aboutsummaryrefslogtreecommitdiff
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/CMakeLists.txt2
-rw-r--r--src/video_core/compatible_formats.cpp162
-rw-r--r--src/video_core/compatible_formats.h32
-rw-r--r--src/video_core/gpu.cpp5
-rw-r--r--src/video_core/gpu.h6
-rw-r--r--src/video_core/gpu_asynch.cpp9
-rw-r--r--src/video_core/gpu_asynch.h2
-rw-r--r--src/video_core/gpu_synch.cpp8
-rw-r--r--src/video_core/gpu_synch.h2
-rw-r--r--src/video_core/gpu_thread.cpp7
-rw-r--r--src/video_core/macro/macro.cpp35
-rw-r--r--src/video_core/renderer_opengl/gl_device.cpp20
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp5
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h44
-rw-r--r--src/video_core/renderer_vulkan/maxwell_to_vk.cpp144
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.cpp28
-rw-r--r--src/video_core/renderer_vulkan/vk_scheduler.cpp2
-rw-r--r--src/video_core/renderer_vulkan/wrapper.cpp16
-rw-r--r--src/video_core/renderer_vulkan/wrapper.h4
-rw-r--r--src/video_core/texture_cache/texture_cache.h25
20 files changed, 407 insertions, 151 deletions
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 2dc752aa9..21c46a567 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -3,6 +3,8 @@ add_library(video_core STATIC
buffer_cache/buffer_cache.h
buffer_cache/map_interval.cpp
buffer_cache/map_interval.h
+ compatible_formats.cpp
+ compatible_formats.h
dirty_flags.cpp
dirty_flags.h
dma_pusher.cpp
diff --git a/src/video_core/compatible_formats.cpp b/src/video_core/compatible_formats.cpp
new file mode 100644
index 000000000..6c426b035
--- /dev/null
+++ b/src/video_core/compatible_formats.cpp
@@ -0,0 +1,162 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <bitset>
+#include <cstddef>
+
+#include "video_core/compatible_formats.h"
+#include "video_core/surface.h"
+
+namespace VideoCore::Surface {
+
+namespace {
+
+// Compatibility table taken from Table 3.X.2 in:
+// https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_texture_view.txt
+
+constexpr std::array VIEW_CLASS_128_BITS = {
+ PixelFormat::RGBA32F,
+ PixelFormat::RGBA32UI,
+};
+// Missing formats:
+// PixelFormat::RGBA32I
+
+constexpr std::array VIEW_CLASS_96_BITS = {
+ PixelFormat::RGB32F,
+};
+// Missing formats:
+// PixelFormat::RGB32UI,
+// PixelFormat::RGB32I,
+
+constexpr std::array VIEW_CLASS_64_BITS = {
+ PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI,
+ PixelFormat::RGBA16U, PixelFormat::RGBA16F, PixelFormat::RGBA16S,
+};
+// Missing formats:
+// PixelFormat::RGBA16I
+// PixelFormat::RG32I
+
+// TODO: How should we handle 48 bits?
+
+constexpr std::array VIEW_CLASS_32_BITS = {
+ PixelFormat::RG16F, PixelFormat::R11FG11FB10F, PixelFormat::R32F,
+ PixelFormat::A2B10G10R10U, PixelFormat::RG16UI, PixelFormat::R32UI,
+ PixelFormat::RG16I, PixelFormat::R32I, PixelFormat::ABGR8U,
+ PixelFormat::RG16, PixelFormat::ABGR8S, PixelFormat::RG16S,
+ PixelFormat::RGBA8_SRGB, PixelFormat::E5B9G9R9F, PixelFormat::BGRA8,
+ PixelFormat::BGRA8_SRGB,
+};
+// Missing formats:
+// PixelFormat::RGBA8UI
+// PixelFormat::RGBA8I
+// PixelFormat::RGB10_A2_UI
+
+// TODO: How should we handle 24 bits?
+
+constexpr std::array VIEW_CLASS_16_BITS = {
+ PixelFormat::R16F, PixelFormat::RG8UI, PixelFormat::R16UI, PixelFormat::R16I,
+ PixelFormat::RG8U, PixelFormat::R16U, PixelFormat::RG8S, PixelFormat::R16S,
+};
+// Missing formats:
+// PixelFormat::RG8I
+
+constexpr std::array VIEW_CLASS_8_BITS = {
+ PixelFormat::R8UI,
+ PixelFormat::R8U,
+};
+// Missing formats:
+// PixelFormat::R8I
+// PixelFormat::R8S
+
+constexpr std::array VIEW_CLASS_RGTC1_RED = {
+ PixelFormat::DXN1,
+};
+// Missing formats:
+// COMPRESSED_SIGNED_RED_RGTC1
+
+constexpr std::array VIEW_CLASS_RGTC2_RG = {
+ PixelFormat::DXN2UNORM,
+ PixelFormat::DXN2SNORM,
+};
+
+constexpr std::array VIEW_CLASS_BPTC_UNORM = {
+ PixelFormat::BC7U,
+ PixelFormat::BC7U_SRGB,
+};
+
+constexpr std::array VIEW_CLASS_BPTC_FLOAT = {
+ PixelFormat::BC6H_SF16,
+ PixelFormat::BC6H_UF16,
+};
+
+// Compatibility table taken from Table 4.X.1 in:
+// https://www.khronos.org/registry/OpenGL/extensions/ARB/ARB_copy_image.txt
+
+constexpr std::array COPY_CLASS_128_BITS = {
+ PixelFormat::RGBA32UI, PixelFormat::RGBA32F, PixelFormat::DXT23,
+ PixelFormat::DXT23_SRGB, PixelFormat::DXT45, PixelFormat::DXT45_SRGB,
+ PixelFormat::DXN2SNORM, PixelFormat::BC7U, PixelFormat::BC7U_SRGB,
+ PixelFormat::BC6H_SF16, PixelFormat::BC6H_UF16,
+};
+// Missing formats:
+// PixelFormat::RGBA32I
+// COMPRESSED_RG_RGTC2
+
+constexpr std::array COPY_CLASS_64_BITS = {
+ PixelFormat::RGBA16F, PixelFormat::RG32F, PixelFormat::RGBA16UI, PixelFormat::RG32UI,
+ PixelFormat::RGBA16U, PixelFormat::RGBA16S, PixelFormat::DXT1_SRGB, PixelFormat::DXT1,
+
+};
+// Missing formats:
+// PixelFormat::RGBA16I
+// PixelFormat::RG32I,
+// COMPRESSED_RGB_S3TC_DXT1_EXT
+// COMPRESSED_SRGB_S3TC_DXT1_EXT
+// COMPRESSED_RGBA_S3TC_DXT1_EXT
+// COMPRESSED_SIGNED_RED_RGTC1
+
+void Enable(FormatCompatibility::Table& compatiblity, size_t format_a, size_t format_b) {
+ compatiblity[format_a][format_b] = true;
+ compatiblity[format_b][format_a] = true;
+}
+
+void Enable(FormatCompatibility::Table& compatibility, PixelFormat format_a, PixelFormat format_b) {
+ Enable(compatibility, static_cast<size_t>(format_a), static_cast<size_t>(format_b));
+}
+
+template <typename Range>
+void EnableRange(FormatCompatibility::Table& compatibility, const Range& range) {
+ for (auto it_a = range.begin(); it_a != range.end(); ++it_a) {
+ for (auto it_b = it_a; it_b != range.end(); ++it_b) {
+ Enable(compatibility, *it_a, *it_b);
+ }
+ }
+}
+
+} // Anonymous namespace
+
+FormatCompatibility::FormatCompatibility() {
+ for (size_t i = 0; i < MaxPixelFormat; ++i) {
+ // Identity is allowed
+ Enable(view, i, i);
+ }
+
+ EnableRange(view, VIEW_CLASS_128_BITS);
+ EnableRange(view, VIEW_CLASS_96_BITS);
+ EnableRange(view, VIEW_CLASS_64_BITS);
+ EnableRange(view, VIEW_CLASS_32_BITS);
+ EnableRange(view, VIEW_CLASS_16_BITS);
+ EnableRange(view, VIEW_CLASS_8_BITS);
+ EnableRange(view, VIEW_CLASS_RGTC1_RED);
+ EnableRange(view, VIEW_CLASS_RGTC2_RG);
+ EnableRange(view, VIEW_CLASS_BPTC_UNORM);
+ EnableRange(view, VIEW_CLASS_BPTC_FLOAT);
+
+ copy = view;
+ EnableRange(copy, COPY_CLASS_128_BITS);
+ EnableRange(copy, COPY_CLASS_64_BITS);
+}
+
+} // namespace VideoCore::Surface
diff --git a/src/video_core/compatible_formats.h b/src/video_core/compatible_formats.h
new file mode 100644
index 000000000..d1082566d
--- /dev/null
+++ b/src/video_core/compatible_formats.h
@@ -0,0 +1,32 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <bitset>
+#include <cstddef>
+
+#include "video_core/surface.h"
+
+namespace VideoCore::Surface {
+
+class FormatCompatibility {
+public:
+ using Table = std::array<std::bitset<MaxPixelFormat>, MaxPixelFormat>;
+
+ explicit FormatCompatibility();
+
+ bool TestView(PixelFormat format_a, PixelFormat format_b) const noexcept {
+ return view[static_cast<size_t>(format_a)][static_cast<size_t>(format_b)];
+ }
+
+ bool TestCopy(PixelFormat format_a, PixelFormat format_b) const noexcept {
+ return copy[static_cast<size_t>(format_a)][static_cast<size_t>(format_b)];
+ }
+
+private:
+ Table view;
+ Table copy;
+};
+
+} // namespace VideoCore::Surface
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 8eb017f65..482e49711 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -2,6 +2,8 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include <chrono>
+
#include "common/assert.h"
#include "common/microprofile.h"
#include "core/core.h"
@@ -154,8 +156,7 @@ u64 GPU::GetTicks() const {
constexpr u64 gpu_ticks_num = 384;
constexpr u64 gpu_ticks_den = 625;
- const u64 cpu_ticks = system.CoreTiming().GetTicks();
- u64 nanoseconds = Core::Timing::CyclesToNs(cpu_ticks).count();
+ u64 nanoseconds = system.CoreTiming().GetGlobalTimeNs().count();
if (Settings::values.use_fast_gpu_time) {
nanoseconds /= 256;
}
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index a1b4c305c..2c42483bd 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -284,6 +284,12 @@ public:
/// core timing events.
virtual void Start() = 0;
+ /// Obtain the CPU Context
+ virtual void ObtainContext() = 0;
+
+ /// Release the CPU Context
+ virtual void ReleaseContext() = 0;
+
/// Push GPU command entries to be processed
virtual void PushGPUEntries(Tegra::CommandList&& entries) = 0;
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
index 53305ab43..7b855f63e 100644
--- a/src/video_core/gpu_asynch.cpp
+++ b/src/video_core/gpu_asynch.cpp
@@ -19,10 +19,17 @@ GPUAsynch::GPUAsynch(Core::System& system, std::unique_ptr<VideoCore::RendererBa
GPUAsynch::~GPUAsynch() = default;
void GPUAsynch::Start() {
- cpu_context->MakeCurrent();
gpu_thread.StartThread(*renderer, *gpu_context, *dma_pusher);
}
+void GPUAsynch::ObtainContext() {
+ cpu_context->MakeCurrent();
+}
+
+void GPUAsynch::ReleaseContext() {
+ cpu_context->DoneCurrent();
+}
+
void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
gpu_thread.SubmitList(std::move(entries));
}
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
index 517658612..15e9f1d38 100644
--- a/src/video_core/gpu_asynch.h
+++ b/src/video_core/gpu_asynch.h
@@ -25,6 +25,8 @@ public:
~GPUAsynch() override;
void Start() override;
+ void ObtainContext() override;
+ void ReleaseContext() override;
void PushGPUEntries(Tegra::CommandList&& entries) override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
void FlushRegion(VAddr addr, u64 size) override;
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
index 6f38a672a..aaeb9811d 100644
--- a/src/video_core/gpu_synch.cpp
+++ b/src/video_core/gpu_synch.cpp
@@ -13,10 +13,16 @@ GPUSynch::GPUSynch(Core::System& system, std::unique_ptr<VideoCore::RendererBase
GPUSynch::~GPUSynch() = default;
-void GPUSynch::Start() {
+void GPUSynch::Start() {}
+
+void GPUSynch::ObtainContext() {
context->MakeCurrent();
}
+void GPUSynch::ReleaseContext() {
+ context->DoneCurrent();
+}
+
void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
dma_pusher->Push(std::move(entries));
dma_pusher->DispatchCalls();
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
index 4a6e9a01d..762c20aa5 100644
--- a/src/video_core/gpu_synch.h
+++ b/src/video_core/gpu_synch.h
@@ -24,6 +24,8 @@ public:
~GPUSynch() override;
void Start() override;
+ void ObtainContext() override;
+ void ReleaseContext() override;
void PushGPUEntries(Tegra::CommandList&& entries) override;
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
void FlushRegion(VAddr addr, u64 size) override;
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index c3bb4fe06..738c6f0c1 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -4,6 +4,7 @@
#include "common/assert.h"
#include "common/microprofile.h"
+#include "common/thread.h"
#include "core/core.h"
#include "core/frontend/emu_window.h"
#include "core/settings.h"
@@ -18,7 +19,11 @@ namespace VideoCommon::GPUThread {
static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
SynchState& state) {
- MicroProfileOnThreadCreate("GpuThread");
+ std::string name = "yuzu:GPU";
+ MicroProfileOnThreadCreate(name.c_str());
+ Common::SetCurrentThreadName(name.c_str());
+ Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
+ system.RegisterHostThread();
// Wait for first GPU command before acquiring the window context
while (state.queue.Empty())
diff --git a/src/video_core/macro/macro.cpp b/src/video_core/macro/macro.cpp
index ef7dad349..a50e7b4e0 100644
--- a/src/video_core/macro/macro.cpp
+++ b/src/video_core/macro/macro.cpp
@@ -2,6 +2,7 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include <optional>
#include <boost/container_hash/hash.hpp>
#include "common/assert.h"
#include "common/logging/log.h"
@@ -35,22 +36,40 @@ void MacroEngine::Execute(Engines::Maxwell3D& maxwell3d, u32 method,
}
} else {
// Macro not compiled, check if it's uploaded and if so, compile it
- auto macro_code = uploaded_macro_code.find(method);
+ std::optional<u32> mid_method = std::nullopt;
+ const auto macro_code = uploaded_macro_code.find(method);
if (macro_code == uploaded_macro_code.end()) {
- UNREACHABLE_MSG("Macro 0x{0:x} was not uploaded", method);
- return;
+ for (const auto& [method_base, code] : uploaded_macro_code) {
+ if (method >= method_base && (method - method_base) < code.size()) {
+ mid_method = method_base;
+ break;
+ }
+ }
+ if (!mid_method.has_value()) {
+ UNREACHABLE_MSG("Macro 0x{0:x} was not uploaded", method);
+ return;
+ }
}
auto& cache_info = macro_cache[method];
- cache_info.hash = boost::hash_value(macro_code->second);
- cache_info.lle_program = Compile(macro_code->second);
+
+ if (!mid_method.has_value()) {
+ cache_info.lle_program = Compile(macro_code->second);
+ cache_info.hash = boost::hash_value(macro_code->second);
+ } else {
+ const auto& macro_cached = uploaded_macro_code[mid_method.value()];
+ const auto rebased_method = method - mid_method.value();
+ auto& code = uploaded_macro_code[method];
+ code.resize(macro_cached.size() - rebased_method);
+ std::memcpy(code.data(), macro_cached.data() + rebased_method,
+ code.size() * sizeof(u32));
+ cache_info.hash = boost::hash_value(code);
+ cache_info.lle_program = Compile(code);
+ }
auto hle_program = hle_macros->GetHLEProgram(cache_info.hash);
if (hle_program.has_value()) {
cache_info.has_hle_program = true;
cache_info.hle_program = std::move(hle_program.value());
- }
-
- if (cache_info.has_hle_program) {
cache_info.hle_program->Execute(parameters, method);
} else {
cache_info.lle_program->Execute(parameters, method);
diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp
index b6b6659c1..208fc6167 100644
--- a/src/video_core/renderer_opengl/gl_device.cpp
+++ b/src/video_core/renderer_opengl/gl_device.cpp
@@ -188,20 +188,6 @@ bool IsASTCSupported() {
return true;
}
-/// @brief Returns true when a GL_RENDERER is a Turing GPU
-/// @param renderer GL_RENDERER string
-bool IsTuring(std::string_view renderer) {
- static constexpr std::array<std::string_view, 12> TURING_GPUS = {
- "GTX 1650", "GTX 1660", "RTX 2060", "RTX 2070",
- "RTX 2080", "TITAN RTX", "Quadro RTX 3000", "Quadro RTX 4000",
- "Quadro RTX 5000", "Quadro RTX 6000", "Quadro RTX 8000", "Tesla T4",
- };
- return std::any_of(TURING_GPUS.begin(), TURING_GPUS.end(),
- [renderer](std::string_view candidate) {
- return renderer.find(candidate) != std::string_view::npos;
- });
-}
-
} // Anonymous namespace
Device::Device()
@@ -213,7 +199,6 @@ Device::Device()
const bool is_nvidia = vendor == "NVIDIA Corporation";
const bool is_amd = vendor == "ATI Technologies Inc.";
- const bool is_turing = is_nvidia && IsTuring(renderer);
bool disable_fast_buffer_sub_data = false;
if (is_nvidia && version == "4.6.0 NVIDIA 443.24") {
@@ -238,15 +223,12 @@ Device::Device()
has_component_indexing_bug = is_amd;
has_precise_bug = TestPreciseBug();
has_nv_viewport_array2 = GLAD_GL_NV_viewport_array2;
+ has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory;
// At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
// uniform buffers as "push constants"
has_fast_buffer_sub_data = is_nvidia && !disable_fast_buffer_sub_data;
- // Nvidia's driver on Turing GPUs randomly crashes when the buffer is made resident, or on
- // DeleteBuffers. Disable unified memory on these devices.
- has_vertex_buffer_unified_memory = GLAD_GL_NV_vertex_buffer_unified_memory && !is_turing;
-
use_assembly_shaders = Settings::values.use_assembly_shaders && GLAD_GL_NV_gpu_program5 &&
GLAD_GL_NV_compute_program5 && GLAD_GL_NV_transform_feedback &&
GLAD_GL_NV_transform_feedback2;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 362457ffe..e960a0ef1 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -213,9 +213,10 @@ void RasterizerOpenGL::SetupVertexFormat() {
if (attrib.type == Maxwell::VertexAttribute::Type::SignedInt ||
attrib.type == Maxwell::VertexAttribute::Type::UnsignedInt) {
glVertexAttribIFormat(gl_index, attrib.ComponentCount(),
- MaxwellToGL::VertexType(attrib), attrib.offset);
+ MaxwellToGL::VertexFormat(attrib), attrib.offset);
} else {
- glVertexAttribFormat(gl_index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
+ glVertexAttribFormat(gl_index, attrib.ComponentCount(),
+ MaxwellToGL::VertexFormat(attrib),
attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
}
glVertexAttribBinding(gl_index, attrib.buffer);
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 35e329240..fe9bd4b5a 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -24,10 +24,11 @@ namespace MaxwellToGL {
using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
+inline GLenum VertexFormat(Maxwell::VertexAttribute attrib) {
switch (attrib.type) {
- case Maxwell::VertexAttribute::Type::UnsignedInt:
case Maxwell::VertexAttribute::Type::UnsignedNorm:
+ case Maxwell::VertexAttribute::Type::UnsignedScaled:
+ case Maxwell::VertexAttribute::Type::UnsignedInt:
switch (attrib.size) {
case Maxwell::VertexAttribute::Size::Size_8:
case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -48,8 +49,9 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_UNSIGNED_INT_2_10_10_10_REV;
}
break;
- case Maxwell::VertexAttribute::Type::SignedInt:
case Maxwell::VertexAttribute::Type::SignedNorm:
+ case Maxwell::VertexAttribute::Type::SignedScaled:
+ case Maxwell::VertexAttribute::Type::SignedInt:
switch (attrib.size) {
case Maxwell::VertexAttribute::Size::Size_8:
case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -84,36 +86,8 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
return GL_FLOAT;
}
break;
- case Maxwell::VertexAttribute::Type::UnsignedScaled:
- switch (attrib.size) {
- case Maxwell::VertexAttribute::Size::Size_8:
- case Maxwell::VertexAttribute::Size::Size_8_8:
- case Maxwell::VertexAttribute::Size::Size_8_8_8:
- case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return GL_UNSIGNED_BYTE;
- case Maxwell::VertexAttribute::Size::Size_16:
- case Maxwell::VertexAttribute::Size::Size_16_16:
- case Maxwell::VertexAttribute::Size::Size_16_16_16:
- case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return GL_UNSIGNED_SHORT;
- }
- break;
- case Maxwell::VertexAttribute::Type::SignedScaled:
- switch (attrib.size) {
- case Maxwell::VertexAttribute::Size::Size_8:
- case Maxwell::VertexAttribute::Size::Size_8_8:
- case Maxwell::VertexAttribute::Size::Size_8_8_8:
- case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return GL_BYTE;
- case Maxwell::VertexAttribute::Size::Size_16:
- case Maxwell::VertexAttribute::Size::Size_16_16:
- case Maxwell::VertexAttribute::Size::Size_16_16_16:
- case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return GL_SHORT;
- }
- break;
}
- UNIMPLEMENTED_MSG("Unimplemented vertex type={} and size={}", attrib.TypeString(),
+ UNIMPLEMENTED_MSG("Unimplemented vertex format of type={} and size={}", attrib.TypeString(),
attrib.SizeString());
return {};
}
@@ -217,6 +191,12 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
} else {
return GL_MIRROR_CLAMP_TO_EDGE;
}
+ case Tegra::Texture::WrapMode::MirrorOnceClampOGL:
+ if (GL_EXT_texture_mirror_clamp) {
+ return GL_MIRROR_CLAMP_EXT;
+ } else {
+ return GL_MIRROR_CLAMP_TO_EDGE;
+ }
}
UNIMPLEMENTED_MSG("Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
return GL_REPEAT;
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 1f2b6734b..d7f1ae89f 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -294,6 +294,28 @@ VkPrimitiveTopology PrimitiveTopology([[maybe_unused]] const VKDevice& device,
VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
switch (type) {
+ case Maxwell::VertexAttribute::Type::UnsignedNorm:
+ switch (size) {
+ case Maxwell::VertexAttribute::Size::Size_8:
+ return VK_FORMAT_R8_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_8_8:
+ return VK_FORMAT_R8G8_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_8_8_8:
+ return VK_FORMAT_R8G8B8_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+ return VK_FORMAT_R8G8B8A8_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_16:
+ return VK_FORMAT_R16_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_16_16:
+ return VK_FORMAT_R16G16_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_16_16_16:
+ return VK_FORMAT_R16G16B16_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+ return VK_FORMAT_R16G16B16A16_UNORM;
+ case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+ return VK_FORMAT_A2B10G10R10_UNORM_PACK32;
+ }
+ break;
case Maxwell::VertexAttribute::Type::SignedNorm:
switch (size) {
case Maxwell::VertexAttribute::Size::Size_8:
@@ -314,62 +336,50 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib
return VK_FORMAT_R16G16B16A16_SNORM;
case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
return VK_FORMAT_A2B10G10R10_SNORM_PACK32;
- default:
- break;
}
break;
- case Maxwell::VertexAttribute::Type::UnsignedNorm:
+ case Maxwell::VertexAttribute::Type::UnsignedScaled:
switch (size) {
case Maxwell::VertexAttribute::Size::Size_8:
- return VK_FORMAT_R8_UNORM;
+ return VK_FORMAT_R8_USCALED;
case Maxwell::VertexAttribute::Size::Size_8_8:
- return VK_FORMAT_R8G8_UNORM;
+ return VK_FORMAT_R8G8_USCALED;
case Maxwell::VertexAttribute::Size::Size_8_8_8:
- return VK_FORMAT_R8G8B8_UNORM;
+ return VK_FORMAT_R8G8B8_USCALED;
case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return VK_FORMAT_R8G8B8A8_UNORM;
+ return VK_FORMAT_R8G8B8A8_USCALED;
case Maxwell::VertexAttribute::Size::Size_16:
- return VK_FORMAT_R16_UNORM;
+ return VK_FORMAT_R16_USCALED;
case Maxwell::VertexAttribute::Size::Size_16_16:
- return VK_FORMAT_R16G16_UNORM;
+ return VK_FORMAT_R16G16_USCALED;
case Maxwell::VertexAttribute::Size::Size_16_16_16:
- return VK_FORMAT_R16G16B16_UNORM;
+ return VK_FORMAT_R16G16B16_USCALED;
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return VK_FORMAT_R16G16B16A16_UNORM;
+ return VK_FORMAT_R16G16B16A16_USCALED;
case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
- return VK_FORMAT_A2B10G10R10_UNORM_PACK32;
- default:
- break;
+ return VK_FORMAT_A2B10G10R10_USCALED_PACK32;
}
break;
- case Maxwell::VertexAttribute::Type::SignedInt:
+ case Maxwell::VertexAttribute::Type::SignedScaled:
switch (size) {
case Maxwell::VertexAttribute::Size::Size_8:
- return VK_FORMAT_R8_SINT;
+ return VK_FORMAT_R8_SSCALED;
case Maxwell::VertexAttribute::Size::Size_8_8:
- return VK_FORMAT_R8G8_SINT;
+ return VK_FORMAT_R8G8_SSCALED;
case Maxwell::VertexAttribute::Size::Size_8_8_8:
- return VK_FORMAT_R8G8B8_SINT;
+ return VK_FORMAT_R8G8B8_SSCALED;
case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return VK_FORMAT_R8G8B8A8_SINT;
+ return VK_FORMAT_R8G8B8A8_SSCALED;
case Maxwell::VertexAttribute::Size::Size_16:
- return VK_FORMAT_R16_SINT;
+ return VK_FORMAT_R16_SSCALED;
case Maxwell::VertexAttribute::Size::Size_16_16:
- return VK_FORMAT_R16G16_SINT;
+ return VK_FORMAT_R16G16_SSCALED;
case Maxwell::VertexAttribute::Size::Size_16_16_16:
- return VK_FORMAT_R16G16B16_SINT;
+ return VK_FORMAT_R16G16B16_SSCALED;
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return VK_FORMAT_R16G16B16A16_SINT;
- case Maxwell::VertexAttribute::Size::Size_32:
- return VK_FORMAT_R32_SINT;
- case Maxwell::VertexAttribute::Size::Size_32_32:
- return VK_FORMAT_R32G32_SINT;
- case Maxwell::VertexAttribute::Size::Size_32_32_32:
- return VK_FORMAT_R32G32B32_SINT;
- case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
- return VK_FORMAT_R32G32B32A32_SINT;
- default:
- break;
+ return VK_FORMAT_R16G16B16A16_SSCALED;
+ case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+ return VK_FORMAT_A2B10G10R10_SSCALED_PACK32;
}
break;
case Maxwell::VertexAttribute::Type::UnsignedInt:
@@ -398,56 +408,50 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib
return VK_FORMAT_R32G32B32_UINT;
case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
return VK_FORMAT_R32G32B32A32_UINT;
- default:
- break;
+ case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+ return VK_FORMAT_A2B10G10R10_UINT_PACK32;
}
break;
- case Maxwell::VertexAttribute::Type::UnsignedScaled:
+ case Maxwell::VertexAttribute::Type::SignedInt:
switch (size) {
case Maxwell::VertexAttribute::Size::Size_8:
- return VK_FORMAT_R8_USCALED;
+ return VK_FORMAT_R8_SINT;
case Maxwell::VertexAttribute::Size::Size_8_8:
- return VK_FORMAT_R8G8_USCALED;
+ return VK_FORMAT_R8G8_SINT;
case Maxwell::VertexAttribute::Size::Size_8_8_8:
- return VK_FORMAT_R8G8B8_USCALED;
+ return VK_FORMAT_R8G8B8_SINT;
case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return VK_FORMAT_R8G8B8A8_USCALED;
+ return VK_FORMAT_R8G8B8A8_SINT;
case Maxwell::VertexAttribute::Size::Size_16:
- return VK_FORMAT_R16_USCALED;
+ return VK_FORMAT_R16_SINT;
case Maxwell::VertexAttribute::Size::Size_16_16:
- return VK_FORMAT_R16G16_USCALED;
+ return VK_FORMAT_R16G16_SINT;
case Maxwell::VertexAttribute::Size::Size_16_16_16:
- return VK_FORMAT_R16G16B16_USCALED;
+ return VK_FORMAT_R16G16B16_SINT;
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return VK_FORMAT_R16G16B16A16_USCALED;
- default:
- break;
+ return VK_FORMAT_R16G16B16A16_SINT;
+ case Maxwell::VertexAttribute::Size::Size_32:
+ return VK_FORMAT_R32_SINT;
+ case Maxwell::VertexAttribute::Size::Size_32_32:
+ return VK_FORMAT_R32G32_SINT;
+ case Maxwell::VertexAttribute::Size::Size_32_32_32:
+ return VK_FORMAT_R32G32B32_SINT;
+ case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
+ return VK_FORMAT_R32G32B32A32_SINT;
+ case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+ return VK_FORMAT_A2B10G10R10_SINT_PACK32;
}
break;
- case Maxwell::VertexAttribute::Type::SignedScaled:
+ case Maxwell::VertexAttribute::Type::Float:
switch (size) {
- case Maxwell::VertexAttribute::Size::Size_8:
- return VK_FORMAT_R8_SSCALED;
- case Maxwell::VertexAttribute::Size::Size_8_8:
- return VK_FORMAT_R8G8_SSCALED;
- case Maxwell::VertexAttribute::Size::Size_8_8_8:
- return VK_FORMAT_R8G8B8_SSCALED;
- case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
- return VK_FORMAT_R8G8B8A8_SSCALED;
case Maxwell::VertexAttribute::Size::Size_16:
- return VK_FORMAT_R16_SSCALED;
+ return VK_FORMAT_R16_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_16_16:
- return VK_FORMAT_R16G16_SSCALED;
+ return VK_FORMAT_R16G16_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_16_16_16:
- return VK_FORMAT_R16G16B16_SSCALED;
+ return VK_FORMAT_R16G16B16_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return VK_FORMAT_R16G16B16A16_SSCALED;
- default:
- break;
- }
- break;
- case Maxwell::VertexAttribute::Type::Float:
- switch (size) {
+ return VK_FORMAT_R16G16B16A16_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_32:
return VK_FORMAT_R32_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_32_32:
@@ -456,16 +460,6 @@ VkFormat VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttrib
return VK_FORMAT_R32G32B32_SFLOAT;
case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
return VK_FORMAT_R32G32B32A32_SFLOAT;
- case Maxwell::VertexAttribute::Size::Size_16:
- return VK_FORMAT_R16_SFLOAT;
- case Maxwell::VertexAttribute::Size::Size_16_16:
- return VK_FORMAT_R16G16_SFLOAT;
- case Maxwell::VertexAttribute::Size::Size_16_16_16:
- return VK_FORMAT_R16G16B16_SFLOAT;
- case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
- return VK_FORMAT_R16G16B16A16_SFLOAT;
- default:
- break;
}
break;
}
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index cd9673d1f..2d9b18ed9 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -155,11 +155,31 @@ vk::Instance CreateInstance(Common::DynamicLibrary& library, vk::InstanceDispatc
}
}
- static constexpr std::array layers_data{"VK_LAYER_LUNARG_standard_validation"};
- vk::Span<const char*> layers = layers_data;
- if (!enable_layers) {
- layers = {};
+ std::vector<const char*> layers;
+ layers.reserve(1);
+ if (enable_layers) {
+ layers.push_back("VK_LAYER_KHRONOS_validation");
+ }
+
+ const std::optional layer_properties = vk::EnumerateInstanceLayerProperties(dld);
+ if (!layer_properties) {
+ LOG_ERROR(Render_Vulkan, "Failed to query layer properties, disabling layers");
+ layers.clear();
+ }
+
+ for (auto layer_it = layers.begin(); layer_it != layers.end();) {
+ const char* const layer = *layer_it;
+ const auto it = std::find_if(
+ layer_properties->begin(), layer_properties->end(),
+ [layer](const VkLayerProperties& prop) { return !std::strcmp(layer, prop.layerName); });
+ if (it == layer_properties->end()) {
+ LOG_ERROR(Render_Vulkan, "Layer {} not available, removing it", layer);
+ layer_it = layers.erase(layer_it);
+ } else {
+ ++layer_it;
+ }
}
+
vk::Instance instance = vk::Instance::Create(layers, extensions, dld);
if (!instance) {
LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance");
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
index 82ec9180e..56524e6f3 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.cpp
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -9,6 +9,7 @@
#include <utility>
#include "common/microprofile.h"
+#include "common/thread.h"
#include "video_core/renderer_vulkan/vk_device.h"
#include "video_core/renderer_vulkan/vk_query_cache.h"
#include "video_core/renderer_vulkan/vk_resource_manager.h"
@@ -133,6 +134,7 @@ void VKScheduler::BindGraphicsPipeline(VkPipeline pipeline) {
}
void VKScheduler::WorkerThread() {
+ Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
std::unique_lock lock{mutex};
do {
cv.wait(lock, [this] { return !chunk_queue.Empty() || quit; });
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
index 42eff85d3..0d485a662 100644
--- a/src/video_core/renderer_vulkan/wrapper.cpp
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -153,7 +153,8 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
bool Load(InstanceDispatch& dld) noexcept {
#define X(name) Proc(dld.name, dld, #name)
- return X(vkCreateInstance) && X(vkEnumerateInstanceExtensionProperties);
+ return X(vkCreateInstance) && X(vkEnumerateInstanceExtensionProperties) &&
+ X(vkEnumerateInstanceLayerProperties);
#undef X
}
@@ -770,4 +771,17 @@ std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProp
return properties;
}
+std::optional<std::vector<VkLayerProperties>> EnumerateInstanceLayerProperties(
+ const InstanceDispatch& dld) {
+ u32 num;
+ if (dld.vkEnumerateInstanceLayerProperties(&num, nullptr) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ std::vector<VkLayerProperties> properties(num);
+ if (dld.vkEnumerateInstanceLayerProperties(&num, properties.data()) != VK_SUCCESS) {
+ return std::nullopt;
+ }
+ return properties;
+}
+
} // namespace Vulkan::vk
diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h
index da42ca88e..d56fdb3f9 100644
--- a/src/video_core/renderer_vulkan/wrapper.h
+++ b/src/video_core/renderer_vulkan/wrapper.h
@@ -141,6 +141,7 @@ struct InstanceDispatch {
PFN_vkCreateInstance vkCreateInstance;
PFN_vkDestroyInstance vkDestroyInstance;
PFN_vkEnumerateInstanceExtensionProperties vkEnumerateInstanceExtensionProperties;
+ PFN_vkEnumerateInstanceLayerProperties vkEnumerateInstanceLayerProperties;
PFN_vkCreateDebugUtilsMessengerEXT vkCreateDebugUtilsMessengerEXT;
PFN_vkCreateDevice vkCreateDevice;
@@ -996,4 +997,7 @@ private:
std::optional<std::vector<VkExtensionProperties>> EnumerateInstanceExtensionProperties(
const InstanceDispatch& dld);
+std::optional<std::vector<VkLayerProperties>> EnumerateInstanceLayerProperties(
+ const InstanceDispatch& dld);
+
} // namespace Vulkan::vk
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 85075e868..6207d8dfe 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -24,6 +24,7 @@
#include "core/core.h"
#include "core/memory.h"
#include "core/settings.h"
+#include "video_core/compatible_formats.h"
#include "video_core/dirty_flags.h"
#include "video_core/engines/fermi_2d.h"
#include "video_core/engines/maxwell_3d.h"
@@ -47,8 +48,8 @@ class RasterizerInterface;
namespace VideoCommon {
+using VideoCore::Surface::FormatCompatibility;
using VideoCore::Surface::PixelFormat;
-
using VideoCore::Surface::SurfaceTarget;
using RenderTargetConfig = Tegra::Engines::Maxwell3D::Regs::RenderTargetConfig;
@@ -595,7 +596,7 @@ private:
} else {
new_surface = GetUncachedSurface(gpu_addr, params);
}
- const auto& final_params = new_surface->GetSurfaceParams();
+ const SurfaceParams& final_params = new_surface->GetSurfaceParams();
if (cr_params.type != final_params.type) {
if (Settings::IsGPULevelExtreme()) {
BufferCopy(current_surface, new_surface);
@@ -603,7 +604,7 @@ private:
} else {
std::vector<CopyParams> bricks = current_surface->BreakDown(final_params);
for (auto& brick : bricks) {
- ImageCopy(current_surface, new_surface, brick);
+ TryCopyImage(current_surface, new_surface, brick);
}
}
Unregister(current_surface);
@@ -694,7 +695,7 @@ private:
}
const CopyParams copy_params(0, 0, 0, 0, 0, base_layer, 0, mipmap, width, height,
src_params.depth);
- ImageCopy(surface, new_surface, copy_params);
+ TryCopyImage(surface, new_surface, copy_params);
}
}
if (passed_tests == 0) {
@@ -791,7 +792,7 @@ private:
const u32 width = params.width;
const u32 height = params.height;
const CopyParams copy_params(0, 0, 0, 0, 0, slice, 0, 0, width, height, 1);
- ImageCopy(surface, new_surface, copy_params);
+ TryCopyImage(surface, new_surface, copy_params);
}
for (const auto& surface : overlaps) {
Unregister(surface);
@@ -1192,6 +1193,19 @@ private:
return {};
}
+ /// Try to do an image copy logging when formats are incompatible.
+ void TryCopyImage(TSurface& src, TSurface& dst, const CopyParams& copy) {
+ const SurfaceParams& src_params = src->GetSurfaceParams();
+ const SurfaceParams& dst_params = dst->GetSurfaceParams();
+ if (!format_compatibility.TestCopy(src_params.pixel_format, dst_params.pixel_format)) {
+ LOG_ERROR(HW_GPU, "Illegal copy between formats={{{}, {}}}",
+ static_cast<int>(dst_params.pixel_format),
+ static_cast<int>(src_params.pixel_format));
+ return;
+ }
+ ImageCopy(src, dst, copy);
+ }
+
constexpr PixelFormat GetSiblingFormat(PixelFormat format) const {
return siblings_table[static_cast<std::size_t>(format)];
}
@@ -1241,6 +1255,7 @@ private:
VideoCore::RasterizerInterface& rasterizer;
FormatLookupTable format_lookup_table;
+ FormatCompatibility format_compatibility;
u64 ticks{};
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-29 16:45:52 +0000