aboutsummaryrefslogtreecommitdiff
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/gl_device.cpp43
-rw-r--r--src/video_core/renderer_opengl/gl_graphics_pipeline.cpp18
-rw-r--r--src/video_core/renderer_opengl/gl_graphics_pipeline.h2
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp7
4 files changed, 51 insertions, 19 deletions
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index ee140c9c2..94258ccd0 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -106,6 +106,43 @@ bool IsASTCSupported() {
return true;
}
+static bool HasSlowSoftwareAstc(std::string_view vendor_name, std::string_view renderer) {
+// ifdef for Unix reduces string comparisons for non-Windows drivers, and Intel
+#ifdef YUZU_UNIX
+ // Sorted vaguely by how likely a vendor is to appear
+ if (vendor_name == "AMD") {
+ // RadeonSI
+ return true;
+ }
+ if (vendor_name == "Intel") {
+ // Must be inside YUZU_UNIX ifdef as the Windows driver uses the same vendor string
+ // iris, crocus
+ const bool is_intel_dg = (renderer.find("DG") != std::string_view::npos);
+ return is_intel_dg;
+ }
+ if (vendor_name == "nouveau") {
+ return true;
+ }
+ if (vendor_name == "X.Org") {
+ // R600
+ return true;
+ }
+#endif
+ if (vendor_name == "Collabora Ltd") {
+ // Zink
+ return true;
+ }
+ if (vendor_name == "Microsoft Corporation") {
+ // d3d12
+ return true;
+ }
+ if (vendor_name == "Mesa/X.org") {
+ // llvmpipe, softpipe, virgl
+ return true;
+ }
+ return false;
+}
+
[[nodiscard]] bool IsDebugToolAttached(std::span<const std::string_view> extensions) {
const bool nsight = std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
return nsight || HasExtension(extensions, "GL_EXT_debug_tool") ||
@@ -120,12 +157,16 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
}
vendor_name = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
const std::string_view version = reinterpret_cast<const char*>(glGetString(GL_VERSION));
+ const std::string_view renderer = reinterpret_cast<const char*>(glGetString(GL_RENDERER));
const std::vector extensions = GetExtensions();
const bool is_nvidia = vendor_name == "NVIDIA Corporation";
const bool is_amd = vendor_name == "ATI Technologies Inc.";
const bool is_intel = vendor_name == "Intel";
+ const bool has_slow_software_astc =
+ !is_nvidia && !is_amd && HasSlowSoftwareAstc(vendor_name, renderer);
+
#ifdef __unix__
constexpr bool is_linux = true;
#else
@@ -152,7 +193,7 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
has_texture_shadow_lod = HasExtension(extensions, "GL_EXT_texture_shadow_lod");
- has_astc = IsASTCSupported();
+ has_astc = !has_slow_software_astc && IsASTCSupported();
has_variable_aoffi = TestVariableAoffi();
has_component_indexing_bug = is_amd;
has_precise_bug = TestPreciseBug();
diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
index f822fa856..44a771d65 100644
--- a/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
+++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.cpp
@@ -220,7 +220,8 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c
ASSERT(num_textures <= MAX_TEXTURES);
ASSERT(num_images <= MAX_IMAGES);
- const bool assembly_shaders{assembly_programs[0].handle != 0};
+ const auto backend = device.GetShaderBackend();
+ const bool assembly_shaders{backend == Settings::ShaderBackend::Glasm};
use_storage_buffers =
!assembly_shaders || num_storage_buffers <= device.GetMaxGLASMStorageBufferBlocks();
writes_global_memory &= !use_storage_buffers;
@@ -230,7 +231,6 @@ GraphicsPipeline::GraphicsPipeline(const Device& device, TextureCache& texture_c
GenerateTransformFeedbackState();
}
const bool in_parallel = thread_worker != nullptr;
- const auto backend = device.GetShaderBackend();
auto func{[this, sources_ = std::move(sources), sources_spirv_ = std::move(sources_spirv),
shader_notify, backend, in_parallel,
force_context_flush](ShaderContext::Context*) mutable {
@@ -559,15 +559,13 @@ void GraphicsPipeline::ConfigureImpl(bool is_indexed) {
}
void GraphicsPipeline::ConfigureTransformFeedbackImpl() const {
- glTransformFeedbackStreamAttribsNV(num_xfb_attribs, xfb_attribs.data(), num_xfb_strides,
- xfb_streams.data(), GL_INTERLEAVED_ATTRIBS);
+ glTransformFeedbackAttribsNV(num_xfb_attribs, xfb_attribs.data(), GL_SEPARATE_ATTRIBS);
}
void GraphicsPipeline::GenerateTransformFeedbackState() {
// TODO(Rodrigo): Inject SKIP_COMPONENTS*_NV when required. An unimplemented message will signal
// when this is required.
GLint* cursor{xfb_attribs.data()};
- GLint* current_stream{xfb_streams.data()};
for (size_t feedback = 0; feedback < Maxwell::NumTransformFeedbackBuffers; ++feedback) {
const auto& layout = key.xfb_state.layouts[feedback];
@@ -575,15 +573,6 @@ void GraphicsPipeline::GenerateTransformFeedbackState() {
if (layout.varying_count == 0) {
continue;
}
- *current_stream = static_cast<GLint>(feedback);
- if (current_stream != xfb_streams.data()) {
- // When stepping one stream, push the expected token
- cursor[0] = GL_NEXT_BUFFER_NV;
- cursor[1] = 0;
- cursor[2] = 0;
- cursor += XFB_ENTRY_STRIDE;
- }
- ++current_stream;
const auto& locations = key.xfb_state.varyings[feedback];
std::optional<u32> current_index;
@@ -619,7 +608,6 @@ void GraphicsPipeline::GenerateTransformFeedbackState() {
}
}
num_xfb_attribs = static_cast<GLsizei>((cursor - xfb_attribs.data()) / XFB_ENTRY_STRIDE);
- num_xfb_strides = static_cast<GLsizei>(current_stream - xfb_streams.data());
}
void GraphicsPipeline::WaitForBuild() {
diff --git a/src/video_core/renderer_opengl/gl_graphics_pipeline.h b/src/video_core/renderer_opengl/gl_graphics_pipeline.h
index 7b3d7eae8..74fc9cc3d 100644
--- a/src/video_core/renderer_opengl/gl_graphics_pipeline.h
+++ b/src/video_core/renderer_opengl/gl_graphics_pipeline.h
@@ -154,9 +154,7 @@ private:
static constexpr std::size_t XFB_ENTRY_STRIDE = 3;
GLsizei num_xfb_attribs{};
- GLsizei num_xfb_strides{};
std::array<GLint, 128 * XFB_ENTRY_STRIDE * Maxwell::NumTransformFeedbackBuffers> xfb_attribs{};
- std::array<GLint, Maxwell::NumTransformFeedbackBuffers> xfb_streams{};
std::mutex built_mutex;
std::condition_variable built_condvar;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index aadd6967c..1ba31be88 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -1335,7 +1335,8 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
}
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 = VideoCommon::ObtainBufferOperation::DoNothing;
+ 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);
@@ -1344,8 +1345,12 @@ bool AccelerateDMA::DmaBufferImageCopy(const Tegra::DMA::ImageCopy& copy_info,
const std::span copy_span{&copy, 1};
if constexpr (IS_IMAGE_UPLOAD) {
+ texture_cache.PrepareImage(image_id, true, false);
image->UploadMemory(buffer->Handle(), offset, copy_span);
} else {
+ if (offset % BytesPerBlock(image->info.format)) {
+ return false;
+ }
texture_cache.DownloadImageIntoBuffer(image, buffer->Handle(), offset, copy_span,
buffer_operand.address, buffer_size);
}
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-28 18:12:44 +0000