Mirrors/shadPS4
1
0
Fork 0
mirror of https://github.com/shadps4-emu/shadPS4.git synced 2025-06-04 17:53:17 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

shader_recompiler: Implement AMD buffer bounds checking behavior. (#2448)

Browse source 
* shader_recompiler: Implement AMD buffer bounds checking behavior.

* shader_recompiler: Use SRT flatbuf for bounds check size.

* shader_recompiler: Fix buffer atomic bounds check.

* buffer_cache: Prevent false image-to-buffer sync.

Lowering vertex fetch to formatted buffer surfaced an issue where a CPU modified range may be overwritten with stale GPU modified image data.

* Address review comments.
This commit is contained in:
squidbus 2025-02-17 06:13:39 -08:00 • committed by GitHub
parent b06790dfe5
commit fd3d3c4158
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
19 changed files with 376 additions and 158 deletions
Download patch file Download diff file Expand all files Collapse all files

6
src/video_core/buffer_cache/buffer_cache.cpp
View file

@ -608,7 +608,11 @@ bool BufferCache::SynchronizeBufferFromImage(Buffer& buffer, VAddr device_addr,
return false;
}
Image& image = texture_cache.GetImage(image_id);
if (False(image.flags & ImageFlagBits::GpuModified)) {
// Only perform sync if image is:
// - GPU modified; otherwise there are no changes to synchronize.
// - Not CPU modified; otherwise we could overwrite CPU changes with stale GPU changes.
if (False(image.flags & ImageFlagBits::GpuModified) ||
True(image.flags & ImageFlagBits::CpuDirty)) {
return false;
}
ASSERT_MSG(device_addr == image.info.guest_address,

37
src/video_core/renderer_vulkan/vk_instance.cpp
View file

@ -210,9 +210,6 @@ bool Instance::CreateDevice() {
vk::PhysicalDevicePrimitiveTopologyListRestartFeaturesEXT,
vk::PhysicalDevicePortabilitySubsetFeaturesKHR>();
features = feature_chain.get().features;
#ifdef __APPLE__
portability_features = feature_chain.get<vk::PhysicalDevicePortabilitySubsetFeaturesKHR>();
#endif
const vk::StructureChain properties_chain = physical_device.getProperties2<
vk::PhysicalDeviceProperties2, vk::PhysicalDeviceVulkan11Properties,
@ -258,16 +255,19 @@ bool Instance::CreateDevice() {
add_extension(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
add_extension(VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME);
add_extension(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME);
dynamic_color_write_mask = add_extension(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME);
if (dynamic_color_write_mask) {
dynamic_color_write_mask =
feature_chain.get<vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT>()
.extendedDynamicState3ColorWriteMask;
dynamic_state_3 = add_extension(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME);
if (dynamic_state_3) {
dynamic_state_3_features =
feature_chain.get<vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT>();
LOG_INFO(Render_Vulkan, "- extendedDynamicState3ColorWriteMask: {}",
dynamic_state_3_features.extendedDynamicState3ColorWriteMask);
}
null_descriptor = add_extension(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
if (null_descriptor) {
null_descriptor =
feature_chain.get<vk::PhysicalDeviceRobustness2FeaturesEXT>().nullDescriptor;
robustness2 = add_extension(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
if (robustness2) {
robustness2_features = feature_chain.get<vk::PhysicalDeviceRobustness2FeaturesEXT>();
LOG_INFO(Render_Vulkan, "- robustBufferAccess2: {}",
robustness2_features.robustBufferAccess2);
LOG_INFO(Render_Vulkan, "- nullDescriptor: {}", robustness2_features.nullDescriptor);
}
custom_border_color = add_extension(VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME);
depth_clip_control = add_extension(VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME);
@ -284,6 +284,9 @@ bool Instance::CreateDevice() {
#ifdef __APPLE__
// Required by Vulkan spec if supported.
portability_subset = add_extension(VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME);
if (portability_subset) {
portability_features = feature_chain.get<vk::PhysicalDevicePortabilitySubsetFeaturesKHR>();
}
#endif
const auto family_properties = physical_device.getQueueFamilyProperties();
@ -387,13 +390,15 @@ bool Instance::CreateDevice() {
.customBorderColorWithoutFormat = true,
},
vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT{
.extendedDynamicState3ColorWriteMask = true,
.extendedDynamicState3ColorWriteMask =
dynamic_state_3_features.extendedDynamicState3ColorWriteMask,
},
vk::PhysicalDeviceDepthClipControlFeaturesEXT{
.depthClipControl = true,
},
vk::PhysicalDeviceRobustness2FeaturesEXT{
.nullDescriptor = true,
.robustBufferAccess2 = robustness2_features.robustBufferAccess2,
.nullDescriptor = robustness2_features.nullDescriptor,
},
vk::PhysicalDeviceVertexInputDynamicStateFeaturesEXT{
.vertexInputDynamicState = true,
@ -420,13 +425,13 @@ bool Instance::CreateDevice() {
if (!custom_border_color) {
device_chain.unlink<vk::PhysicalDeviceCustomBorderColorFeaturesEXT>();
}
if (!dynamic_color_write_mask) {
if (!dynamic_state_3) {
device_chain.unlink<vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT>();
}
if (!depth_clip_control) {
device_chain.unlink<vk::PhysicalDeviceDepthClipControlFeaturesEXT>();
}
if (!null_descriptor) {
if (!robustness2) {
device_chain.unlink<vk::PhysicalDeviceRobustness2FeaturesEXT>();
}
if (!vertex_input_dynamic_state) {

18
src/video_core/renderer_vulkan/vk_instance.h
View file

@ -99,9 +99,10 @@ public:
return depth_clip_control;
}
/// Returns true when dynamic color write mask state is supported
/// Returns true when the extendedDynamicState3ColorWriteMask feature of
/// VK_EXT_extended_dynamic_state3 is supported.
bool IsDynamicColorWriteMaskSupported() const {
return dynamic_color_write_mask;
return dynamic_state_3 && dynamic_state_3_features.extendedDynamicState3ColorWriteMask;
}
/// Returns true when VK_EXT_vertex_input_dynamic_state is supported.
@ -109,9 +110,14 @@ public:
return vertex_input_dynamic_state;
}
/// Returns true when the robustBufferAccess2 feature of VK_EXT_robustness2 is supported.
bool IsRobustBufferAccess2Supported() const {
return robustness2 && robustness2_features.robustBufferAccess2;
}
/// Returns true when the nullDescriptor feature of VK_EXT_robustness2 is supported.
bool IsNullDescriptorSupported() const {
return null_descriptor;
return robustness2 && robustness2_features.nullDescriptor;
}
/// Returns true when VK_KHR_fragment_shader_barycentric is supported.
@ -303,6 +309,8 @@ private:
vk::PhysicalDevicePushDescriptorPropertiesKHR push_descriptor_props;
vk::PhysicalDeviceFeatures features;
vk::PhysicalDevicePortabilitySubsetFeaturesKHR portability_features;
vk::PhysicalDeviceExtendedDynamicState3FeaturesEXT dynamic_state_3_features;
vk::PhysicalDeviceRobustness2FeaturesEXT robustness2_features;
vk::DriverIdKHR driver_id;
vk::UniqueDebugUtilsMessengerEXT debug_callback{};
std::string vendor_name;
@ -317,9 +325,9 @@ private:
bool custom_border_color{};
bool fragment_shader_barycentric{};
bool depth_clip_control{};
bool dynamic_color_write_mask{};
bool dynamic_state_3{};
bool vertex_input_dynamic_state{};
bool null_descriptor{};
bool robustness2{};
bool list_restart{};
bool legacy_vertex_attributes{};
bool shader_stencil_export{};

1
src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
View file

@ -200,6 +200,7 @@ PipelineCache::PipelineCache(const Instance& instance_, Scheduler& scheduler_,
.support_legacy_vertex_attributes = instance_.IsLegacyVertexAttributesSupported(),
.supports_image_load_store_lod = instance_.IsImageLoadStoreLodSupported(),
.supports_native_cube_calc = instance_.IsAmdGcnShaderSupported(),
.supports_robust_buffer_access = instance_.IsRobustBufferAccess2Supported(),
.needs_manual_interpolation = instance.IsFragmentShaderBarycentricSupported() &&
instance.GetDriverID() == vk::DriverId::eNvidiaProprietary,
.needs_lds_barriers = instance.GetDriverID() == vk::DriverId::eNvidiaProprietary ||

1
src/video_core/renderer_vulkan/vk_rasterizer.cpp
View file

@ -447,7 +447,6 @@ bool Rasterizer::BindResources(const Pipeline* pipeline) {
set_writes.clear();
buffer_barriers.clear();
buffer_infos.clear();
buffer_views.clear();
image_infos.clear();
// Bind resource buffers and textures.

11
src/video_core/renderer_vulkan/vk_rasterizer.h
View file

@ -110,18 +110,17 @@ private:
std::pair<VideoCore::ImageId, VideoCore::TextureCache::RenderTargetDesc>, 8>
cb_descs;
std::optional<std::pair<VideoCore::ImageId, VideoCore::TextureCache::DepthTargetDesc>> db_desc;
boost::container::static_vector<vk::DescriptorImageInfo, 64> image_infos;
boost::container::static_vector<vk::BufferView, 16> buffer_views;
boost::container::static_vector<vk::DescriptorBufferInfo, 32> buffer_infos;
boost::container::static_vector<VideoCore::ImageId, 64> bound_images;
boost::container::static_vector<vk::DescriptorImageInfo, Shader::NumImages> image_infos;
boost::container::static_vector<vk::DescriptorBufferInfo, Shader::NumBuffers> buffer_infos;
boost::container::static_vector<VideoCore::ImageId, Shader::NumImages> bound_images;
Pipeline::DescriptorWrites set_writes;
Pipeline::BufferBarriers buffer_barriers;
using BufferBindingInfo = std::pair<VideoCore::BufferId, AmdGpu::Buffer>;
boost::container::static_vector<BufferBindingInfo, 32> buffer_bindings;
boost::container::static_vector<BufferBindingInfo, Shader::NumBuffers> buffer_bindings;
using ImageBindingInfo = std::pair<VideoCore::ImageId, VideoCore::TextureCache::TextureDesc>;
boost::container::static_vector<ImageBindingInfo, 64> image_bindings;
boost::container::static_vector<ImageBindingInfo, Shader::NumImages> image_bindings;
};
} // namespace Vulkan