
struct SurfaceParams {
private:
static constexpr std::array<unsigned int, 18> BPP_TABLE = {
32, // RGBA8
24, // RGB8
16, // RGB5A1
16, // RGB565
16, // RGBA4
16, // IA8
16, // RG8
8, // I8
8, // A8
8, // IA4
4, // I4
4, // A4
4, // ETC1
8, // ETC1A4
16, // D16
0,
24, // D24
32, // D24S8
};
public:
enum class PixelFormat {
// First 5 formats are shared between textures and color buffers
RGBA8 = 0,
RGB8 = 1,
RGB5A1 = 2,
RGB565 = 3,
RGBA4 = 4,
// Texture-only formats
IA8 = 5,
RG8 = 6,
I8 = 7,
A8 = 8,
IA4 = 9,
I4 = 10,
A4 = 11,
ETC1 = 12,
ETC1A4 = 13,
// Depth buffer-only formats
D16 = 14,
// gap
D24 = 16,
D24S8 = 17,
Invalid = 255,
};
enum class SurfaceType {
Color = 0,
Texture = 1,
Depth = 2,
DepthStencil = 3,
Fill = 4,
Invalid = 5
};
static constexpr unsigned int GetFormatBpp(PixelFormat format) {
const auto format_idx = static_cast<std::size_t>(format);
DEBUG_ASSERT_MSG(format_idx < BPP_TABLE.size(), "Invalid pixel format {}", format_idx);
return BPP_TABLE[format_idx];
}
unsigned int GetFormatBpp() const {
return GetFormatBpp(pixel_format);
}
static std::string_view PixelFormatAsString(PixelFormat format) {
switch (format) {
case PixelFormat::RGBA8:
return "RGBA8";
case PixelFormat::RGB8:
return "RGB8";
case PixelFormat::RGB5A1:
return "RGB5A1";
case PixelFormat::RGB565:
return "RGB565";
case PixelFormat::RGBA4:
return "RGBA4";
case PixelFormat::IA8:
return "IA8";
case PixelFormat::RG8:
return "RG8";
case PixelFormat::I8:
return "I8";
case PixelFormat::A8:
return "A8";
case PixelFormat::IA4:
return "IA4";
case PixelFormat::I4:
return "I4";
case PixelFormat::A4:
return "A4";
case PixelFormat::ETC1:
return "ETC1";
case PixelFormat::ETC1A4:
return "ETC1A4";
case PixelFormat::D16:
return "D16";
case PixelFormat::D24:
return "D24";
case PixelFormat::D24S8:
return "D24S8";
default:
return "Not a real pixel format";
}
}
static PixelFormat PixelFormatFromTextureFormat(Pica::TexturingRegs::TextureFormat format) {
return ((unsigned int)format < 14) ? (PixelFormat)format : PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromColorFormat(Pica::FramebufferRegs::ColorFormat format) {
return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromDepthFormat(Pica::FramebufferRegs::DepthFormat format) {
return ((unsigned int)format < 4) ? (PixelFormat)((unsigned int)format + 14)
: PixelFormat::Invalid;
}
static PixelFormat PixelFormatFromGPUPixelFormat(GPU::Regs::PixelFormat format) {
switch (format) {
// RGB565 and RGB5A1 are switched in PixelFormat compared to ColorFormat
case GPU::Regs::PixelFormat::RGB565:
return PixelFormat::RGB565;
case GPU::Regs::PixelFormat::RGB5A1:
return PixelFormat::RGB5A1;
default:
return ((unsigned int)format < 5) ? (PixelFormat)format : PixelFormat::Invalid;
}
}
static bool CheckFormatsBlittable(PixelFormat pixel_format_a, PixelFormat pixel_format_b) {
SurfaceType a_type = GetFormatType(pixel_format_a);
SurfaceType b_type = GetFormatType(pixel_format_b);
if ((a_type == SurfaceType::Color || a_type == SurfaceType::Texture) &&
(b_type == SurfaceType::Color || b_type == SurfaceType::Texture)) {
return true;
}
if (a_type == SurfaceType::Depth && b_type == SurfaceType::Depth) {
return true;
}
if (a_type == SurfaceType::DepthStencil && b_type == SurfaceType::DepthStencil) {
return true;
}
return false;
}
static constexpr SurfaceType GetFormatType(PixelFormat pixel_format) {
if ((unsigned int)pixel_format < 5) {
return SurfaceType::Color;
}
if ((unsigned int)pixel_format < 14) {
return SurfaceType::Texture;
}
if (pixel_format == PixelFormat::D16 || pixel_format == PixelFormat::D24) {
return SurfaceType::Depth;
}
if (pixel_format == PixelFormat::D24S8) {
return SurfaceType::DepthStencil;
}
return SurfaceType::Invalid;
}
/// Update the params "size", "end" and "type" from the already set "addr", "width", "height"
/// and "pixel_format"
void UpdateParams() {
if (stride == 0) {
stride = width;
}
type = GetFormatType(pixel_format);
size = !is_tiled ? BytesInPixels(stride * (height - 1) + width)
: BytesInPixels(stride * 8 * (height / 8 - 1) + width * 8);
end = addr + size;
}
SurfaceInterval GetInterval() const {
return SurfaceInterval(addr, end);
}
// Returns the outer rectangle containing "interval"
SurfaceParams FromInterval(SurfaceInterval interval) const;
SurfaceInterval GetSubRectInterval(Common::Rectangle<u32> unscaled_rect) const;
// Returns the region of the biggest valid rectange within interval
SurfaceInterval GetCopyableInterval(const Surface& src_surface) const;
u32 GetScaledWidth() const {
return width * res_scale;
}
u32 GetScaledHeight() const {
return height * res_scale;
}
Common::Rectangle<u32> GetRect() const {
return {0, height, width, 0};
}
Common::Rectangle<u32> GetScaledRect() const {
return {0, GetScaledHeight(), GetScaledWidth(), 0};
}
u32 PixelsInBytes(u32 size) const {
return size * CHAR_BIT / GetFormatBpp(pixel_format);
}
u32 BytesInPixels(u32 pixels) const {
return pixels * GetFormatBpp(pixel_format) / CHAR_BIT;
}
bool ExactMatch(const SurfaceParams& other_surface) const;
bool CanSubRect(const SurfaceParams& sub_surface) const;
bool CanExpand(const SurfaceParams& expanded_surface) const;
bool CanTexCopy(const SurfaceParams& texcopy_params) const;
Common::Rectangle<u32> GetSubRect(const SurfaceParams& sub_surface) const;
Common::Rectangle<u32> GetScaledSubRect(const SurfaceParams& sub_surface) const;
PAddr addr = 0;
PAddr end = 0;
u32 size = 0;
u32 width = 0;
u32 height = 0;
u32 stride = 0;
u16 res_scale = 1;
bool is_tiled = false;
PixelFormat pixel_format = PixelFormat::Invalid;
SurfaceType type = SurfaceType::Invalid;
};
} // namespace OpenGL

282
src/video_core/renderer_opengl/renderer_opengl.cpp
View file

@ -32,23 +32,8 @@
#include "video_core/renderer_opengl/gl_vars.h"
#include "video_core/renderer_opengl/post_processing_opengl.h"
#include "video_core/renderer_opengl/renderer_opengl.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_manager.h"
#include "video_core/video_core.h"
namespace Frontend {
struct Frame {
u32 width{}; /// Width of the frame (to detect resize)
u32 height{}; /// Height of the frame
bool color_reloaded = false; /// Texture attachment was recreated (ie: resized)
OpenGL::OGLRenderbuffer color{}; /// Buffer shared between the render/present FBO
OpenGL::OGLFramebuffer render{}; /// FBO created on the render thread
OpenGL::OGLFramebuffer present{}; /// FBO created on the present thread
GLsync render_fence{}; /// Fence created on the render thread
GLsync present_fence{}; /// Fence created on the presentation thread
};
} // namespace Frontend
namespace OpenGL {
// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
@ -79,6 +64,7 @@ public:
std::queue<Frontend::Frame*>().swap(free_queue);
present_queue.clear();
present_cv.notify_all();
free_cv.notify_all();
}
void ReloadPresentFrame(Frontend::Frame* frame, u32 height, u32 width) override {
@ -89,7 +75,7 @@ public:
glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
frame->color.handle);
if (!glCheckFramebufferStatus(GL_FRAMEBUFFER) == GL_FRAMEBUFFER_COMPLETE) {
if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
}
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
@ -115,7 +101,7 @@ public:
state.Apply();
glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
frame->color.handle);
if (!glCheckFramebufferStatus(GL_FRAMEBUFFER) == GL_FRAMEBUFFER_COMPLETE) {
if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
}
prev_state.Apply();
@ -145,19 +131,12 @@ public:
present_cv.notify_one();
}
Frontend::Frame* TryGetPresentFrame(int timeout_ms) override {
std::unique_lock<std::mutex> lock(swap_chain_lock);
// wait for new entries in the present_queue
present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
[&] { return !present_queue.empty(); });
if (present_queue.empty()) {
// timed out waiting for a frame to draw so return the previous frame
return previous_frame;
}
// This is virtual as it is to be overriden in OGLVideoDumpingMailbox below.
virtual void LoadPresentFrame() {
// free the previous frame and add it back to the free queue
if (previous_frame) {
free_queue.push(previous_frame);
free_cv.notify_one();
}
// the newest entries are pushed to the front of the queue
@ -169,8 +148,72 @@ public:
}
present_queue.clear();
previous_frame = frame;
}
Frontend::Frame* TryGetPresentFrame(int timeout_ms) override {
std::unique_lock<std::mutex> lock(swap_chain_lock);
// wait for new entries in the present_queue
present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
[&] { return !present_queue.empty(); });
if (present_queue.empty()) {
// timed out waiting for a frame to draw so return the previous frame
return previous_frame;
}
LoadPresentFrame();
return previous_frame;
}
};
/// This mailbox is different in that it will never discard rendered frames
class OGLVideoDumpingMailbox : public OGLTextureMailbox {
public:
Frontend::Frame* GetRenderFrame() override {
std::unique_lock<std::mutex> lock(swap_chain_lock);
// If theres no free frames, we will wait until one shows up
if (free_queue.empty()) {
free_cv.wait(lock, [&] { return !free_queue.empty(); });
}
if (free_queue.empty()) {
LOG_CRITICAL(Render_OpenGL, "Could not get free frame");
return nullptr;
}
Frontend::Frame* frame = free_queue.front();
free_queue.pop();
return frame;
}
void LoadPresentFrame() override {
// free the previous frame and add it back to the free queue
if (previous_frame) {
free_queue.push(previous_frame);
free_cv.notify_one();
}
Frontend::Frame* frame = present_queue.back();
present_queue.pop_back();
previous_frame = frame;
// Do not remove entries from the present_queue, as video dumping would require
// that we preserve all frames
}
Frontend::Frame* TryGetPresentFrame(int timeout_ms) override {
std::unique_lock<std::mutex> lock(swap_chain_lock);
// wait for new entries in the present_queue
present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
[&] { return !present_queue.empty(); });
if (present_queue.empty()) {
// timed out waiting for a frame
return nullptr;
}
LoadPresentFrame();
return previous_frame;
}
};
static const char vertex_shader[] = R"(
@ -279,21 +322,35 @@ struct ScreenRectVertex {
*
* The projection part of the matrix is trivial, hence these operations are represented
* by a 3x2 matrix.
*
* @param flipped Whether the frame should be flipped upside down.
*/
static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, const float height) {
static std::array<GLfloat, 3 * 2> MakeOrthographicMatrix(const float width, const float height,
bool flipped) {
std::array<GLfloat, 3 * 2> matrix; // Laid out in column-major order
// clang-format off
matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f;
matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f;
// Last matrix row is implicitly assumed to be [0, 0, 1].
// clang-format on
if (flipped) {
// clang-format off
matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f;
matrix[1] = 0.f; matrix[3] = 2.f / height; matrix[5] = -1.f;
// clang-format on
} else {
// clang-format off
matrix[0] = 2.f / width; matrix[2] = 0.f; matrix[4] = -1.f;
matrix[1] = 0.f; matrix[3] = -2.f / height; matrix[5] = 1.f;
// clang-format on
}
return matrix;
}
RendererOpenGL::RendererOpenGL(Frontend::EmuWindow& window) : RendererBase{window} {
RendererOpenGL::RendererOpenGL(Frontend::EmuWindow& window)
: RendererBase{window}, frame_dumper(Core::System::GetInstance().VideoDumper(), window) {
window.mailbox = std::make_unique<OGLTextureMailbox>();
frame_dumper.mailbox = std::make_unique<OGLVideoDumpingMailbox>();
}
RendererOpenGL::~RendererOpenGL() = default;
@ -311,56 +368,14 @@ void RendererOpenGL::SwapBuffers() {
RenderScreenshot();
RenderVideoDumping();
const auto& layout = render_window.GetFramebufferLayout();
RenderToMailbox(layout, render_window.mailbox, false);
Frontend::Frame* frame;
{
MICROPROFILE_SCOPE(OpenGL_WaitPresent);
frame = render_window.mailbox->GetRenderFrame();
// Clean up sync objects before drawing
// INTEL driver workaround. We can't delete the previous render sync object until we are
// sure that the presentation is done
if (frame->present_fence) {
glClientWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
}
// delete the draw fence if the frame wasn't presented
if (frame->render_fence) {
glDeleteSync(frame->render_fence);
frame->render_fence = 0;
}
// wait for the presentation to be done
if (frame->present_fence) {
glWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
glDeleteSync(frame->present_fence);
frame->present_fence = 0;
}
if (frame_dumper.IsDumping()) {
RenderToMailbox(frame_dumper.GetLayout(), frame_dumper.mailbox, true);
}
{
MICROPROFILE_SCOPE(OpenGL_RenderFrame);
// Recreate the frame if the size of the window has changed
if (layout.width != frame->width || layout.height != frame->height) {
LOG_DEBUG(Render_OpenGL, "Reloading render frame");
render_window.mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
}
GLuint render_texture = frame->color.handle;
state.draw.draw_framebuffer = frame->render.handle;
state.Apply();
DrawScreens(layout);
// Create a fence for the frontend to wait on and swap this frame to OffTex
frame->render_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
glFlush();
render_window.mailbox->ReleaseRenderFrame(frame);
m_current_frame++;
}
m_current_frame++;
Core::System::GetInstance().perf_stats->EndSystemFrame();
@ -396,7 +411,7 @@ void RendererOpenGL::RenderScreenshot() {
glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
renderbuffer);
DrawScreens(layout);
DrawScreens(layout, false);
glReadPixels(0, 0, layout.width, layout.height, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV,
VideoCore::g_screenshot_bits);
@ -449,33 +464,54 @@ void RendererOpenGL::PrepareRendertarget() {
}
}
void RendererOpenGL::RenderVideoDumping() {
if (cleanup_video_dumping.exchange(false)) {
ReleaseVideoDumpingGLObjects();
}
void RendererOpenGL::RenderToMailbox(const Layout::FramebufferLayout& layout,
std::unique_ptr<Frontend::TextureMailbox>& mailbox,
bool flipped) {
if (Core::System::GetInstance().VideoDumper().IsDumping()) {
if (prepare_video_dumping.exchange(false)) {
InitVideoDumpingGLObjects();
Frontend::Frame* frame;
{
MICROPROFILE_SCOPE(OpenGL_WaitPresent);
frame = mailbox->GetRenderFrame();
// Clean up sync objects before drawing
// INTEL driver workaround. We can't delete the previous render sync object until we are
// sure that the presentation is done
if (frame->present_fence) {
glClientWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
}
const auto& layout = Core::System::GetInstance().VideoDumper().GetLayout();
glBindFramebuffer(GL_READ_FRAMEBUFFER, frame_dumping_framebuffer.handle);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, frame_dumping_framebuffer.handle);
DrawScreens(layout);
// delete the draw fence if the frame wasn't presented
if (frame->render_fence) {
glDeleteSync(frame->render_fence);
frame->render_fence = 0;
}
glBindBuffer(GL_PIXEL_PACK_BUFFER, frame_dumping_pbos[current_pbo].handle);
glReadPixels(0, 0, layout.width, layout.height, GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV, 0);
glBindBuffer(GL_PIXEL_PACK_BUFFER, frame_dumping_pbos[next_pbo].handle);
// wait for the presentation to be done
if (frame->present_fence) {
glWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
glDeleteSync(frame->present_fence);
frame->present_fence = 0;
}
}
GLubyte* pixels = static_cast<GLubyte*>(glMapBuffer(GL_PIXEL_PACK_BUFFER, GL_READ_ONLY));
VideoDumper::VideoFrame frame_data{layout.width, layout.height, pixels};
Core::System::GetInstance().VideoDumper().AddVideoFrame(frame_data);
{
MICROPROFILE_SCOPE(OpenGL_RenderFrame);
// Recreate the frame if the size of the window has changed
if (layout.width != frame->width || layout.height != frame->height) {
LOG_DEBUG(Render_OpenGL, "Reloading render frame");
mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
}
glUnmapBuffer(GL_PIXEL_PACK_BUFFER);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
current_pbo = (current_pbo + 1) % 2;
next_pbo = (current_pbo + 1) % 2;
GLuint render_texture = frame->color.handle;
state.draw.draw_framebuffer = frame->render.handle;
state.Apply();
DrawScreens(layout, flipped);
// Create a fence for the frontend to wait on and swap this frame to OffTex
frame->render_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
glFlush();
mailbox->ReleaseRenderFrame(frame);
}
}
@ -886,7 +922,7 @@ void RendererOpenGL::DrawSingleScreenStereo(const ScreenInfo& screen_info_l,
/**
* Draws the emulated screens to the emulator window.
*/
void RendererOpenGL::DrawScreens(const Layout::FramebufferLayout& layout) {
void RendererOpenGL::DrawScreens(const Layout::FramebufferLayout& layout, bool flipped) {
if (VideoCore::g_renderer_bg_color_update_requested.exchange(false)) {
// Update background color before drawing
glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
@ -913,7 +949,7 @@ void RendererOpenGL::DrawScreens(const Layout::FramebufferLayout& layout) {
// Set projection matrix
std::array<GLfloat, 3 * 2> ortho_matrix =
MakeOrthographicMatrix((float)layout.width, (float)layout.height);
MakeOrthographicMatrix((float)layout.width, (float)layout.height, flipped);
glUniformMatrix3x2fv(uniform_modelview_matrix, 1, GL_FALSE, ortho_matrix.data());
// Bind texture in Texture Unit 0
@ -1052,41 +1088,11 @@ void RendererOpenGL::TryPresent(int timeout_ms) {
void RendererOpenGL::UpdateFramerate() {}
void RendererOpenGL::PrepareVideoDumping() {
prepare_video_dumping = true;
frame_dumper.StartDumping();
}
void RendererOpenGL::CleanupVideoDumping() {
cleanup_video_dumping = true;
}
void RendererOpenGL::InitVideoDumpingGLObjects() {
const auto& layout = Core::System::GetInstance().VideoDumper().GetLayout();
frame_dumping_framebuffer.Create();
glGenRenderbuffers(1, &frame_dumping_renderbuffer);
glBindRenderbuffer(GL_RENDERBUFFER, frame_dumping_renderbuffer);
glRenderbufferStorage(GL_RENDERBUFFER, GL_RGB8, layout.width, layout.height);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, frame_dumping_framebuffer.handle);
glFramebufferRenderbuffer(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
frame_dumping_renderbuffer);
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, 0);
for (auto& buffer : frame_dumping_pbos) {
buffer.Create();
glBindBuffer(GL_PIXEL_PACK_BUFFER, buffer.handle);
glBufferData(GL_PIXEL_PACK_BUFFER, layout.width * layout.height * 4, nullptr,
GL_STREAM_READ);
glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
}
}
void RendererOpenGL::ReleaseVideoDumpingGLObjects() {
frame_dumping_framebuffer.Release();
glDeleteRenderbuffers(1, &frame_dumping_renderbuffer);
for (auto& buffer : frame_dumping_pbos) {
buffer.Release();
}
frame_dumper.StopDumping();
}
static const char* GetSource(GLenum source) {
@ -1179,14 +1185,10 @@ VideoCore::ResultStatus RendererOpenGL::Init() {
RefreshRasterizerSetting();
TextureFilterManager::GetInstance().Reset();
return VideoCore::ResultStatus::Success;
}
/// Shutdown the renderer
void RendererOpenGL::ShutDown() {
TextureFilterManager::GetInstance().Destroy();
}
void RendererOpenGL::ShutDown() {}
} // namespace OpenGL

37
src/video_core/renderer_opengl/renderer_opengl.h
View file

@ -10,6 +10,7 @@
#include "common/math_util.h"
#include "core/hw/gpu.h"
#include "video_core/renderer_base.h"
#include "video_core/renderer_opengl/frame_dumper_opengl.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_state.h"
@ -17,6 +18,20 @@ namespace Layout {
struct FramebufferLayout;
}
namespace Frontend {
struct Frame {
u32 width{}; /// Width of the frame (to detect resize)
u32 height{}; /// Height of the frame
bool color_reloaded = false; /// Texture attachment was recreated (ie: resized)
OpenGL::OGLRenderbuffer color{}; /// Buffer shared between the render/present FBO
OpenGL::OGLFramebuffer render{}; /// FBO created on the render thread
OpenGL::OGLFramebuffer present{}; /// FBO created on the present thread
GLsync render_fence{}; /// Fence created on the render thread
GLsync present_fence{}; /// Fence created on the presentation thread
};
} // namespace Frontend
namespace OpenGL {
/// Structure used for storing information about the textures for each 3DS screen
@ -72,10 +87,11 @@ private:
void ReloadShader();
void PrepareRendertarget();
void RenderScreenshot();
void RenderVideoDumping();
void RenderToMailbox(const Layout::FramebufferLayout& layout,
std::unique_ptr<Frontend::TextureMailbox>& mailbox, bool flipped);
void ConfigureFramebufferTexture(TextureInfo& texture,
const GPU::Regs::FramebufferConfig& framebuffer);
void DrawScreens(const Layout::FramebufferLayout& layout);
void DrawScreens(const Layout::FramebufferLayout& layout, bool flipped);
void DrawSingleScreenRotated(const ScreenInfo& screen_info, float x, float y, float w, float h);
void DrawSingleScreen(const ScreenInfo& screen_info, float x, float y, float w, float h);
void DrawSingleScreenStereoRotated(const ScreenInfo& screen_info_l,
@ -91,9 +107,6 @@ private:
// Fills active OpenGL texture with the given RGB color.
void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, const TextureInfo& texture);
void InitVideoDumpingGLObjects();
void ReleaseVideoDumpingGLObjects();
OpenGLState state;
// OpenGL object IDs
@ -120,19 +133,7 @@ private:
GLuint attrib_position;
GLuint attrib_tex_coord;
// Frame dumping
OGLFramebuffer frame_dumping_framebuffer;
GLuint frame_dumping_renderbuffer;
// Whether prepare/cleanup video dumping has been requested.
// They will be executed on next frame.
std::atomic_bool prepare_video_dumping = false;
std::atomic_bool cleanup_video_dumping = false;
// PBOs used to dump frames faster
std::array<OGLBuffer, 2> frame_dumping_pbos;
GLuint current_pbo = 1;
GLuint next_pbo = 0;
FrameDumperOpenGL frame_dumper;
};
} // namespace OpenGL

45
src/video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.cpp
View file

@ -42,18 +42,17 @@
namespace OpenGL {
Anime4kUltrafast::Anime4kUltrafast(u16 scale_factor) : TextureFilterInterface(scale_factor) {
Anime4kUltrafast::Anime4kUltrafast(u16 scale_factor) : TextureFilterBase(scale_factor) {
const OpenGLState cur_state = OpenGLState::GetCurState();
const auto setup_temp_tex = [this, scale_factor](TempTex& texture, GLint internal_format,
GLint format) {
const auto setup_temp_tex = [this](TempTex& texture, GLint internal_format, GLint format) {
texture.fbo.Create();
texture.tex.Create();
state.draw.draw_framebuffer = texture.fbo.handle;
state.Apply();
glActiveTexture(GL_TEXTURE0);
glBindTexture(GL_TEXTURE_RECTANGLE, texture.tex.handle);
glTexImage2D(GL_TEXTURE_RECTANGLE, 0, internal_format, 1024 * scale_factor,
1024 * scale_factor, 0, format, GL_HALF_FLOAT, nullptr);
glTexImage2D(GL_TEXTURE_RECTANGLE, 0, internal_format, 1024 * internal_scale_factor,
1024 * internal_scale_factor, 0, format, GL_HALF_FLOAT, nullptr);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_RECTANGLE,
texture.tex.handle, 0);
};
@ -61,7 +60,6 @@ Anime4kUltrafast::Anime4kUltrafast(u16 scale_factor) : TextureFilterInterface(sc
setup_temp_tex(XY, GL_RG16F, GL_RG);
vao.Create();
out_fbo.Create();
for (std::size_t idx = 0; idx < samplers.size(); ++idx) {
samplers[idx].Create();
@ -86,30 +84,26 @@ Anime4kUltrafast::Anime4kUltrafast(u16 scale_factor) : TextureFilterInterface(sc
state.draw.shader_program = refine_program.handle;
state.Apply();
glUniform1i(glGetUniformLocation(refine_program.handle, "LUMAD"), 1);
glUniform1f(glGetUniformLocation(refine_program.handle, "final_scale"),
static_cast<GLfloat>(internal_scale_factor) / scale_factor);
cur_state.Apply();
}
void Anime4kUltrafast::scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) {
void Anime4kUltrafast::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect,
GLuint dst_tex, const Common::Rectangle<u32>& dst_rect,
GLuint read_fb_handle, GLuint draw_fb_handle) {
const OpenGLState cur_state = OpenGLState::GetCurState();
OGLTexture src_tex;
src_tex.Create();
state.viewport = RectToViewport(rect);
state.texture_units[0].texture_2d = src_tex.handle;
state.viewport = {static_cast<GLint>(src_rect.left * internal_scale_factor),
static_cast<GLint>(src_rect.bottom * internal_scale_factor),
static_cast<GLsizei>(src_rect.GetWidth() * internal_scale_factor),
static_cast<GLsizei>(src_rect.GetHeight() * internal_scale_factor)};
state.texture_units[0].texture_2d = src_tex;
state.draw.draw_framebuffer = XY.fbo.handle;
state.draw.shader_program = gradient_x_program.handle;
state.Apply();
const FormatTuple tuple = GetFormatTuple(surface.pixel_format);
glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(surface.stride));
glActiveTexture(GL_TEXTURE0);
glTexImage2D(GL_TEXTURE_2D, 0, tuple.internal_format, rect.GetWidth(), rect.GetHeight(), 0,
tuple.format, tuple.type, &surface.gl_buffer[buffer_offset]);
glActiveTexture(GL_TEXTURE1);
glBindTexture(GL_TEXTURE_RECTANGLE, LUMAD.tex.handle);
glActiveTexture(GL_TEXTURE2);
@ -124,14 +118,17 @@ void Anime4kUltrafast::scale(CachedSurface& surface, const Common::Rectangle<u32
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
// refine pass
state.draw.draw_framebuffer = out_fbo.handle;
state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
static_cast<GLsizei>(dst_rect.GetWidth()),
static_cast<GLsizei>(dst_rect.GetHeight())};
state.draw.draw_framebuffer = draw_fb_handle;
state.draw.shader_program = refine_program.handle;
state.Apply();
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
cur_state.texture_units[0].texture_2d, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
cur_state.Apply();
}

22
src/video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h
View file

@ -6,29 +6,25 @@
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_interface.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
namespace OpenGL {
class Anime4kUltrafast : public TextureFilterInterface {
class Anime4kUltrafast : public TextureFilterBase {
public:
static TextureFilterInfo GetInfo() {
TextureFilterInfo info;
info.name = "Anime4K Ultrafast";
info.clamp_scale = {2, 2};
info.constructor = std::make_unique<Anime4kUltrafast, u16>;
return info;
}
static constexpr std::string_view NAME = "Anime4K Ultrafast";
Anime4kUltrafast(u16 scale_factor);
void scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) override;
explicit Anime4kUltrafast(u16 scale_factor);
void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) override;
private:
static constexpr u8 internal_scale_factor = 2;
OpenGLState state{};
OGLVertexArray vao;
OGLFramebuffer out_fbo;
struct TempTex {
OGLTexture tex;

4
src/video_core/renderer_opengl/texture_filters/anime4k/refine.frag
View file

@ -8,6 +8,8 @@ uniform sampler2D HOOKED;
uniform sampler2DRect LUMAD;
uniform sampler2DRect LUMAG;
uniform float final_scale;
const float LINE_DETECT_THRESHOLD = 0.4;
const float STRENGTH = 0.6;
@ -24,7 +26,7 @@ vec4 getAverage(vec4 cc, vec4 a, vec4 b, vec4 c) {
#define GetRGBAL(offset) \
RGBAL(textureOffset(HOOKED, tex_coord, offset), \
texture(LUMAD, clamp(gl_FragCoord.xy + offset, vec2(0.0), input_max)).x)
texture(LUMAD, clamp((gl_FragCoord.xy + offset) * final_scale, vec2(0.0), input_max)).x)
float min3v(float a, float b, float c) {
return min(min(a, b), c);

33
src/video_core/renderer_opengl/texture_filters/bicubic/bicubic.cpp
View file

@ -10,45 +10,36 @@
namespace OpenGL {
Bicubic::Bicubic(u16 scale_factor) : TextureFilterInterface(scale_factor) {
Bicubic::Bicubic(u16 scale_factor) : TextureFilterBase(scale_factor) {
program.Create(tex_coord_vert.data(), bicubic_frag.data());
vao.Create();
draw_fbo.Create();
src_sampler.Create();
state.draw.shader_program = program.handle;
state.draw.vertex_array = vao.handle;
state.draw.shader_program = program.handle;
state.draw.draw_framebuffer = draw_fbo.handle;
state.texture_units[0].sampler = src_sampler.handle;
glSamplerParameteri(src_sampler.handle, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glSamplerParameteri(src_sampler.handle, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glSamplerParameteri(src_sampler.handle, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
}
} // namespace OpenGL
void Bicubic::scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) {
void Bicubic::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) {
const OpenGLState cur_state = OpenGLState::GetCurState();
OGLTexture src_tex;
src_tex.Create();
state.texture_units[0].texture_2d = src_tex.handle;
state.viewport = RectToViewport(rect);
state.texture_units[0].texture_2d = src_tex;
state.draw.draw_framebuffer = draw_fb_handle;
state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
static_cast<GLsizei>(dst_rect.GetWidth()),
static_cast<GLsizei>(dst_rect.GetHeight())};
state.Apply();
const FormatTuple tuple = GetFormatTuple(surface.pixel_format);
glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(surface.stride));
glActiveTexture(GL_TEXTURE0);
glTexImage2D(GL_TEXTURE_2D, 0, tuple.internal_format, rect.GetWidth(), rect.GetHeight(), 0,
tuple.format, tuple.type, &surface.gl_buffer[buffer_offset]);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
cur_state.texture_units[0].texture_2d, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
cur_state.Apply();
}

23
src/video_core/renderer_opengl/texture_filters/bicubic/bicubic.h
View file

@ -6,27 +6,24 @@
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_interface.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
namespace OpenGL {
class Bicubic : public TextureFilterInterface {
public:
static TextureFilterInfo GetInfo() {
TextureFilterInfo info;
info.name = "Bicubic";
info.constructor = std::make_unique<Bicubic, u16>;
return info;
}
Bicubic(u16 scale_factor);
void scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) override;
class Bicubic : public TextureFilterBase {
public:
static constexpr std::string_view NAME = "Bicubic";
explicit Bicubic(u16 scale_factor);
void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) override;
private:
OpenGLState state{};
OGLProgram program{};
OGLVertexArray vao{};
OGLFramebuffer draw_fbo{};
OGLSampler src_sampler{};
};
} // namespace OpenGL

26
src/video_core/renderer_opengl/texture_filters/texture_filter_base.h Normal file
View file

@ -0,0 +1,26 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include "common/common_types.h"
#include "common/math_util.h"
#include "video_core/renderer_opengl/gl_surface_params.h"
namespace OpenGL {
class TextureFilterBase {
friend class TextureFilterer;
virtual void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) = 0;
public:
explicit TextureFilterBase(u16 scale_factor) : scale_factor{scale_factor} {};
virtual ~TextureFilterBase() = default;
const u16 scale_factor{};
};
} // namespace OpenGL

38
src/video_core/renderer_opengl/texture_filters/texture_filter_interface.h
View file

@ -1,38 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <functional>
#include <string_view>
#include "common/common_types.h"
#include "common/math_util.h"
namespace OpenGL {
struct CachedSurface;
struct Viewport;
class TextureFilterInterface {
public:
const u16 scale_factor{};
TextureFilterInterface(u16 scale_factor) : scale_factor{scale_factor} {}
virtual void scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) = 0;
virtual ~TextureFilterInterface() = default;
protected:
Viewport RectToViewport(const Common::Rectangle<u32>& rect);
};
// every texture filter should have a static GetInfo function
struct TextureFilterInfo {
std::string_view name;
struct {
u16 min, max;
} clamp_scale{1, 10};
std::function<std::unique_ptr<TextureFilterInterface>(u16 scale_factor)> constructor;
};
} // namespace OpenGL

89
src/video_core/renderer_opengl/texture_filters/texture_filter_manager.cpp
View file

@ -1,89 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include "common/logging/log.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h"
#include "video_core/renderer_opengl/texture_filters/bicubic/bicubic.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_manager.h"
#include "video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.h"
namespace OpenGL {
Viewport TextureFilterInterface::RectToViewport(const Common::Rectangle<u32>& rect) {
return {
static_cast<GLint>(rect.left) * scale_factor,
static_cast<GLint>(rect.top) * scale_factor,
static_cast<GLsizei>(rect.GetWidth()) * scale_factor,
static_cast<GLsizei>(rect.GetHeight()) * scale_factor,
};
}
namespace {
template <typename T>
std::pair<std::string_view, TextureFilterInfo> FilterMapPair() {
return {T::GetInfo().name, T::GetInfo()};
};
struct NoFilter {
static TextureFilterInfo GetInfo() {
TextureFilterInfo info;
info.name = TextureFilterManager::NONE;
info.clamp_scale = {1, 1};
info.constructor = [](u16) { return nullptr; };
return info;
}
};
} // namespace
const std::map<std::string_view, TextureFilterInfo, TextureFilterManager::FilterNameComp>&
TextureFilterManager::TextureFilterMap() {
static const std::map<std::string_view, TextureFilterInfo, FilterNameComp> filter_map{
FilterMapPair<NoFilter>(),
FilterMapPair<Anime4kUltrafast>(),
FilterMapPair<Bicubic>(),
FilterMapPair<XbrzFreescale>(),
};
return filter_map;
}
void TextureFilterManager::SetTextureFilter(std::string filter_name, u16 new_scale_factor) {
if (name == filter_name && scale_factor == new_scale_factor)
return;
std::lock_guard<std::mutex> lock{mutex};
name = std::move(filter_name);
scale_factor = new_scale_factor;
updated = true;
}
TextureFilterInterface* TextureFilterManager::GetTextureFilter() const {
return filter.get();
}
bool TextureFilterManager::IsUpdated() const {
return updated;
}
void TextureFilterManager::Reset() {
std::lock_guard<std::mutex> lock{mutex};
updated = false;
auto iter = TextureFilterMap().find(name);
if (iter == TextureFilterMap().end()) {
LOG_ERROR(Render_OpenGL, "Invalid texture filter: {}", name);
filter = nullptr;
return;
}
const auto& filter_info = iter->second;
u16 clamped_scale =
std::clamp(scale_factor, filter_info.clamp_scale.min, filter_info.clamp_scale.max);
if (clamped_scale != scale_factor)
LOG_ERROR(Render_OpenGL, "Invalid scale factor {} for texture filter {}, clamped to {}",
scale_factor, filter_info.name, clamped_scale);
filter = filter_info.constructor(clamped_scale);
}
} // namespace OpenGL

55
src/video_core/renderer_opengl/texture_filters/texture_filter_manager.h
View file

@ -1,55 +0,0 @@
// Copyright 2019 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <atomic>
#include <map>
#include <memory>
#include <mutex>
#include <string_view>
#include <tuple>
#include "video_core/renderer_opengl/texture_filters/texture_filter_interface.h"
namespace OpenGL {
class TextureFilterManager {
public:
static constexpr std::string_view NONE = "none";
struct FilterNameComp {
bool operator()(const std::string_view a, const std::string_view b) const {
bool na = a == NONE;
bool nb = b == NONE;
if (na | nb)
return na & !nb;
return a < b;
}
};
// function ensures map is initialized before use
static const std::map<std::string_view, TextureFilterInfo, FilterNameComp>& TextureFilterMap();
static TextureFilterManager& GetInstance() {
static TextureFilterManager singleton;
return singleton;
}
void Destroy() {
filter.reset();
}
void SetTextureFilter(std::string filter_name, u16 new_scale_factor);
TextureFilterInterface* GetTextureFilter() const;
// returns true if filter has been changed and a cache reset is needed
bool IsUpdated() const;
void Reset();
private:
std::atomic<bool> updated{false};
std::mutex mutex;
std::string name{"none"};
u16 scale_factor{1};
std::unique_ptr<TextureFilterInterface> filter;
};
} // namespace OpenGL

86
src/video_core/renderer_opengl/texture_filters/texture_filterer.cpp Normal file
View file

@ -0,0 +1,86 @@
/// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <algorithm>
#include <functional>
#include <unordered_map>
#include "common/logging/log.h"
#include "video_core/renderer_opengl/texture_filters/anime4k/anime4k_ultrafast.h"
#include "video_core/renderer_opengl/texture_filters/bicubic/bicubic.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
#include "video_core/renderer_opengl/texture_filters/texture_filterer.h"
#include "video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.h"
namespace OpenGL {
namespace {
using TextureFilterContructor = std::function<std::unique_ptr<TextureFilterBase>(u16)>;
template <typename T>
std::pair<std::string_view, TextureFilterContructor> FilterMapPair() {
return {T::NAME, std::make_unique<T, u16>};
};
static const std::unordered_map<std::string_view, TextureFilterContructor> filter_map{
{TextureFilterer::NONE, [](u16) { return nullptr; }},
FilterMapPair<Anime4kUltrafast>(),
FilterMapPair<Bicubic>(),
FilterMapPair<XbrzFreescale>(),
};
} // namespace
TextureFilterer::TextureFilterer(std::string_view filter_name, u16 scale_factor) {
Reset(filter_name, scale_factor);
}
bool TextureFilterer::Reset(std::string_view new_filter_name, u16 new_scale_factor) {
if (filter_name == new_filter_name && (IsNull() || filter->scale_factor == new_scale_factor))
return false;
auto iter = filter_map.find(new_filter_name);
if (iter == filter_map.end()) {
LOG_ERROR(Render_OpenGL, "Invalid texture filter: {}", new_filter_name);
filter = nullptr;
return true;
}
filter_name = iter->first;
filter = iter->second(new_scale_factor);
return true;
}
bool TextureFilterer::IsNull() const {
return !filter;
}
bool TextureFilterer::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect,
SurfaceParams::SurfaceType type, GLuint read_fb_handle,
GLuint draw_fb_handle) {
// depth / stencil texture filtering is not supported for now
if (IsNull() ||
(type != SurfaceParams::SurfaceType::Color && type != SurfaceParams::SurfaceType::Texture))
return false;
filter->Filter(src_tex, src_rect, dst_tex, dst_rect, read_fb_handle, draw_fb_handle);
return true;
}
std::vector<std::string_view> TextureFilterer::GetFilterNames() {
std::vector<std::string_view> ret;
std::transform(filter_map.begin(), filter_map.end(), std::back_inserter(ret),
[](auto pair) { return pair.first; });
std::sort(ret.begin(), ret.end(), [](std::string_view lhs, std::string_view rhs) {
// sort lexicographically with none at the top
bool lhs_is_none{lhs == NONE};
bool rhs_is_none{rhs == NONE};
if (lhs_is_none || rhs_is_none)
return lhs_is_none && !rhs_is_none;
return lhs < rhs;
});
return ret;
}
} // namespace OpenGL

39
src/video_core/renderer_opengl/texture_filters/texture_filterer.h Normal file
View file

@ -0,0 +1,39 @@
// Copyright 2020 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#pragma once
#include <memory>
#include <string_view>
#include <vector>
#include <glad/glad.h>
#include "common/common_types.h"
#include "common/math_util.h"
#include "video_core/renderer_opengl/gl_surface_params.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
namespace OpenGL {
class TextureFilterer {
public:
static constexpr std::string_view NONE = "none";
explicit TextureFilterer(std::string_view filter_name, u16 scale_factor);
// returns true if the filter actually changed
bool Reset(std::string_view new_filter_name, u16 new_scale_factor);
// returns true if there is no active filter
bool IsNull() const;
// returns true if the texture was able to be filtered
bool Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, SurfaceParams::SurfaceType type,
GLuint read_fb_handle, GLuint draw_fb_handle);
static std::vector<std::string_view> GetFilterNames();
private:
std::string_view filter_name = NONE;
std::unique_ptr<TextureFilterBase> filter;
};
} // namespace OpenGL

30
src/video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.cpp
View file

@ -48,12 +48,11 @@
namespace OpenGL {
XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterInterface(scale_factor) {
XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterBase(scale_factor) {
const OpenGLState cur_state = OpenGLState::GetCurState();
program.Create(xbrz_freescale_vert.data(), xbrz_freescale_frag.data());
vao.Create();
draw_fbo.Create();
src_sampler.Create();
state.draw.shader_program = program.handle;
@ -68,31 +67,24 @@ XbrzFreescale::XbrzFreescale(u16 scale_factor) : TextureFilterInterface(scale_fa
cur_state.Apply();
state.draw.vertex_array = vao.handle;
state.draw.shader_program = program.handle;
state.draw.draw_framebuffer = draw_fbo.handle;
state.texture_units[0].sampler = src_sampler.handle;
}
void XbrzFreescale::scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) {
void XbrzFreescale::Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) {
const OpenGLState cur_state = OpenGLState::GetCurState();
OGLTexture src_tex;
src_tex.Create();
state.texture_units[0].texture_2d = src_tex.handle;
state.viewport = RectToViewport(rect);
state.texture_units[0].texture_2d = src_tex;
state.draw.draw_framebuffer = draw_fb_handle;
state.viewport = {static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.bottom),
static_cast<GLsizei>(dst_rect.GetWidth()),
static_cast<GLsizei>(dst_rect.GetHeight())};
state.Apply();
const FormatTuple tuple = GetFormatTuple(surface.pixel_format);
glPixelStorei(GL_UNPACK_ROW_LENGTH, static_cast<GLint>(surface.stride));
glActiveTexture(GL_TEXTURE0);
glTexImage2D(GL_TEXTURE_2D, 0, tuple.internal_format, rect.GetWidth(), rect.GetHeight(), 0,
tuple.format, tuple.type, &surface.gl_buffer[buffer_offset]);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
cur_state.texture_units[0].texture_2d, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, dst_tex, 0);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, 0, 0);
cur_state.Apply();
}

19
src/video_core/renderer_opengl/texture_filters/xbrz/xbrz_freescale.h
View file

@ -6,28 +6,23 @@
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_state.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_interface.h"
#include "video_core/renderer_opengl/texture_filters/texture_filter_base.h"
namespace OpenGL {
class XbrzFreescale : public TextureFilterInterface {
class XbrzFreescale : public TextureFilterBase {
public:
static TextureFilterInfo GetInfo() {
TextureFilterInfo info;
info.name = "xBRZ freescale";
info.constructor = std::make_unique<XbrzFreescale, u16>;
return info;
}
static constexpr std::string_view NAME = "xBRZ freescale";
XbrzFreescale(u16 scale_factor);
void scale(CachedSurface& surface, const Common::Rectangle<u32>& rect,
std::size_t buffer_offset) override;
explicit XbrzFreescale(u16 scale_factor);
void Filter(GLuint src_tex, const Common::Rectangle<u32>& src_rect, GLuint dst_tex,
const Common::Rectangle<u32>& dst_rect, GLuint read_fb_handle,
GLuint draw_fb_handle) override;
private:
OpenGLState state{};
OGLProgram program{};
OGLVertexArray vao{};
OGLFramebuffer draw_fbo{};
OGLSampler src_sampler{};
};
} // namespace OpenGL

1
src/video_core/video_core.cpp
View file

@ -28,6 +28,7 @@ std::atomic<bool> g_use_disk_shader_cache;
std::atomic<bool> g_renderer_bg_color_update_requested;
std::atomic<bool> g_renderer_sampler_update_requested;
std::atomic<bool> g_renderer_shader_update_requested;
std::atomic<bool> g_texture_filter_update_requested;
// Screenshot
std::atomic<bool> g_renderer_screenshot_requested;
void* g_screenshot_bits;

1
src/video_core/video_core.h
View file

@ -36,6 +36,7 @@ extern std::atomic<bool> g_use_disk_shader_cache;
extern std::atomic<bool> g_renderer_bg_color_update_requested;
extern std::atomic<bool> g_renderer_sampler_update_requested;
extern std::atomic<bool> g_renderer_shader_update_requested;
extern std::atomic<bool> g_texture_filter_update_requested;
// Screenshot
extern std::atomic<bool> g_renderer_screenshot_requested;
extern void* g_screenshot_bits;