Xargon/Duckstation
1
0
Fork 0
mirror of https://github.com/RetroDECK/Duckstation.git synced 2024-11-22 13:55:38 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

GPU: Drop OpenGL ES renderer

Browse source 
We can put this through the normal GL renderer instead.
This commit is contained in:
Connor McLaughlin 2020-04-11 00:00:10 +10:00
parent fbfd838e22
commit 6775100170
9 changed files with 15 additions and 769 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
src/core/CMakeLists.txt
View file

@ -34,8 +34,6 @@ add_library(core
gpu_hw.h
gpu_hw_opengl.cpp
gpu_hw_opengl.h
gpu_hw_opengl_es.cpp
gpu_hw_opengl_es.h
gpu_hw_shadergen.cpp
gpu_hw_shadergen.h
gpu_sw.cpp

2
src/core/core.vcxproj
View file

@ -63,7 +63,6 @@
<ClCompile Include="game_list.cpp" />
<ClCompile Include="gpu_commands.cpp" />
<ClCompile Include="gpu_hw_d3d11.cpp" />
<ClCompile Include="gpu_hw_opengl_es.cpp" />
<ClCompile Include="gpu_hw_shadergen.cpp" />
<ClCompile Include="gpu_sw.cpp" />
<ClCompile Include="gte.cpp" />
@ -101,7 +100,6 @@
<ClInclude Include="digital_controller.h" />
<ClInclude Include="game_list.h" />
<ClInclude Include="gpu_hw_d3d11.h" />
<ClInclude Include="gpu_hw_opengl_es.h" />
<ClInclude Include="gpu_hw_shadergen.h" />
<ClInclude Include="gpu_sw.h" />
<ClInclude Include="gte.h" />

2
src/core/core.vcxproj.filters
View file

@ -34,7 +34,6 @@
<ClCompile Include="cpu_types.cpp" />
<ClCompile Include="game_list.cpp" />
<ClCompile Include="cpu_recompiler_code_generator_aarch64.cpp" />
<ClCompile Include="gpu_hw_opengl_es.cpp" />
<ClCompile Include="sio.cpp" />
<ClCompile Include="controller.cpp" />
<ClCompile Include="analog_controller.cpp" />
@ -77,7 +76,6 @@
<ClInclude Include="cpu_recompiler_thunks.h" />
<ClInclude Include="cpu_recompiler_code_generator.h" />
<ClInclude Include="game_list.h" />
<ClInclude Include="gpu_hw_opengl_es.h" />
<ClInclude Include="sio.h" />
<ClInclude Include="controller.h" />
<ClInclude Include="analog_controller.h" />

3
src/core/gpu.h
View file

@ -155,9 +155,6 @@ public:
// gpu_hw_opengl.cpp
static std::unique_ptr<GPU> CreateHardwareOpenGLRenderer();
// gpu_hw_opengl_es.cpp
static std::unique_ptr<GPU> CreateHardwareOpenGLESRenderer();
// gpu_sw.cpp
static std::unique_ptr<GPU> CreateSoftwareRenderer();

662
src/core/gpu_hw_opengl_es.cpp
View file

@ -1,662 +0,0 @@
#include "gpu_hw_opengl_es.h"
#include "common/assert.h"
#include "common/log.h"
#include "gpu_hw_shadergen.h"
#include "host_display.h"
#include "system.h"
Log_SetChannel(GPU_HW_OpenGL_ES);
GPU_HW_OpenGL_ES::GPU_HW_OpenGL_ES() : GPU_HW(), m_vertex_buffer(VERTEX_BUFFER_SIZE / sizeof(BatchVertex)) {}
GPU_HW_OpenGL_ES::~GPU_HW_OpenGL_ES()
{
// TODO: Destroy objects...
if (m_host_display)
{
m_host_display->ClearDisplayTexture();
ResetGraphicsAPIState();
}
}
bool GPU_HW_OpenGL_ES::Initialize(HostDisplay* host_display, System* system, DMA* dma,
InterruptController* interrupt_controller, Timers* timers)
{
if (host_display->GetRenderAPI() != HostDisplay::RenderAPI::OpenGLES)
{
Log_ErrorPrintf("Host render API type is incompatible");
return false;
}
SetCapabilities(host_display);
m_shader_cache.Open(true, system->GetHostInterface()->GetUserDirectoryRelativePath("cache"));
if (!GPU_HW::Initialize(host_display, system, dma, interrupt_controller, timers))
return false;
if (!CreateFramebuffer())
{
Log_ErrorPrintf("Failed to create framebuffer");
return false;
}
if (!CompilePrograms())
{
Log_ErrorPrintf("Failed to compile programs");
return false;
}
RestoreGraphicsAPIState();
return true;
}
void GPU_HW_OpenGL_ES::Reset()
{
GPU_HW::Reset();
ClearFramebuffer();
}
void GPU_HW_OpenGL_ES::ResetGraphicsAPIState()
{
GPU_HW::ResetGraphicsAPIState();
glEnable(GL_CULL_FACE);
glDisable(GL_SCISSOR_TEST);
glDisable(GL_BLEND);
glDepthMask(GL_TRUE);
glLineWidth(1.0f);
glDisableVertexAttribArray(0);
glDisableVertexAttribArray(1);
glDisableVertexAttribArray(2);
glDisableVertexAttribArray(3);
}
void GPU_HW_OpenGL_ES::RestoreGraphicsAPIState()
{
m_vram_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
glViewport(0, 0, m_vram_texture.GetWidth(), m_vram_texture.GetHeight());
glDisable(GL_CULL_FACE);
glDisable(GL_DEPTH_TEST);
glEnable(GL_SCISSOR_TEST);
glDepthMask(GL_FALSE);
glLineWidth(static_cast<float>(m_resolution_scale));
glBindVertexArray(0);
SetScissorFromDrawingArea();
SetVertexPointers();
}
void GPU_HW_OpenGL_ES::UpdateSettings()
{
GPU_HW::UpdateSettings();
CreateFramebuffer();
CompilePrograms();
UpdateDisplay();
}
void GPU_HW_OpenGL_ES::MapBatchVertexPointer(u32 required_vertices)
{
Assert(!m_batch_start_vertex_ptr);
m_batch_start_vertex_ptr = m_vertex_buffer.data();
m_batch_current_vertex_ptr = m_batch_start_vertex_ptr;
m_batch_end_vertex_ptr = m_vertex_buffer.data() + m_vertex_buffer.size();
m_batch_base_vertex = 0;
}
std::tuple<s32, s32> GPU_HW_OpenGL_ES::ConvertToFramebufferCoordinates(s32 x, s32 y)
{
return std::make_tuple(x, static_cast<s32>(static_cast<s32>(VRAM_HEIGHT) - y));
}
void GPU_HW_OpenGL_ES::SetCapabilities(HostDisplay* host_display)
{
Log_InfoPrintf("GL_VERSION: %s", glGetString(GL_VERSION));
Log_InfoPrintf("GL_RENDERER: %s", glGetString(GL_VERSION));
GLint max_texture_size = VRAM_WIDTH;
glGetIntegerv(GL_MAX_TEXTURE_SIZE, &max_texture_size);
Log_InfoPrintf("Max texture size: %dx%d", max_texture_size, max_texture_size);
const int max_texture_scale = max_texture_size / VRAM_WIDTH;
std::array<int, 2> line_width_range = {{1, 1}};
glGetIntegerv(GL_ALIASED_LINE_WIDTH_RANGE, line_width_range.data());
Log_InfoPrintf("Max line width: %d", line_width_range[1]);
m_max_resolution_scale = std::min(max_texture_scale, line_width_range[1]);
m_supports_dual_source_blend = false;
}
bool GPU_HW_OpenGL_ES::CreateFramebuffer()
{
// save old vram texture/fbo, in case we're changing scale
GL::Texture old_vram_texture = std::move(m_vram_texture);
// scale vram size to internal resolution
const u32 texture_width = VRAM_WIDTH * m_resolution_scale;
const u32 texture_height = VRAM_HEIGHT * m_resolution_scale;
if (!m_vram_texture.Create(texture_width, texture_height, GL_RGBA, GL_UNSIGNED_BYTE, nullptr, false) ||
!m_vram_texture.CreateFramebuffer())
{
return false;
}
// do we need to restore the framebuffer after a size change?
if (old_vram_texture.IsValid())
{
const bool linear_filter = old_vram_texture.GetWidth() > m_vram_texture.GetWidth();
Log_DevPrintf("Scaling %ux%u VRAM texture to %ux%u using %s filter", old_vram_texture.GetWidth(),
old_vram_texture.GetHeight(), m_vram_texture.GetWidth(), m_vram_texture.GetHeight(),
linear_filter ? "linear" : "nearest");
glDisable(GL_SCISSOR_TEST);
old_vram_texture.BindFramebuffer(GL_READ_FRAMEBUFFER);
glBlitFramebuffer(0, 0, old_vram_texture.GetWidth(), old_vram_texture.GetHeight(), 0, 0, m_vram_texture.GetWidth(),
m_vram_texture.GetHeight(), GL_COLOR_BUFFER_BIT, linear_filter ? GL_LINEAR : GL_NEAREST);
glEnable(GL_SCISSOR_TEST);
old_vram_texture.Destroy();
}
if (!m_vram_read_texture.Create(texture_width, texture_height, GL_RGBA, GL_UNSIGNED_BYTE, nullptr, false) ||
!m_vram_read_texture.CreateFramebuffer() ||
!m_vram_encoding_texture.Create(VRAM_WIDTH, VRAM_HEIGHT, GL_RGBA, GL_UNSIGNED_BYTE, nullptr, false) ||
!m_vram_encoding_texture.CreateFramebuffer() ||
!m_display_texture.Create(texture_width, texture_height, GL_RGBA, GL_UNSIGNED_BYTE, nullptr, false) ||
!m_display_texture.CreateFramebuffer())
{
return false;
}
m_vram_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
SetFullVRAMDirtyRectangle();
return true;
}
void GPU_HW_OpenGL_ES::ClearFramebuffer()
{
glDisable(GL_SCISSOR_TEST);
glClearColor(0.0f, 0.0f, 0.0f, 0.0f);
glClear(GL_COLOR_BUFFER_BIT);
glEnable(GL_SCISSOR_TEST);
SetFullVRAMDirtyRectangle();
}
bool GPU_HW_OpenGL_ES::CompilePrograms()
{
GPU_HW_ShaderGen shadergen(m_host_display->GetRenderAPI(), m_resolution_scale, m_true_color, m_scaled_dithering,
m_texture_filtering, m_supports_dual_source_blend);
for (u32 render_mode = 0; render_mode < 4; render_mode++)
{
for (u32 texture_mode = 0; texture_mode < 9; texture_mode++)
{
for (u8 dithering = 0; dithering < 2; dithering++)
{
for (u8 interlacing = 0; interlacing < 2; interlacing++)
{
const bool textured = (static_cast<TextureMode>(texture_mode) != TextureMode::Disabled);
const std::string vs = shadergen.GenerateBatchVertexShader(textured);
const std::string fs = shadergen.GenerateBatchFragmentShader(
static_cast<BatchRenderMode>(render_mode), static_cast<TextureMode>(texture_mode),
ConvertToBoolUnchecked(dithering), ConvertToBoolUnchecked(interlacing));
std::optional<GL::Program> prog = m_shader_cache.GetProgram(vs, fs, [this, textured](GL::Program& prog) {
prog.BindAttribute(0, "a_pos");
prog.BindAttribute(1, "a_col0");
if (textured)
{
prog.BindAttribute(2, "a_texcoord");
prog.BindAttribute(3, "a_texpage");
}
});
if (!prog)
return false;
prog->Bind();
prog->RegisterUniform("u_texture_window_mask");
prog->RegisterUniform("u_texture_window_offset");
prog->RegisterUniform("u_src_alpha_factor");
prog->RegisterUniform("u_dst_alpha_factor");
prog->RegisterUniform("u_set_mask_while_drawing");
prog->RegisterUniform("u_interlaced_displayed_field");
if (textured)
prog->Uniform1i("samp0", 0);
m_render_programs[render_mode][texture_mode][dithering][interlacing] = std::move(*prog);
}
}
}
}
for (u8 depth_24bit = 0; depth_24bit < 2; depth_24bit++)
{
for (u8 interlaced = 0; interlaced < 2; interlaced++)
{
const std::string vs = shadergen.GenerateScreenQuadVertexShader();
const std::string fs = shadergen.GenerateDisplayFragmentShader(ConvertToBoolUnchecked(depth_24bit),
ConvertToBoolUnchecked(interlaced));
std::optional<GL::Program> prog = m_shader_cache.GetProgram(vs, fs);
if (!prog)
return false;
prog->Bind();
prog->RegisterUniform("u_base_coords");
prog->Uniform1i("samp0", 0);
m_display_programs[depth_24bit][interlaced] = std::move(*prog);
}
}
std::optional<GL::Program> prog = m_shader_cache.GetProgram(shadergen.GenerateScreenQuadVertexShader(),
shadergen.GenerateInterlacedFillFragmentShader());
if (!prog)
return false;
prog->Bind();
prog->RegisterUniform("u_fill_color");
prog->RegisterUniform("u_u_interlaced_displayed_field");
m_vram_interlaced_fill_program = std::move(*prog);
prog =
m_shader_cache.GetProgram(shadergen.GenerateScreenQuadVertexShader(), shadergen.GenerateVRAMReadFragmentShader());
if (!prog)
return false;
prog->Bind();
prog->RegisterUniform("u_base_coords");
prog->RegisterUniform("u_size");
prog->Uniform1i("samp0", 0);
m_vram_read_program = std::move(*prog);
return true;
}
void GPU_HW_OpenGL_ES::SetVertexPointers()
{
glEnableVertexAttribArray(0);
glEnableVertexAttribArray(1);
glEnableVertexAttribArray(2);
glEnableVertexAttribArray(3);
glVertexAttribIPointer(0, 2, GL_INT, sizeof(BatchVertex), &m_vertex_buffer[0].x);
glVertexAttribPointer(1, 4, GL_UNSIGNED_BYTE, true, sizeof(BatchVertex), &m_vertex_buffer[0].color);
glVertexAttribIPointer(2, 1, GL_INT, sizeof(BatchVertex), &m_vertex_buffer[0].u);
glVertexAttribIPointer(3, 1, GL_INT, sizeof(BatchVertex), &m_vertex_buffer[0].texpage);
}
void GPU_HW_OpenGL_ES::SetDrawState(BatchRenderMode render_mode)
{
const GL::Program& prog = m_render_programs[static_cast<u8>(render_mode)][static_cast<u8>(m_batch.texture_mode)]
[BoolToUInt8(m_batch.dithering)][BoolToUInt8(m_batch.interlacing)];
m_batch_ubo_dirty |= !prog.IsBound();
prog.Bind();
if (m_batch.texture_mode != TextureMode::Disabled)
m_vram_read_texture.Bind();
if (m_batch.transparency_mode == TransparencyMode::Disabled || render_mode == BatchRenderMode::OnlyOpaque)
{
glDisable(GL_BLEND);
}
else
{
glEnable(GL_BLEND);
glBlendEquationSeparate(
m_batch.transparency_mode == TransparencyMode::BackgroundMinusForeground ? GL_FUNC_REVERSE_SUBTRACT : GL_FUNC_ADD,
GL_FUNC_ADD);
glBlendFuncSeparate(GL_ONE, GL_SRC_ALPHA, GL_ONE, GL_ZERO);
}
if (m_drawing_area_changed)
{
m_drawing_area_changed = false;
m_vram_dirty_rect.Include(m_drawing_area);
SetScissorFromDrawingArea();
}
if (m_batch_ubo_dirty)
{
prog.Uniform2uiv(0, m_batch_ubo_data.u_texture_window_mask);
prog.Uniform2uiv(1, m_batch_ubo_data.u_texture_window_offset);
prog.Uniform1f(2, m_batch_ubo_data.u_src_alpha_factor);
prog.Uniform1f(3, m_batch_ubo_data.u_dst_alpha_factor);
prog.Uniform1i(4, static_cast<s32>(m_batch_ubo_data.u_set_mask_while_drawing));
prog.Uniform1i(5, static_cast<s32>(m_batch_ubo_data.u_interlaced_displayed_field));
m_batch_ubo_dirty = false;
}
}
void GPU_HW_OpenGL_ES::SetScissorFromDrawingArea()
{
int left, top, right, bottom;
CalcScissorRect(&left, &top, &right, &bottom);
const int width = right - left;
const int height = bottom - top;
const int x = left;
const int y = m_vram_texture.GetHeight() - bottom;
Log_DebugPrintf("SetScissor: (%d-%d, %d-%d)", x, x + width, y, y + height);
glScissor(x, y, width, height);
}
void GPU_HW_OpenGL_ES::UpdateDisplay()
{
GPU_HW::UpdateDisplay();
if (m_system->GetSettings().debugging.show_vram)
{
m_host_display->SetDisplayTexture(reinterpret_cast<void*>(static_cast<uintptr_t>(m_vram_texture.GetGLId())),
m_vram_texture.GetWidth(), static_cast<s32>(m_vram_texture.GetHeight()), 0,
m_vram_texture.GetHeight(), m_vram_texture.GetWidth(),
-static_cast<s32>(m_vram_texture.GetHeight()));
m_host_display->SetDisplayParameters(VRAM_WIDTH, VRAM_HEIGHT, 0, 0, VRAM_WIDTH, VRAM_HEIGHT,
static_cast<float>(VRAM_WIDTH) / static_cast<float>(VRAM_HEIGHT));
}
else
{
const u32 vram_offset_x = m_crtc_state.display_vram_left;
const u32 vram_offset_y = m_crtc_state.display_vram_top;
const u32 scaled_vram_offset_x = vram_offset_x * m_resolution_scale;
const u32 scaled_vram_offset_y = vram_offset_y * m_resolution_scale;
const u32 display_width = m_crtc_state.display_vram_width;
const u32 display_height = m_crtc_state.display_vram_height;
const u32 scaled_display_width = display_width * m_resolution_scale;
const u32 scaled_display_height = display_height * m_resolution_scale;
const bool interlaced = IsInterlacedDisplayEnabled();
if (m_GPUSTAT.display_disable)
{
m_host_display->ClearDisplayTexture();
}
else if (!m_GPUSTAT.display_area_color_depth_24 && !interlaced)
{
m_host_display->SetDisplayTexture(reinterpret_cast<void*>(static_cast<uintptr_t>(m_vram_texture.GetGLId())),
m_vram_texture.GetWidth(), m_vram_texture.GetHeight(), scaled_vram_offset_x,
m_vram_texture.GetHeight() - scaled_vram_offset_y, scaled_display_width,
-static_cast<s32>(scaled_display_height));
}
else
{
glDisable(GL_BLEND);
glDisable(GL_SCISSOR_TEST);
GL::Program& prog =
m_display_programs[BoolToUInt8(m_GPUSTAT.display_area_color_depth_24)][BoolToUInt8(interlaced)];
prog.Bind();
m_display_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
m_vram_texture.Bind();
const u32 flipped_vram_offset_y = VRAM_HEIGHT - vram_offset_y - display_height;
const u32 scaled_flipped_vram_offset_y =
m_vram_texture.GetHeight() - scaled_vram_offset_y - scaled_display_height;
const u32 reinterpret_field_offset = GetInterlacedField();
const u32 reinterpret_start_x = m_crtc_state.regs.X * m_resolution_scale;
const u32 reinterpret_width = scaled_display_width + (m_crtc_state.display_vram_left - m_crtc_state.regs.X);
prog.Uniform2i(0, reinterpret_field_offset, reinterpret_start_x);
m_batch_ubo_dirty = true;
glViewport(reinterpret_start_x, scaled_flipped_vram_offset_y, reinterpret_width, scaled_display_height);
glDrawArrays(GL_TRIANGLES, 0, 3);
m_host_display->SetDisplayTexture(reinterpret_cast<void*>(static_cast<uintptr_t>(m_display_texture.GetGLId())),
m_display_texture.GetWidth(), m_display_texture.GetHeight(),
scaled_vram_offset_x, m_vram_texture.GetHeight() - scaled_vram_offset_y,
scaled_display_width, -static_cast<s32>(scaled_display_height));
}
// restore state
m_vram_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
glViewport(0, 0, m_vram_texture.GetWidth(), m_vram_texture.GetHeight());
glEnable(GL_SCISSOR_TEST);
}
m_host_display->SetDisplayParameters(m_crtc_state.display_width, m_crtc_state.display_height,
m_crtc_state.display_origin_left, m_crtc_state.display_origin_top,
m_crtc_state.display_vram_width, m_crtc_state.display_vram_height,
m_crtc_state.display_aspect_ratio);
}
void GPU_HW_OpenGL_ES::ReadVRAM(u32 x, u32 y, u32 width, u32 height)
{
// Get bounds with wrap-around handled.
const Common::Rectangle<u32> copy_rect = GetVRAMTransferBounds(x, y, width, height);
const u32 encoded_width = (copy_rect.GetWidth() + 1) / 2;
const u32 encoded_height = copy_rect.GetHeight();
// Encode the 24-bit texture as 16-bit.
m_vram_encoding_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
m_vram_texture.Bind();
m_vram_read_program.Bind();
m_vram_read_program.Uniform2i(0, copy_rect.left, VRAM_HEIGHT - copy_rect.top - copy_rect.GetHeight());
m_vram_read_program.Uniform2i(1, copy_rect.GetWidth(), copy_rect.GetHeight());
glDisable(GL_BLEND);
glDisable(GL_SCISSOR_TEST);
glViewport(0, 0, encoded_width, encoded_height);
glDrawArrays(GL_TRIANGLES, 0, 3);
// Readback encoded texture.
m_vram_encoding_texture.BindFramebuffer(GL_READ_FRAMEBUFFER);
glPixelStorei(GL_PACK_ALIGNMENT, 2);
glPixelStorei(GL_PACK_ROW_LENGTH, VRAM_WIDTH / 2);
glReadPixels(0, 0, encoded_width, encoded_height, GL_RGBA, GL_UNSIGNED_BYTE,
&m_vram_shadow[copy_rect.top * VRAM_WIDTH + copy_rect.left]);
glPixelStorei(GL_PACK_ALIGNMENT, 4);
glPixelStorei(GL_PACK_ROW_LENGTH, 0);
RestoreGraphicsAPIState();
}
void GPU_HW_OpenGL_ES::FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color)
{
if ((x + width) > VRAM_WIDTH || (y + height) > VRAM_HEIGHT)
{
// CPU round trip if oversized for now.
Log_WarningPrintf("Oversized VRAM fill (%u-%u, %u-%u), CPU round trip", x, x + width, y, y + height);
ReadVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT);
GPU::FillVRAM(x, y, width, height, color);
UpdateVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT, m_vram_shadow.data());
return;
}
GPU_HW::FillVRAM(x, y, width, height, color);
// scale coordinates
x *= m_resolution_scale;
y *= m_resolution_scale;
width *= m_resolution_scale;
height *= m_resolution_scale;
glScissor(x, m_vram_texture.GetHeight() - y - height, width, height);
// drop precision unless true colour is enabled
if (!m_true_color)
color = RGBA5551ToRGBA8888(RGBA8888ToRGBA5551(color));
const auto [r, g, b, a] = RGBA8ToFloat(color);
if (!IsInterlacedRenderingEnabled())
{
glClearColor(r, g, b, a);
glClear(GL_COLOR_BUFFER_BIT);
SetScissorFromDrawingArea();
}
else
{
m_vram_interlaced_fill_program.Bind();
m_vram_interlaced_fill_program.Uniform4f(0, r, g, b, a);
m_vram_interlaced_fill_program.Uniform1i(1, GetInterlacedField());
glDrawArrays(GL_TRIANGLES, 0, 3);
RestoreGraphicsAPIState();
}
}
void GPU_HW_OpenGL_ES::UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data)
{
if ((x + width) > VRAM_WIDTH || (y + height) > VRAM_HEIGHT)
{
// CPU round trip if oversized for now.
Log_WarningPrintf("Oversized VRAM update (%u-%u, %u-%u), CPU round trip", x, x + width, y, y + height);
ReadVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT);
GPU::UpdateVRAM(x, y, width, height, data);
UpdateVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT, m_vram_shadow.data());
return;
}
GPU_HW::UpdateVRAM(x, y, width, height, data);
const u32 num_pixels = width * height;
std::vector<u32> staging_buffer(num_pixels);
// reverse copy the rows so it matches opengl's lower-left origin
const u32 source_stride = width * sizeof(u16);
const u8* source_ptr = static_cast<const u8*>(data) + (source_stride * (height - 1));
u32* dest_ptr = static_cast<u32*>(staging_buffer.data());
for (u32 row = 0; row < height; row++)
{
const u8* source_row_ptr = source_ptr;
for (u32 col = 0; col < width; col++)
{
u16 src_col;
std::memcpy(&src_col, source_row_ptr, sizeof(src_col));
source_row_ptr += sizeof(src_col);
*(dest_ptr++) = RGBA5551ToRGBA8888(src_col);
}
source_ptr -= source_stride;
}
// have to write to the 1x texture first
if (m_resolution_scale > 1)
m_vram_encoding_texture.Bind();
else
m_vram_texture.Bind();
// lower-left origin flip happens here
const u32 flipped_y = VRAM_HEIGHT - y - height;
// update texture data
glTexSubImage2D(GL_TEXTURE_2D, 0, x, flipped_y, width, height, GL_RGBA, GL_UNSIGNED_BYTE, staging_buffer.data());
if (m_resolution_scale > 1)
{
// scale to internal resolution
const u32 scaled_width = width * m_resolution_scale;
const u32 scaled_height = height * m_resolution_scale;
const u32 scaled_x = x * m_resolution_scale;
const u32 scaled_y = y * m_resolution_scale;
const u32 scaled_flipped_y = m_vram_texture.GetHeight() - scaled_y - scaled_height;
glDisable(GL_SCISSOR_TEST);
m_vram_encoding_texture.BindFramebuffer(GL_READ_FRAMEBUFFER);
glBlitFramebuffer(x, flipped_y, x + width, flipped_y + height, scaled_x, scaled_flipped_y, scaled_x + scaled_width,
scaled_flipped_y + scaled_height, GL_COLOR_BUFFER_BIT, GL_NEAREST);
glEnable(GL_SCISSOR_TEST);
}
}
void GPU_HW_OpenGL_ES::CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height)
{
if ((src_x + width) > VRAM_WIDTH || (src_y + height) > VRAM_HEIGHT || (dst_x + width) > VRAM_WIDTH ||
(dst_y + height) > VRAM_HEIGHT)
{
Log_WarningPrintf("Oversized VRAM copy (%u,%u, %u,%u, %u,%u), CPU round trip", src_x, src_y, dst_x, dst_y, width,
height);
ReadVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT);
GPU::CopyVRAM(src_x, src_y, dst_x, dst_y, width, height);
UpdateVRAM(0, 0, VRAM_WIDTH, VRAM_HEIGHT, m_vram_shadow.data());
return;
}
GPU_HW::CopyVRAM(src_x, src_y, dst_x, dst_y, width, height);
src_x *= m_resolution_scale;
src_y *= m_resolution_scale;
dst_x *= m_resolution_scale;
dst_y *= m_resolution_scale;
width *= m_resolution_scale;
height *= m_resolution_scale;
// lower-left origin flip
src_y = m_vram_texture.GetHeight() - src_y - height;
dst_y = m_vram_texture.GetHeight() - dst_y - height;
if (GLAD_GL_EXT_copy_image)
{
glCopyImageSubDataEXT(m_vram_texture.GetGLId(), GL_TEXTURE_2D, 0, src_x, src_y, 0, m_vram_texture.GetGLId(),
GL_TEXTURE_2D, 0, dst_x, dst_y, 0, width, height, 1);
}
else
{
glDisable(GL_SCISSOR_TEST);
m_vram_texture.BindFramebuffer(GL_READ_FRAMEBUFFER);
glBlitFramebuffer(src_x, src_y, src_x + width, src_y + height, dst_x, dst_y, dst_x + width, dst_y + height,
GL_COLOR_BUFFER_BIT, GL_NEAREST);
glEnable(GL_SCISSOR_TEST);
}
}
void GPU_HW_OpenGL_ES::UpdateVRAMReadTexture()
{
const auto scaled_rect = m_vram_dirty_rect * m_resolution_scale;
const u32 width = scaled_rect.GetWidth();
const u32 height = scaled_rect.GetHeight();
const u32 x = scaled_rect.left;
const u32 y = m_vram_texture.GetHeight() - scaled_rect.top - height;
if (GLAD_GL_EXT_copy_image)
{
glCopyImageSubDataEXT(m_vram_texture.GetGLId(), GL_TEXTURE_2D, 0, x, y, 0, m_vram_read_texture.GetGLId(),
GL_TEXTURE_2D, 0, x, y, 0, width, height, 1);
}
else
{
m_vram_read_texture.BindFramebuffer(GL_DRAW_FRAMEBUFFER);
m_vram_texture.BindFramebuffer(GL_READ_FRAMEBUFFER);
glDisable(GL_SCISSOR_TEST);
glBlitFramebuffer(x, y, x + width, y + height, x, y, x + width, y + height, GL_COLOR_BUFFER_BIT, GL_NEAREST);
glEnable(GL_SCISSOR_TEST);
m_vram_texture.BindFramebuffer(GL_FRAMEBUFFER);
}
}
void GPU_HW_OpenGL_ES::FlushRender()
{
static constexpr std::array<GLenum, 4> gl_primitives = {{GL_LINES, GL_LINE_STRIP, GL_TRIANGLES, GL_TRIANGLE_STRIP}};
if (!m_batch_current_vertex_ptr)
return;
const u32 vertex_count = GetBatchVertexCount();
m_batch_start_vertex_ptr = nullptr;
m_batch_end_vertex_ptr = nullptr;
m_batch_current_vertex_ptr = nullptr;
if (vertex_count == 0)
return;
m_renderer_stats.num_batches++;
if (m_batch.NeedsTwoPassRendering())
{
SetDrawState(BatchRenderMode::OnlyTransparent);
glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
SetDrawState(BatchRenderMode::OnlyOpaque);
glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
}
else
{
SetDrawState(m_batch.GetRenderMode());
glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
}
}
std::unique_ptr<GPU> GPU::CreateHardwareOpenGLESRenderer()
{
return std::make_unique<GPU_HW_OpenGL_ES>();
}

72
src/core/gpu_hw_opengl_es.h
View file

@ -1,72 +0,0 @@
#pragma once
#include "common/gl/program.h"
#include "common/gl/shader_cache.h"
#include "common/gl/stream_buffer.h"
#include "common/gl/texture.h"
#include "glad.h"
#include "gpu_hw.h"
#include <array>
#include <tuple>
class GPU_HW_OpenGL_ES : public GPU_HW
{
public:
GPU_HW_OpenGL_ES();
~GPU_HW_OpenGL_ES() override;
bool Initialize(HostDisplay* host_display, System* system, DMA* dma, InterruptController* interrupt_controller,
Timers* timers) override;
void Reset() override;
void ResetGraphicsAPIState() override;
void RestoreGraphicsAPIState() override;
void UpdateSettings() override;
protected:
void UpdateDisplay() override;
void ReadVRAM(u32 x, u32 y, u32 width, u32 height) override;
void FillVRAM(u32 x, u32 y, u32 width, u32 height, u32 color) override;
void UpdateVRAM(u32 x, u32 y, u32 width, u32 height, const void* data) override;
void CopyVRAM(u32 src_x, u32 src_y, u32 dst_x, u32 dst_y, u32 width, u32 height) override;
void FlushRender() override;
void MapBatchVertexPointer(u32 required_vertices) override;
void UpdateVRAMReadTexture() override;
private:
struct GLStats
{
u32 num_batches;
u32 num_vertices;
u32 num_vram_reads;
u32 num_vram_writes;
u32 num_vram_read_texture_updates;
u32 num_uniform_buffer_updates;
};
std::tuple<s32, s32> ConvertToFramebufferCoordinates(s32 x, s32 y);
void SetCapabilities(HostDisplay* host_display);
bool CreateFramebuffer();
void ClearFramebuffer();
bool CompilePrograms();
void SetVertexPointers();
void SetDrawState(BatchRenderMode render_mode);
void SetScissorFromDrawingArea();
GL::ShaderCache m_shader_cache;
// downsample texture - used for readbacks at >1xIR.
GL::Texture m_vram_texture;
GL::Texture m_vram_read_texture;
GL::Texture m_vram_encoding_texture;
GL::Texture m_display_texture;
std::vector<BatchVertex> m_vertex_buffer;
std::array<std::array<std::array<std::array<GL::Program, 2>, 2>, 9>, 4>
m_render_programs; // [render_mode][texture_mode][dithering][interlacing]
std::array<std::array<GL::Program, 2>, 2> m_display_programs; // [depth_24][interlaced]
GL::Program m_vram_interlaced_fill_program;
GL::Program m_vram_read_program;
};

36
src/core/gpu_hw_shadergen.cpp
View file

@ -9,8 +9,7 @@ GPU_HW_ShaderGen::GPU_HW_ShaderGen(HostDisplay::RenderAPI render_api, u32 resolu
bool scaled_dithering, bool texture_filtering, bool supports_dual_source_blend)
: m_render_api(render_api), m_resolution_scale(resolution_scale), m_true_color(true_color),
m_scaled_dithering(scaled_dithering), m_texture_filering(texture_filtering),
m_glsl(render_api != HostDisplay::RenderAPI::D3D11), m_glsl_es(render_api == HostDisplay::RenderAPI::OpenGLES),
m_supports_dual_source_blend(supports_dual_source_blend)
m_glsl(render_api != HostDisplay::RenderAPI::D3D11), m_supports_dual_source_blend(supports_dual_source_blend)
{
if (m_glsl)
SetGLSLVersionString();
@ -26,6 +25,7 @@ static void DefineMacro(std::stringstream& ss, const char* name, bool enabled)
void GPU_HW_ShaderGen::SetGLSLVersionString()
{
const char* glsl_version = reinterpret_cast<const char*>(glGetString(GL_SHADING_LANGUAGE_VERSION));
const bool glsl_es = (m_render_api == HostDisplay::RenderAPI::OpenGLES);
Assert(glsl_version != nullptr);
// Skip any strings in front of the version code.
@ -37,12 +37,12 @@ void GPU_HW_ShaderGen::SetGLSLVersionString()
if (std::sscanf(glsl_version_start, "%d.%d", &major_version, &minor_version) == 2)
{
// Cap at GLSL 3.3, we're not using anything newer for now.
if (!m_glsl_es && major_version >= 4)
if (!glsl_es && major_version >= 4)
{
major_version = 3;
minor_version = 30;
}
else if (m_glsl_es && (major_version > 3 || minor_version > 20))
else if (glsl_es && (major_version > 3 || minor_version > 20))
{
major_version = 3;
minor_version = 20;
@ -51,17 +51,17 @@ void GPU_HW_ShaderGen::SetGLSLVersionString()
else
{
Log_ErrorPrintf("Invalid GLSL version string: '%s' ('%s')", glsl_version, glsl_version_start);
if (m_glsl_es)
if (glsl_es)
{
major_version = 3;
minor_version = 0;
}
m_glsl_version_string = m_glsl_es ? "300" : "130";
m_glsl_version_string = glsl_es ? "300" : "130";
}
char buf[128];
std::snprintf(buf, sizeof(buf), "#version %d%02d%s", major_version, minor_version,
(m_glsl_es && major_version >= 3) ? " es" : "");
(glsl_es && major_version >= 3) ? " es" : "");
m_glsl_version_string = buf;
}
@ -178,23 +178,15 @@ float4 RGBA5551ToRGBA8(uint v)
void GPU_HW_ShaderGen::DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members)
{
if (m_glsl_es)
{
for (const char* member : members)
ss << "uniform " << member << ";\n";
}
if (m_glsl)
ss << "layout(std140) uniform UBOBlock\n";
else
{
if (m_glsl)
ss << "layout(std140) uniform UBOBlock\n";
else
ss << "cbuffer UBOBlock : register(b0)\n";
ss << "cbuffer UBOBlock : register(b0)\n";
ss << "{\n";
for (const char* member : members)
ss << member << ";\n";
ss << "};\n\n";
}
ss << "{\n";
for (const char* member : members)
ss << member << ";\n";
ss << "};\n\n";
}
void GPU_HW_ShaderGen::DeclareTexture(std::stringstream& ss, const char* name, u32 index)

1
src/core/gpu_hw_shadergen.h
View file

@ -29,7 +29,6 @@ public:
bool m_scaled_dithering;
bool m_texture_filering;
bool m_glsl;
bool m_glsl_es;
bool m_supports_dual_source_blend;
std::string m_glsl_version_string;

4
src/core/system.cpp
View file

@ -264,9 +264,7 @@ bool System::CreateGPU(GPURenderer renderer)
switch (renderer)
{
case GPURenderer::HardwareOpenGL:
m_gpu = m_host_interface->GetDisplay()->GetRenderAPI() == HostDisplay::RenderAPI::OpenGLES ?
GPU::CreateHardwareOpenGLESRenderer() :
GPU::CreateHardwareOpenGLRenderer();
m_gpu = GPU::CreateHardwareOpenGLRenderer();
break;
#ifdef WIN32