summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r--src/video_core/renderer_opengl/blit_image.cpp3
-rw-r--r--src/video_core/renderer_opengl/gl_fence_manager.cpp4
-rw-r--r--src/video_core/renderer_opengl/gl_graphics_pipeline.cpp5
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp48
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h20
-rw-r--r--src/video_core/renderer_opengl/gl_resource_manager.cpp10
-rw-r--r--src/video_core/renderer_opengl/gl_resource_manager.h3
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.cpp28
-rw-r--r--src/video_core/renderer_opengl/gl_texture_cache.h6
9 files changed, 96 insertions, 31 deletions
diff --git a/src/video_core/renderer_opengl/blit_image.cpp b/src/video_core/renderer_opengl/blit_image.cpp
index 9a560a73b..3b03e8d5a 100644
--- a/src/video_core/renderer_opengl/blit_image.cpp
+++ b/src/video_core/renderer_opengl/blit_image.cpp
@@ -22,7 +22,7 @@ BlitImageHelper::~BlitImageHelper() = default;
void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, GLuint src_sampler,
const Region2D& dst_region, const Region2D& src_region,
const Extent3D& src_size) {
- glEnable(GL_CULL_FACE);
+ glDisable(GL_CULL_FACE);
glDisable(GL_COLOR_LOGIC_OP);
glDisable(GL_DEPTH_TEST);
glDisable(GL_STENCIL_TEST);
@@ -31,7 +31,6 @@ void BlitImageHelper::BlitColor(GLuint dst_framebuffer, GLuint src_image_view, G
glDisable(GL_ALPHA_TEST);
glDisablei(GL_BLEND, 0);
glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
- glCullFace(GL_BACK);
glFrontFace(GL_CW);
glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
glDepthRangeIndexed(0, 0.0, 0.0);
diff --git a/src/video_core/renderer_opengl/gl_fence_manager.cpp b/src/video_core/renderer_opengl/gl_fence_manager.cpp
index 91463f854..5326172af 100644
--- a/src/video_core/renderer_opengl/gl_fence_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_fence_manager.cpp
@@ -27,9 +27,7 @@ bool GLInnerFence::IsSignaled() const {
return true;
}
ASSERT(sync_object.handle != 0);
- GLint sync_status;
- glGetSynciv(sync_object.handle, GL_SYNC_STATUS, 1, nullptr, &sync_status);
- return sync_status == GL_SIGNALED;
+ return sync_object.IsSignaled();
}
void GLInnerFence::Wait() {
diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
index 29491e762..89000d6e0 100644
--- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
+++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
@@ -621,10 +621,7 @@ bool GraphicsPipeline::IsBuilt() noexcept {
if (built_fence.handle == 0) {
return false;
}
- // Timeout of zero means this is non-blocking
- const auto sync_status = glClientWaitSync(built_fence.handle, 0, 0);
- ASSERT(sync_status != GL_WAIT_FAILED);
- is_built = sync_status != GL_TIMEOUT_EXPIRED;
+ is_built = built_fence.IsSignaled();
return is_built;
}
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 7bced675c..90e35e307 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -63,7 +63,7 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra
buffer_cache(*this, cpu_memory_, buffer_cache_runtime),
shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager,
state_tracker, gpu.ShaderNotify()),
- query_cache(*this), accelerate_dma(buffer_cache),
+ query_cache(*this), accelerate_dma(buffer_cache, texture_cache),
fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache),
blit_image(program_manager_) {}
@@ -357,6 +357,7 @@ void RasterizerOpenGL::DrawTexture() {
.y = static_cast<s32>(draw_texture_state.src_y1)}};
blit_image.BlitColor(texture_cache.GetFramebuffer()->Handle(), texture.DefaultHandle(),
sampler->Handle(), dst_region, src_region, texture.size);
+ state_tracker.InvalidateState();
}
++num_queued_commands;
@@ -576,7 +577,7 @@ bool RasterizerOpenGL::AccelerateConditionalRendering() {
// Reimplement Host conditional rendering.
return false;
}
- // Medium / Low Hack: stub any checks on queries writen into the buffer cache.
+ // Medium / Low Hack: stub any checks on queries written into the buffer cache.
const GPUVAddr condition_address{maxwell3d->regs.render_enable.Address()};
Maxwell::ReportSemaphore::Compare cmp;
if (gpu_memory->IsMemoryDirty(condition_address, sizeof(cmp),
@@ -1262,7 +1263,8 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
query_cache.EraseChannel(channel_id);
}
-AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {}
+AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_, TextureCache& texture_cache_)
+ : buffer_cache{buffer_cache_}, texture_cache{texture_cache_} {}
bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {
std::scoped_lock lock{buffer_cache.mutex};
@@ -1274,4 +1276,44 @@ bool AccelerateDMA::BufferClear(GPUVAddr src_address, u64 amount, u32 value) {
return buffer_cache.DMAClear(src_address, amount, value);
}
+template <bool IS_IMAGE_UPLOAD>
+bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
+ const Tegra::DMA::BufferOperand& buffer_operand,
+ const Tegra::DMA::ImageOperand& image_operand) {
+ std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+ const auto image_id = texture_cache.DmaImageId(image_operand);
+ if (image_id == VideoCommon::NULL_IMAGE_ID) {
+ return false;
+ }
+ const u32 buffer_size = static_cast<u32>(buffer_operand.pitch * buffer_operand.height);
+ static constexpr auto sync_info = VideoCommon::ObtainBufferSynchronize::FullSynchronize;
+ const auto post_op = IS_IMAGE_UPLOAD ? VideoCommon::ObtainBufferOperation::DoNothing
+ : VideoCommon::ObtainBufferOperation::MarkAsWritten;
+ const auto [buffer, offset] =
+ buffer_cache.ObtainBuffer(buffer_operand.address, buffer_size, sync_info, post_op);
+
+ const auto [image, copy] = texture_cache.DmaBufferImageCopy(
+ copy_info, buffer_operand, image_operand, image_id, IS_IMAGE_UPLOAD);
+ const std::span copy_span{&copy, 1};
+
+ if constexpr (IS_IMAGE_UPLOAD) {
+ image->UploadMemory(buffer->Handle(), offset, copy_span);
+ } else {
+ image->DownloadMemory(buffer->Handle(), offset, copy_span);
+ }
+ return true;
+}
+
+bool AccelerateDMA::ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info,
+ const Tegra::DMA::ImageOperand& image_operand,
+ const Tegra::DMA::BufferOperand& buffer_operand) {
+ return DmaBufferImageCopy<false>(copy_info, buffer_operand, image_operand);
+}
+
+bool AccelerateDMA::BufferToImage(const Tegra::DMA::ImageCopy& copy_info,
+ const Tegra::DMA::BufferOperand& buffer_operand,
+ const Tegra::DMA::ImageOperand& image_operand) {
+ return DmaBufferImageCopy<true>(copy_info, buffer_operand, image_operand);
+}
+
} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 7e21fc43d..ad6978bd0 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -50,24 +50,26 @@ static_assert(sizeof(BindlessSSBO) * CHAR_BIT == 128);
class AccelerateDMA : public Tegra::Engines::AccelerateDMAInterface {
public:
- explicit AccelerateDMA(BufferCache& buffer_cache);
+ explicit AccelerateDMA(BufferCache& buffer_cache, TextureCache& texture_cache);
bool BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) override;
bool BufferClear(GPUVAddr src_address, u64 amount, u32 value) override;
bool ImageToBuffer(const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::ImageOperand& src,
- const Tegra::DMA::BufferOperand& dst) override {
- return false;
- }
+ const Tegra::DMA::BufferOperand& dst) override;
bool BufferToImage(const Tegra::DMA::ImageCopy& copy_info, const Tegra::DMA::BufferOperand& src,
- const Tegra::DMA::ImageOperand& dst) override {
- return false;
- }
+ const Tegra::DMA::ImageOperand& dst) override;
private:
+ template <bool IS_IMAGE_UPLOAD>
+ bool DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
+ const Tegra::DMA::BufferOperand& src,
+ const Tegra::DMA::ImageOperand& dst);
+
BufferCache& buffer_cache;
+ TextureCache& texture_cache;
};
class RasterizerOpenGL : public VideoCore::RasterizerAccelerated,
@@ -160,7 +162,7 @@ private:
/// Syncs the cull mode to match the guest state
void SyncCullMode();
- /// Syncs the primitve restart to match the guest state
+ /// Syncs the primitive restart to match the guest state
void SyncPrimitiveRestart();
/// Syncs the depth test state to match the guest state
@@ -244,7 +246,7 @@ private:
std::array<GLuint, MAX_TEXTURES> texture_handles{};
std::array<GLuint, MAX_IMAGES> image_handles{};
- /// Number of commands queued to the OpenGL driver. Resetted on flush.
+ /// Number of commands queued to the OpenGL driver. Reset on flush.
size_t num_queued_commands = 0;
bool has_written_global_memory = false;
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp
index 3a664fdec..eae8fd110 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp
@@ -3,6 +3,7 @@
#include <string_view>
#include <glad/glad.h>
+#include "common/assert.h"
#include "common/microprofile.h"
#include "video_core/renderer_opengl/gl_resource_manager.h"
#include "video_core/renderer_opengl/gl_shader_util.h"
@@ -158,6 +159,15 @@ void OGLSync::Release() {
handle = 0;
}
+bool OGLSync::IsSignaled() const noexcept {
+ // At least on Nvidia, glClientWaitSync with a timeout of 0
+ // is faster than glGetSynciv of GL_SYNC_STATUS.
+ // Timeout of 0 means this check is non-blocking.
+ const auto sync_status = glClientWaitSync(handle, 0, 0);
+ ASSERT(sync_status != GL_WAIT_FAILED);
+ return sync_status != GL_TIMEOUT_EXPIRED;
+}
+
void OGLFramebuffer::Create() {
if (handle != 0)
return;
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h
index bc05ba4bd..77362acd2 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.h
+++ b/src/video_core/renderer_opengl/gl_resource_manager.h
@@ -263,6 +263,9 @@ public:
/// Deletes the internal OpenGL resource
void Release();
+ /// Checks if the sync has been signaled
+ bool IsSignaled() const noexcept;
+
GLsync handle = 0;
};
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index d3eabd686..0b9c4a904 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -717,9 +717,7 @@ std::optional<size_t> TextureCacheRuntime::StagingBuffers::FindBuffer(size_t req
continue;
}
if (syncs[index].handle != 0) {
- GLint status;
- glGetSynciv(syncs[index].handle, GL_SYNC_STATUS, 1, nullptr, &status);
- if (status != GL_SIGNALED) {
+ if (!syncs[index].IsSignaled()) {
continue;
}
syncs[index].Release();
@@ -765,14 +763,14 @@ Image::Image(const VideoCommon::NullImageParams& params) : VideoCommon::ImageBas
Image::~Image() = default;
-void Image::UploadMemory(const ImageBufferMap& map,
+void Image::UploadMemory(GLuint buffer_handle, size_t buffer_offset,
std::span<const VideoCommon::BufferImageCopy> copies) {
const bool is_rescaled = True(flags & ImageFlagBits::Rescaled);
if (is_rescaled) {
ScaleDown(true);
}
- glBindBuffer(GL_PIXEL_UNPACK_BUFFER, map.buffer);
- glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, map.offset, unswizzled_size_bytes);
+ glBindBuffer(GL_PIXEL_UNPACK_BUFFER, buffer_handle);
+ glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, buffer_offset, unswizzled_size_bytes);
glPixelStorei(GL_UNPACK_ALIGNMENT, 1);
@@ -791,21 +789,26 @@ void Image::UploadMemory(const ImageBufferMap& map,
current_image_height = copy.buffer_image_height;
glPixelStorei(GL_UNPACK_IMAGE_HEIGHT, current_image_height);
}
- CopyBufferToImage(copy, map.offset);
+ CopyBufferToImage(copy, buffer_offset);
}
if (is_rescaled) {
ScaleUp();
}
}
-void Image::DownloadMemory(ImageBufferMap& map,
+void Image::UploadMemory(const ImageBufferMap& map,
+ std::span<const VideoCommon::BufferImageCopy> copies) {
+ UploadMemory(map.buffer, map.offset, copies);
+}
+
+void Image::DownloadMemory(GLuint buffer_handle, size_t buffer_offset,
std::span<const VideoCommon::BufferImageCopy> copies) {
const bool is_rescaled = True(flags & ImageFlagBits::Rescaled);
if (is_rescaled) {
ScaleDown();
}
glMemoryBarrier(GL_PIXEL_BUFFER_BARRIER_BIT); // TODO: Move this to its own API
- glBindBuffer(GL_PIXEL_PACK_BUFFER, map.buffer);
+ glBindBuffer(GL_PIXEL_PACK_BUFFER, buffer_handle);
glPixelStorei(GL_PACK_ALIGNMENT, 1);
u32 current_row_length = std::numeric_limits<u32>::max();
@@ -823,13 +826,18 @@ void Image::DownloadMemory(ImageBufferMap& map,
current_image_height = copy.buffer_image_height;
glPixelStorei(GL_PACK_IMAGE_HEIGHT, current_image_height);
}
- CopyImageToBuffer(copy, map.offset);
+ CopyImageToBuffer(copy, buffer_offset);
}
if (is_rescaled) {
ScaleUp(true);
}
}
+void Image::DownloadMemory(ImageBufferMap& map,
+ std::span<const VideoCommon::BufferImageCopy> copies) {
+ DownloadMemory(map.buffer, map.offset, copies);
+}
+
GLuint Image::StorageHandle() noexcept {
switch (info.format) {
case PixelFormat::A8B8G8R8_SRGB:
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h
index e30875496..911e4607a 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.h
+++ b/src/video_core/renderer_opengl/gl_texture_cache.h
@@ -206,9 +206,15 @@ public:
Image(Image&&) = default;
Image& operator=(Image&&) = default;
+ void UploadMemory(GLuint buffer_handle, size_t buffer_offset,
+ std::span<const VideoCommon::BufferImageCopy> copies);
+
void UploadMemory(const ImageBufferMap& map,
std::span<const VideoCommon::BufferImageCopy> copies);
+ void DownloadMemory(GLuint buffer_handle, size_t buffer_offset,
+ std::span<const VideoCommon::BufferImageCopy> copies);
+
void DownloadMemory(ImageBufferMap& map, std::span<const VideoCommon::BufferImageCopy> copies);
GLuint StorageHandle() noexcept;
generated by cgit v1.2.3 (git 2.25.1) at 2024-07-04 13:57:35 +0200