summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--src/video_core/engines/maxwell_3d.h67
-rw-r--r--src/video_core/engines/shader_bytecode.h57
-rw-r--r--src/video_core/renderer_vulkan/declarations.h2
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.cpp264
-rw-r--r--src/video_core/renderer_vulkan/renderer_vulkan.h19
-rw-r--r--src/video_core/renderer_vulkan/vk_device.cpp92
-rw-r--r--src/video_core/renderer_vulkan/vk_device.h21
-rw-r--r--src/video_core/shader/decode/video.cpp58
-rw-r--r--src/video_core/shader/shader_ir.h3
-rw-r--r--src/video_core/textures/texture.h15
10 files changed, 495 insertions, 103 deletions
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index d24c9f657..4637ddabd 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -312,6 +312,35 @@ public:
}
};
+ struct MsaaSampleLocation {
+ union {
+ BitField<0, 4, u32> x0;
+ BitField<4, 4, u32> y0;
+ BitField<8, 4, u32> x1;
+ BitField<12, 4, u32> y1;
+ BitField<16, 4, u32> x2;
+ BitField<20, 4, u32> y2;
+ BitField<24, 4, u32> x3;
+ BitField<28, 4, u32> y3;
+ };
+
+ constexpr std::pair<u32, u32> Location(int index) const {
+ switch (index) {
+ case 0:
+ return {x0, y0};
+ case 1:
+ return {x1, y1};
+ case 2:
+ return {x2, y2};
+ case 3:
+ return {x3, y3};
+ default:
+ UNREACHABLE();
+ return {0, 0};
+ }
+ }
+ };
+
enum class DepthMode : u32 {
MinusOneToOne = 0,
ZeroToOne = 1,
@@ -793,7 +822,13 @@ public:
u32 rt_separate_frag_data;
- INSERT_UNION_PADDING_WORDS(0xC);
+ INSERT_UNION_PADDING_WORDS(0x1);
+
+ u32 multisample_raster_enable;
+ u32 multisample_raster_samples;
+ std::array<u32, 4> multisample_sample_mask;
+
+ INSERT_UNION_PADDING_WORDS(0x5);
struct {
u32 address_high;
@@ -830,7 +865,16 @@ public:
std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
- INSERT_UNION_PADDING_WORDS(0xF);
+ std::array<MsaaSampleLocation, 4> multisample_sample_locations;
+
+ INSERT_UNION_PADDING_WORDS(0x2);
+
+ union {
+ BitField<0, 1, u32> enable;
+ BitField<4, 3, u32> target;
+ } multisample_coverage_to_color;
+
+ INSERT_UNION_PADDING_WORDS(0x8);
struct {
union {
@@ -943,7 +987,7 @@ public:
CounterReset counter_reset;
- INSERT_UNION_PADDING_WORDS(0x1);
+ u32 multisample_enable;
u32 zeta_enable;
@@ -1007,7 +1051,11 @@ public:
float polygon_offset_units;
- INSERT_UNION_PADDING_WORDS(0x11);
+ INSERT_UNION_PADDING_WORDS(0x4);
+
+ Tegra::Texture::MsaaMode multisample_mode;
+
+ INSERT_UNION_PADDING_WORDS(0xC);
union {
BitField<2, 1, u32> coord_origin;
@@ -1507,12 +1555,17 @@ ASSERT_REG_POSITION(stencil_back_func_ref, 0x3D5);
ASSERT_REG_POSITION(stencil_back_mask, 0x3D6);
ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
ASSERT_REG_POSITION(color_mask_common, 0x3E4);
-ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
ASSERT_REG_POSITION(depth_bounds, 0x3E7);
+ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
+ASSERT_REG_POSITION(multisample_raster_enable, 0x3ED);
+ASSERT_REG_POSITION(multisample_raster_samples, 0x3EE);
+ASSERT_REG_POSITION(multisample_sample_mask, 0x3EF);
ASSERT_REG_POSITION(zeta, 0x3F8);
ASSERT_REG_POSITION(clear_flags, 0x43E);
ASSERT_REG_POSITION(fill_rectangle, 0x44F);
ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
+ASSERT_REG_POSITION(multisample_sample_locations, 0x478);
+ASSERT_REG_POSITION(multisample_coverage_to_color, 0x47E);
ASSERT_REG_POSITION(rt_control, 0x487);
ASSERT_REG_POSITION(zeta_width, 0x48a);
ASSERT_REG_POSITION(zeta_height, 0x48b);
@@ -1545,11 +1598,12 @@ ASSERT_REG_POSITION(samplecnt_enable, 0x545);
ASSERT_REG_POSITION(point_size, 0x546);
ASSERT_REG_POSITION(point_sprite_enable, 0x548);
ASSERT_REG_POSITION(counter_reset, 0x54C);
+ASSERT_REG_POSITION(multisample_enable, 0x54D);
ASSERT_REG_POSITION(zeta_enable, 0x54E);
ASSERT_REG_POSITION(multisample_control, 0x54F);
ASSERT_REG_POSITION(condition, 0x554);
ASSERT_REG_POSITION(tsc, 0x557);
-ASSERT_REG_POSITION(polygon_offset_factor, 0x55b);
+ASSERT_REG_POSITION(polygon_offset_factor, 0x55B);
ASSERT_REG_POSITION(tic, 0x55D);
ASSERT_REG_POSITION(stencil_two_side_enable, 0x565);
ASSERT_REG_POSITION(stencil_back_op_fail, 0x566);
@@ -1558,6 +1612,7 @@ ASSERT_REG_POSITION(stencil_back_op_zpass, 0x568);
ASSERT_REG_POSITION(stencil_back_func_func, 0x569);
ASSERT_REG_POSITION(framebuffer_srgb, 0x56E);
ASSERT_REG_POSITION(polygon_offset_units, 0x56F);
+ASSERT_REG_POSITION(multisample_mode, 0x574);
ASSERT_REG_POSITION(point_coord_replace, 0x581);
ASSERT_REG_POSITION(code_address, 0x582);
ASSERT_REG_POSITION(draw, 0x585);
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 498936f0c..c66c66f6c 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -290,6 +290,23 @@ enum class VmadShr : u64 {
Shr15 = 2,
};
+enum class VmnmxType : u64 {
+ Bits8,
+ Bits16,
+ Bits32,
+};
+
+enum class VmnmxOperation : u64 {
+ Mrg_16H = 0,
+ Mrg_16L = 1,
+ Mrg_8B0 = 2,
+ Mrg_8B2 = 3,
+ Acc = 4,
+ Min = 5,
+ Max = 6,
+ Nop = 7,
+};
+
enum class XmadMode : u64 {
None = 0,
CLo = 1,
@@ -1651,6 +1668,42 @@ union Instruction {
} vmad;
union {
+ BitField<54, 1, u64> is_dest_signed;
+ BitField<48, 1, u64> is_src_a_signed;
+ BitField<49, 1, u64> is_src_b_signed;
+ BitField<37, 2, u64> src_format_a;
+ BitField<29, 2, u64> src_format_b;
+ BitField<56, 1, u64> mx;
+ BitField<55, 1, u64> sat;
+ BitField<36, 2, u64> selector_a;
+ BitField<28, 2, u64> selector_b;
+ BitField<50, 1, u64> is_op_b_register;
+ BitField<51, 3, VmnmxOperation> operation;
+
+ VmnmxType SourceFormatA() const {
+ switch (src_format_a) {
+ case 0b11:
+ return VmnmxType::Bits32;
+ case 0b10:
+ return VmnmxType::Bits16;
+ default:
+ return VmnmxType::Bits8;
+ }
+ }
+
+ VmnmxType SourceFormatB() const {
+ switch (src_format_b) {
+ case 0b11:
+ return VmnmxType::Bits32;
+ case 0b10:
+ return VmnmxType::Bits16;
+ default:
+ return VmnmxType::Bits8;
+ }
+ }
+ } vmnmx;
+
+ union {
BitField<20, 16, u64> imm20_16;
BitField<35, 1, u64> high_b_rr; // used on RR
BitField<36, 1, u64> product_shift_left;
@@ -1763,6 +1816,7 @@ public:
MEMBAR,
VMAD,
VSETP,
+ VMNMX,
FFMA_IMM, // Fused Multiply and Add
FFMA_CR,
FFMA_RC,
@@ -2070,6 +2124,7 @@ private:
INST("1110111110011---", Id::MEMBAR, Type::Trivial, "MEMBAR"),
INST("01011111--------", Id::VMAD, Type::Video, "VMAD"),
INST("0101000011110---", Id::VSETP, Type::Video, "VSETP"),
+ INST("0011101---------", Id::VMNMX, Type::Video, "VMNMX"),
INST("0011001-1-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),
@@ -2170,7 +2225,7 @@ private:
INST("0011011-11111---", Id::SHF_LEFT_IMM, Type::Shift, "SHF_LEFT_IMM"),
INST("0100110011100---", Id::I2I_C, Type::Conversion, "I2I_C"),
INST("0101110011100---", Id::I2I_R, Type::Conversion, "I2I_R"),
- INST("0011101-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"),
+ INST("0011100-11100---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"),
INST("0100110010111---", Id::I2F_C, Type::Conversion, "I2F_C"),
INST("0101110010111---", Id::I2F_R, Type::Conversion, "I2F_R"),
INST("0011100-10111---", Id::I2F_IMM, Type::Conversion, "I2F_IMM"),
diff --git a/src/video_core/renderer_vulkan/declarations.h b/src/video_core/renderer_vulkan/declarations.h
index 323bf6b39..89a035ca4 100644
--- a/src/video_core/renderer_vulkan/declarations.h
+++ b/src/video_core/renderer_vulkan/declarations.h
@@ -39,6 +39,7 @@ using UniqueFence = UniqueHandle<vk::Fence>;
using UniqueFramebuffer = UniqueHandle<vk::Framebuffer>;
using UniqueImage = UniqueHandle<vk::Image>;
using UniqueImageView = UniqueHandle<vk::ImageView>;
+using UniqueInstance = UniqueHandle<vk::Instance>;
using UniqueIndirectCommandsLayoutNVX = UniqueHandle<vk::IndirectCommandsLayoutNVX>;
using UniqueObjectTableNVX = UniqueHandle<vk::ObjectTableNVX>;
using UniquePipeline = UniqueHandle<vk::Pipeline>;
@@ -50,6 +51,7 @@ using UniqueSampler = UniqueHandle<vk::Sampler>;
using UniqueSamplerYcbcrConversion = UniqueHandle<vk::SamplerYcbcrConversion>;
using UniqueSemaphore = UniqueHandle<vk::Semaphore>;
using UniqueShaderModule = UniqueHandle<vk::ShaderModule>;
+using UniqueSurfaceKHR = UniqueHandle<vk::SurfaceKHR>;
using UniqueSwapchainKHR = UniqueHandle<vk::SwapchainKHR>;
using UniqueValidationCacheEXT = UniqueHandle<vk::ValidationCacheEXT>;
using UniqueDebugReportCallbackEXT = UniqueHandle<vk::DebugReportCallbackEXT>;
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index 6953aaafe..9cdb4b627 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -2,13 +2,18 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include <algorithm>
+#include <array>
+#include <cstring>
#include <memory>
#include <optional>
+#include <string>
#include <vector>
#include <fmt/format.h>
#include "common/assert.h"
+#include "common/dynamic_library.h"
#include "common/logging/log.h"
#include "common/telemetry.h"
#include "core/core.h"
@@ -30,15 +35,30 @@
#include "video_core/renderer_vulkan/vk_state_tracker.h"
#include "video_core/renderer_vulkan/vk_swapchain.h"
+// Include these late to avoid changing Vulkan-Hpp's dynamic dispatcher size
+#ifdef _WIN32
+#include <windows.h>
+// ensure include order
+#include <vulkan/vulkan_win32.h>
+#endif
+
+#ifdef __linux__
+#include <X11/Xlib.h>
+#include <vulkan/vulkan_wayland.h>
+#include <vulkan/vulkan_xlib.h>
+#endif
+
namespace Vulkan {
namespace {
+using Core::Frontend::WindowSystemType;
+
VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_,
VkDebugUtilsMessageTypeFlagsEXT type,
const VkDebugUtilsMessengerCallbackDataEXT* data,
[[maybe_unused]] void* user_data) {
- const vk::DebugUtilsMessageSeverityFlagBitsEXT severity{severity_};
+ const auto severity{static_cast<vk::DebugUtilsMessageSeverityFlagBitsEXT>(severity_)};
const char* message{data->pMessage};
if (severity & vk::DebugUtilsMessageSeverityFlagBitsEXT::eError) {
@@ -53,6 +73,110 @@ VkBool32 DebugCallback(VkDebugUtilsMessageSeverityFlagBitsEXT severity_,
return VK_FALSE;
}
+Common::DynamicLibrary OpenVulkanLibrary() {
+ Common::DynamicLibrary library;
+#ifdef __APPLE__
+ // Check if a path to a specific Vulkan library has been specified.
+ char* libvulkan_env = getenv("LIBVULKAN_PATH");
+ if (!libvulkan_env || !library.Open(libvulkan_env)) {
+ // Use the libvulkan.dylib from the application bundle.
+ std::string filename = File::GetBundleDirectory() + "/Contents/Frameworks/libvulkan.dylib";
+ library.Open(filename.c_str());
+ }
+#else
+ std::string filename = Common::DynamicLibrary::GetVersionedFilename("vulkan", 1);
+ if (!library.Open(filename.c_str())) {
+ // Android devices may not have libvulkan.so.1, only libvulkan.so.
+ filename = Common::DynamicLibrary::GetVersionedFilename("vulkan");
+ library.Open(filename.c_str());
+ }
+#endif
+ return library;
+}
+
+UniqueInstance CreateInstance(Common::DynamicLibrary& library, vk::DispatchLoaderDynamic& dld,
+ WindowSystemType window_type = WindowSystemType::Headless,
+ bool enable_layers = false) {
+ if (!library.IsOpen()) {
+ LOG_ERROR(Render_Vulkan, "Vulkan library not available");
+ return UniqueInstance{};
+ }
+ PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr;
+ if (!library.GetSymbol("vkGetInstanceProcAddr", &vkGetInstanceProcAddr)) {
+ LOG_ERROR(Render_Vulkan, "vkGetInstanceProcAddr not present in Vulkan");
+ return UniqueInstance{};
+ }
+ dld.init(vkGetInstanceProcAddr);
+
+ std::vector<const char*> extensions;
+ extensions.reserve(4);
+ switch (window_type) {
+ case Core::Frontend::WindowSystemType::Headless:
+ break;
+#ifdef _WIN32
+ case Core::Frontend::WindowSystemType::Windows:
+ extensions.push_back(VK_KHR_WIN32_SURFACE_EXTENSION_NAME);
+ break;
+#endif
+#ifdef __linux__
+ case Core::Frontend::WindowSystemType::X11:
+ extensions.push_back(VK_KHR_XLIB_SURFACE_EXTENSION_NAME);
+ break;
+ case Core::Frontend::WindowSystemType::Wayland:
+ extensions.push_back(VK_KHR_WAYLAND_SURFACE_EXTENSION_NAME);
+ break;
+#endif
+ default:
+ LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+ break;
+ }
+ if (window_type != Core::Frontend::WindowSystemType::Headless) {
+ extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
+ }
+ if (enable_layers) {
+ extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
+ }
+
+ u32 num_properties;
+ if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, nullptr, dld) !=
+ vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to query number of extension properties");
+ return UniqueInstance{};
+ }
+ std::vector<vk::ExtensionProperties> properties(num_properties);
+ if (vk::enumerateInstanceExtensionProperties(nullptr, &num_properties, properties.data(),
+ dld) != vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to query extension properties");
+ return UniqueInstance{};
+ }
+
+ for (const char* extension : extensions) {
+ const auto it =
+ std::find_if(properties.begin(), properties.end(), [extension](const auto& prop) {
+ return !std::strcmp(extension, prop.extensionName);
+ });
+ if (it == properties.end()) {
+ LOG_ERROR(Render_Vulkan, "Required instance extension {} is not available", extension);
+ return UniqueInstance{};
+ }
+ }
+
+ const vk::ApplicationInfo application_info("yuzu Emulator", VK_MAKE_VERSION(0, 1, 0),
+ "yuzu Emulator", VK_MAKE_VERSION(0, 1, 0),
+ VK_API_VERSION_1_1);
+ const std::array layers = {"VK_LAYER_LUNARG_standard_validation"};
+ const vk::InstanceCreateInfo instance_ci(
+ {}, &application_info, enable_layers ? static_cast<u32>(layers.size()) : 0, layers.data(),
+ static_cast<u32>(extensions.size()), extensions.data());
+ vk::Instance unsafe_instance;
+ if (vk::createInstance(&instance_ci, nullptr, &unsafe_instance, dld) != vk::Result::eSuccess) {
+ LOG_ERROR(Render_Vulkan, "Failed to create Vulkan instance");
+ return UniqueInstance{};
+ }
+ dld.init(unsafe_instance);
+ return UniqueInstance(unsafe_instance, {nullptr, dld});
+}
+
std::string GetReadableVersion(u32 version) {
return fmt::format("{}.{}.{}", VK_VERSION_MAJOR(version), VK_VERSION_MINOR(version),
VK_VERSION_PATCH(version));
@@ -147,27 +271,12 @@ bool RendererVulkan::TryPresent(int /*timeout_ms*/) {
}
bool RendererVulkan::Init() {
- PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
- render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
- const vk::DispatchLoaderDynamic dldi(instance, vkGetInstanceProcAddr);
-
- std::optional<vk::DebugUtilsMessengerEXT> callback;
- if (Settings::values.renderer_debug && dldi.vkCreateDebugUtilsMessengerEXT) {
- callback = CreateDebugCallback(dldi);
- if (!callback) {
- return false;
- }
- }
-
- if (!PickDevices(dldi)) {
- if (callback) {
- instance.destroy(*callback, nullptr, dldi);
- }
+ library = OpenVulkanLibrary();
+ instance = CreateInstance(library, dld, render_window.GetWindowInfo().type,
+ Settings::values.renderer_debug);
+ if (!instance || !CreateDebugCallback() || !CreateSurface() || !PickDevices()) {
return false;
}
- debug_callback = UniqueDebugUtilsMessengerEXT(
- *callback, vk::ObjectDestroy<vk::Instance, vk::DispatchLoaderDynamic>(
- instance, nullptr, device->GetDispatchLoader()));
Report();
@@ -176,7 +285,7 @@ bool RendererVulkan::Init() {
resource_manager = std::make_unique<VKResourceManager>(*device);
const auto& framebuffer = render_window.GetFramebufferLayout();
- swapchain = std::make_unique<VKSwapchain>(surface, *device);
+ swapchain = std::make_unique<VKSwapchain>(*surface, *device);
swapchain->Create(framebuffer.width, framebuffer.height, false);
state_tracker = std::make_unique<StateTracker>(system);
@@ -213,8 +322,10 @@ void RendererVulkan::ShutDown() {
device.reset();
}
-std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback(
- const vk::DispatchLoaderDynamic& dldi) {
+bool RendererVulkan::CreateDebugCallback() {
+ if (!Settings::values.renderer_debug) {
+ return true;
+ }
const vk::DebugUtilsMessengerCreateInfoEXT callback_ci(
{},
vk::DebugUtilsMessageSeverityFlagBitsEXT::eError |
@@ -225,32 +336,88 @@ std::optional<vk::DebugUtilsMessengerEXT> RendererVulkan::CreateDebugCallback(
vk::DebugUtilsMessageTypeFlagBitsEXT::eValidation |
vk::DebugUtilsMessageTypeFlagBitsEXT::ePerformance,
&DebugCallback, nullptr);
- vk::DebugUtilsMessengerEXT callback;
- if (instance.createDebugUtilsMessengerEXT(&callback_ci, nullptr, &callback, dldi) !=
+ vk::DebugUtilsMessengerEXT unsafe_callback;
+ if (instance->createDebugUtilsMessengerEXT(&callback_ci, nullptr, &unsafe_callback, dld) !=
vk::Result::eSuccess) {
LOG_ERROR(Render_Vulkan, "Failed to create debug callback");
- return {};
+ return false;
+ }
+ debug_callback = UniqueDebugUtilsMessengerEXT(unsafe_callback, {*instance, nullptr, dld});
+ return true;
+}
+
+bool RendererVulkan::CreateSurface() {
+ [[maybe_unused]] const auto& window_info = render_window.GetWindowInfo();
+ VkSurfaceKHR unsafe_surface = nullptr;
+
+#ifdef _WIN32
+ if (window_info.type == Core::Frontend::WindowSystemType::Windows) {
+ const HWND hWnd = static_cast<HWND>(window_info.render_surface);
+ const VkWin32SurfaceCreateInfoKHR win32_ci{VK_STRUCTURE_TYPE_WIN32_SURFACE_CREATE_INFO_KHR,
+ nullptr, 0, nullptr, hWnd};
+ const auto vkCreateWin32SurfaceKHR = reinterpret_cast<PFN_vkCreateWin32SurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateWin32SurfaceKHR"));
+ if (!vkCreateWin32SurfaceKHR || vkCreateWin32SurfaceKHR(instance.get(), &win32_ci, nullptr,
+ &unsafe_surface) != VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Win32 surface");
+ return false;
+ }
+ }
+#endif
+#ifdef __linux__
+ if (window_info.type == Core::Frontend::WindowSystemType::X11) {
+ const VkXlibSurfaceCreateInfoKHR xlib_ci{
+ VK_STRUCTURE_TYPE_XLIB_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+ static_cast<Display*>(window_info.display_connection),
+ reinterpret_cast<Window>(window_info.render_surface)};
+ const auto vkCreateXlibSurfaceKHR = reinterpret_cast<PFN_vkCreateXlibSurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateXlibSurfaceKHR"));
+ if (!vkCreateXlibSurfaceKHR || vkCreateXlibSurfaceKHR(instance.get(), &xlib_ci, nullptr,
+ &unsafe_surface) != VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Xlib surface");
+ return false;
+ }
+ }
+ if (window_info.type == Core::Frontend::WindowSystemType::Wayland) {
+ const VkWaylandSurfaceCreateInfoKHR wayland_ci{
+ VK_STRUCTURE_TYPE_WAYLAND_SURFACE_CREATE_INFO_KHR, nullptr, 0,
+ static_cast<wl_display*>(window_info.display_connection),
+ static_cast<wl_surface*>(window_info.render_surface)};
+ const auto vkCreateWaylandSurfaceKHR = reinterpret_cast<PFN_vkCreateWaylandSurfaceKHR>(
+ dld.vkGetInstanceProcAddr(*instance, "vkCreateWaylandSurfaceKHR"));
+ if (!vkCreateWaylandSurfaceKHR ||
+ vkCreateWaylandSurfaceKHR(instance.get(), &wayland_ci, nullptr, &unsafe_surface) !=
+ VK_SUCCESS) {
+ LOG_ERROR(Render_Vulkan, "Failed to initialize Wayland surface");
+ return false;
+ }
+ }
+#endif
+ if (!unsafe_surface) {
+ LOG_ERROR(Render_Vulkan, "Presentation not supported on this platform");
+ return false;
}
- return callback;
+
+ surface = UniqueSurfaceKHR(unsafe_surface, {*instance, nullptr, dld});
+ return true;
}
-bool RendererVulkan::PickDevices(const vk::DispatchLoaderDynamic& dldi) {
- const auto devices = instance.enumeratePhysicalDevices(dldi);
+bool RendererVulkan::PickDevices() {
+ const auto devices = instance->enumeratePhysicalDevices(dld);
- // TODO(Rodrigo): Choose device from config file
const s32 device_index = Settings::values.vulkan_device;
if (device_index < 0 || device_index >= static_cast<s32>(devices.size())) {
LOG_ERROR(Render_Vulkan, "Invalid device index {}!", device_index);
return false;
}
- const vk::PhysicalDevice physical_device = devices[device_index];
+ const vk::PhysicalDevice physical_device = devices[static_cast<std::size_t>(device_index)];
- if (!VKDevice::IsSuitable(dldi, physical_device, surface)) {
+ if (!VKDevice::IsSuitable(physical_device, *surface, dld)) {
return false;
}
- device = std::make_unique<VKDevice>(dldi, physical_device, surface);
- return device->Create(dldi, instance);
+ device = std::make_unique<VKDevice>(dld, physical_device, *surface);
+ return device->Create(*instance);
}
void RendererVulkan::Report() const {
@@ -276,4 +443,33 @@ void RendererVulkan::Report() const {
telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
}
+std::vector<std::string> RendererVulkan::EnumerateDevices() {
+ // Avoid putting DispatchLoaderDynamic, it's too large
+ auto dld_memory = std::make_unique<vk::DispatchLoaderDynamic>();
+ auto& dld = *dld_memory;
+
+ Common::DynamicLibrary library = OpenVulkanLibrary();
+ UniqueInstance instance = CreateInstance(library, dld);
+ if (!instance) {
+ return {};
+ }
+
+ u32 num_devices;
+ if (instance->enumeratePhysicalDevices(&num_devices, nullptr, dld) != vk::Result::eSuccess) {
+ return {};
+ }
+ std::vector<vk::PhysicalDevice> devices(num_devices);
+ if (instance->enumeratePhysicalDevices(&num_devices, devices.data(), dld) !=
+ vk::Result::eSuccess) {
+ return {};
+ }
+
+ std::vector<std::string> names;
+ names.reserve(num_devices);
+ for (auto& device : devices) {
+ names.push_back(device.getProperties(dld).deviceName);
+ }
+ return names;
+}
+
} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index d14384e79..42e253de5 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -6,8 +6,11 @@
#include <memory>
#include <optional>
+#include <string>
#include <vector>
+#include "common/dynamic_library.h"
+
#include "video_core/renderer_base.h"
#include "video_core/renderer_vulkan/declarations.h"
@@ -44,18 +47,24 @@ public:
void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
bool TryPresent(int timeout_ms) override;
+ static std::vector<std::string> EnumerateDevices();
+
private:
- std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
- const vk::DispatchLoaderDynamic& dldi);
+ bool CreateDebugCallback();
- bool PickDevices(const vk::DispatchLoaderDynamic& dldi);
+ bool CreateSurface();
+
+ bool PickDevices();
void Report() const;
Core::System& system;
- vk::Instance instance;
- vk::SurfaceKHR surface;
+ Common::DynamicLibrary library;
+ vk::DispatchLoaderDynamic dld;
+
+ UniqueInstance instance;
+ UniqueSurfaceKHR surface;
VKScreenInfo screen_info;
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index 7aafb5e59..6f4ae9132 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -10,6 +10,7 @@
#include <string_view>
#include <thread>
#include <vector>
+
#include "common/assert.h"
#include "core/settings.h"
#include "video_core/renderer_vulkan/declarations.h"
@@ -35,20 +36,20 @@ void SetNext(void**& next, T& data) {
}
template <typename T>
-T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
+T GetFeatures(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) {
vk::PhysicalDeviceFeatures2 features;
T extension_features;
features.pNext = &extension_features;
- physical.getFeatures2(&features, dldi);
+ physical.getFeatures2(&features, dld);
return extension_features;
}
template <typename T>
-T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dldi) {
+T GetProperties(vk::PhysicalDevice physical, const vk::DispatchLoaderDynamic& dld) {
vk::PhysicalDeviceProperties2 properties;
T extension_properties;
properties.pNext = &extension_properties;
- physical.getProperties2(&properties, dldi);
+ physical.getProperties2(&properties, dld);
return extension_properties;
}
@@ -78,19 +79,19 @@ vk::FormatFeatureFlags GetFormatFeatures(vk::FormatProperties properties, Format
} // Anonymous namespace
-VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+VKDevice::VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical,
vk::SurfaceKHR surface)
- : physical{physical}, properties{physical.getProperties(dldi)},
- format_properties{GetFormatProperties(dldi, physical)} {
- SetupFamilies(dldi, surface);
- SetupFeatures(dldi);
+ : dld{dld}, physical{physical}, properties{physical.getProperties(dld)},
+ format_properties{GetFormatProperties(dld, physical)} {
+ SetupFamilies(surface);
+ SetupFeatures();
}
VKDevice::~VKDevice() = default;
-bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance) {
+bool VKDevice::Create(vk::Instance instance) {
const auto queue_cis = GetDeviceQueueCreateInfos();
- const std::vector extensions = LoadExtensions(dldi);
+ const std::vector extensions = LoadExtensions();
vk::PhysicalDeviceFeatures2 features2;
void** next = &features2.pNext;
@@ -165,15 +166,13 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
nullptr);
device_ci.pNext = &features2;
- vk::Device dummy_logical;
- if (physical.createDevice(&device_ci, nullptr, &dummy_logical, dldi) != vk::Result::eSuccess) {
+ vk::Device unsafe_logical;
+ if (physical.createDevice(&device_ci, nullptr, &unsafe_logical, dld) != vk::Result::eSuccess) {
LOG_CRITICAL(Render_Vulkan, "Logical device failed to be created!");
return false;
}
-
- dld.init(instance, dldi.vkGetInstanceProcAddr, dummy_logical, dldi.vkGetDeviceProcAddr);
- logical = UniqueDevice(
- dummy_logical, vk::ObjectDestroy<vk::NoParent, vk::DispatchLoaderDynamic>(nullptr, dld));
+ dld.init(instance, dld.vkGetInstanceProcAddr, unsafe_logical);
+ logical = UniqueDevice(unsafe_logical, {nullptr, dld});
CollectTelemetryParameters();
@@ -235,8 +234,8 @@ void VKDevice::ReportLoss() const {
// *(VKGraphicsPipeline*)data[0]
}
-bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
- const vk::DispatchLoaderDynamic& dldi) const {
+bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const {
+ // Disable for now to avoid converting ASTC twice.
static constexpr std::array astc_formats = {
vk::Format::eAstc4x4UnormBlock, vk::Format::eAstc4x4SrgbBlock,
vk::Format::eAstc5x4UnormBlock, vk::Format::eAstc5x4SrgbBlock,
@@ -260,7 +259,7 @@ bool VKDevice::IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features
vk::FormatFeatureFlagBits::eBlitDst | vk::FormatFeatureFlagBits::eTransferSrc |
vk::FormatFeatureFlagBits::eTransferDst};
for (const auto format : astc_formats) {
- const auto format_properties{physical.getFormatProperties(format, dldi)};
+ const auto format_properties{physical.getFormatProperties(format, dld)};
if (!(format_properties.optimalTilingFeatures & format_feature_usage)) {
return false;
}
@@ -279,11 +278,9 @@ bool VKDevice::IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlag
return (supported_usage & wanted_usage) == wanted_usage;
}
-bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
- vk::SurfaceKHR surface) {
- bool is_suitable = true;
-
- constexpr std::array required_extensions = {
+bool VKDevice::IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface,
+ const vk::DispatchLoaderDynamic& dld) {
+ static constexpr std::array required_extensions = {
VK_KHR_SWAPCHAIN_EXTENSION_NAME,
VK_KHR_16BIT_STORAGE_EXTENSION_NAME,
VK_KHR_8BIT_STORAGE_EXTENSION_NAME,
@@ -293,9 +290,10 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME,
};
+ bool is_suitable = true;
std::bitset<required_extensions.size()> available_extensions{};
- for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+ for (const auto& prop : physical.enumerateDeviceExtensionProperties(nullptr, dld)) {
for (std::size_t i = 0; i < required_extensions.size(); ++i) {
if (available_extensions[i]) {
continue;
@@ -315,7 +313,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
bool has_graphics{}, has_present{};
- const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+ const auto queue_family_properties = physical.getQueueFamilyProperties(dld);
for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
const auto& family = queue_family_properties[i];
if (family.queueCount == 0) {
@@ -323,7 +321,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
has_graphics |=
(family.queueFlags & vk::QueueFlagBits::eGraphics) != static_cast<vk::QueueFlagBits>(0);
- has_present |= physical.getSurfaceSupportKHR(i, surface, dldi) != 0;
+ has_present |= physical.getSurfaceSupportKHR(i, surface, dld) != 0;
}
if (!has_graphics || !has_present) {
LOG_ERROR(Render_Vulkan, "Device lacks a graphics and present queue");
@@ -331,7 +329,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
}
// TODO(Rodrigo): Check if the device matches all requeriments.
- const auto properties{physical.getProperties(dldi)};
+ const auto properties{physical.getProperties(dld)};
const auto& limits{properties.limits};
constexpr u32 required_ubo_size = 65536;
@@ -348,7 +346,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
is_suitable = false;
}
- const auto features{physical.getFeatures(dldi)};
+ const auto features{physical.getFeatures(dld)};
const std::array feature_report = {
std::make_pair(features.vertexPipelineStoresAndAtomics, "vertexPipelineStoresAndAtomics"),
std::make_pair(features.independentBlend, "independentBlend"),
@@ -380,7 +378,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
return is_suitable;
}
-std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynamic& dldi) {
+std::vector<const char*> VKDevice::LoadExtensions() {
std::vector<const char*> extensions;
const auto Test = [&](const vk::ExtensionProperties& extension,
std::optional<std::reference_wrapper<bool>> status, const char* name,
@@ -411,7 +409,7 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
bool has_khr_shader_float16_int8{};
bool has_ext_subgroup_size_control{};
bool has_ext_transform_feedback{};
- for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
+ for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dld)) {
Test(extension, khr_uniform_buffer_standard_layout,
VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true);
Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME,
@@ -433,15 +431,15 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
if (has_khr_shader_float16_int8) {
is_float16_supported =
- GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16;
+ GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dld).shaderFloat16;
extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME);
}
if (has_ext_subgroup_size_control) {
const auto features =
- GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dldi);
+ GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dld);
const auto properties =
- GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dldi);
+ GetProperties<vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT>(physical, dld);
is_warp_potentially_bigger = properties.maxSubgroupSize > GuestWarpSize;
@@ -456,9 +454,9 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
if (has_ext_transform_feedback) {
const auto features =
- GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dldi);
+ GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dld);
const auto properties =
- GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dldi);
+ GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dld);
if (features.transformFeedback && features.geometryStreams &&
properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers &&
@@ -471,10 +469,10 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
return extensions;
}
-void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface) {
+void VKDevice::SetupFamilies(vk::SurfaceKHR surface) {
std::optional<u32> graphics_family_, present_family_;
- const auto queue_family_properties = physical.getQueueFamilyProperties(dldi);
+ const auto queue_family_properties = physical.getQueueFamilyProperties(dld);
for (u32 i = 0; i < static_cast<u32>(queue_family_properties.size()); ++i) {
if (graphics_family_ && present_family_)
break;
@@ -483,10 +481,12 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
if (queue_family.queueCount == 0)
continue;
- if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics)
+ if (queue_family.queueFlags & vk::QueueFlagBits::eGraphics) {
graphics_family_ = i;
- if (physical.getSurfaceSupportKHR(i, surface, dldi))
+ }
+ if (physical.getSurfaceSupportKHR(i, surface, dld)) {
present_family_ = i;
+ }
}
ASSERT(graphics_family_ && present_family_);
@@ -494,10 +494,10 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
present_family = *present_family_;
}
-void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
- const auto supported_features{physical.getFeatures(dldi)};
+void VKDevice::SetupFeatures() {
+ const auto supported_features{physical.getFeatures(dld)};
is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat;
- is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
+ is_optimal_astc_supported = IsOptimalAstcSupported(supported_features);
}
void VKDevice::CollectTelemetryParameters() {
@@ -525,7 +525,7 @@ std::vector<vk::DeviceQueueCreateInfo> VKDevice::GetDeviceQueueCreateInfos() con
}
std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperties(
- const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical) {
+ const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical) {
static constexpr std::array formats{vk::Format::eA8B8G8R8UnormPack32,
vk::Format::eA8B8G8R8UintPack32,
vk::Format::eA8B8G8R8SnormPack32,
@@ -606,7 +606,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
vk::Format::eE5B9G9R9UfloatPack32};
std::unordered_map<vk::Format, vk::FormatProperties> format_properties;
for (const auto format : formats) {
- format_properties.emplace(format, physical.getFormatProperties(format, dldi));
+ format_properties.emplace(format, physical.getFormatProperties(format, dld));
}
return format_properties;
}
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index 6e656517f..d9d809852 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -22,12 +22,12 @@ const u32 GuestWarpSize = 32;
/// Handles data specific to a physical device.
class VKDevice final {
public:
- explicit VKDevice(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
+ explicit VKDevice(const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical,
vk::SurfaceKHR surface);
~VKDevice();
/// Initializes the device. Returns true on success.
- bool Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instance);
+ bool Create(vk::Instance instance);
/**
* Returns a format supported by the device for the passed requeriments.
@@ -188,18 +188,18 @@ public:
}
/// Checks if the physical device is suitable.
- static bool IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical,
- vk::SurfaceKHR surface);
+ static bool IsSuitable(vk::PhysicalDevice physical, vk::SurfaceKHR surface,
+ const vk::DispatchLoaderDynamic& dld);
private:
/// Loads extensions into a vector and stores available ones in this object.
- std::vector<const char*> LoadExtensions(const vk::DispatchLoaderDynamic& dldi);
+ std::vector<const char*> LoadExtensions();
/// Sets up queue families.
- void SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceKHR surface);
+ void SetupFamilies(vk::SurfaceKHR surface);
/// Sets up device features.
- void SetupFeatures(const vk::DispatchLoaderDynamic& dldi);
+ void SetupFeatures();
/// Collects telemetry information from the device.
void CollectTelemetryParameters();
@@ -208,8 +208,7 @@ private:
std::vector<vk::DeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
/// Returns true if ASTC textures are natively supported.
- bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features,
- const vk::DispatchLoaderDynamic& dldi) const;
+ bool IsOptimalAstcSupported(const vk::PhysicalDeviceFeatures& features) const;
/// Returns true if a format is supported.
bool IsFormatSupported(vk::Format wanted_format, vk::FormatFeatureFlags wanted_usage,
@@ -217,10 +216,10 @@ private:
/// Returns the device properties for Vulkan formats.
static std::unordered_map<vk::Format, vk::FormatProperties> GetFormatProperties(
- const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDevice physical);
+ const vk::DispatchLoaderDynamic& dld, vk::PhysicalDevice physical);
- const vk::PhysicalDevice physical; ///< Physical device.
vk::DispatchLoaderDynamic dld; ///< Device function pointers.
+ vk::PhysicalDevice physical; ///< Physical device.
vk::PhysicalDeviceProperties properties; ///< Device properties.
UniqueDevice logical; ///< Logical device.
vk::Queue graphics_queue; ///< Main graphics queue.
diff --git a/src/video_core/shader/decode/video.cpp b/src/video_core/shader/decode/video.cpp
index b047cf870..64ba60ea2 100644
--- a/src/video_core/shader/decode/video.cpp
+++ b/src/video_core/shader/decode/video.cpp
@@ -10,16 +10,24 @@
namespace VideoCommon::Shader {
+using std::move;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
using Tegra::Shader::Pred;
using Tegra::Shader::VideoType;
using Tegra::Shader::VmadShr;
+using Tegra::Shader::VmnmxOperation;
+using Tegra::Shader::VmnmxType;
u32 ShaderIR::DecodeVideo(NodeBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
const auto opcode = OpCode::Decode(instr);
+ if (opcode->get().GetId() == OpCode::Id::VMNMX) {
+ DecodeVMNMX(bb, instr);
+ return pc;
+ }
+
const Node op_a =
GetVideoOperand(GetRegister(instr.gpr8), instr.video.is_byte_chunk_a, instr.video.signed_a,
instr.video.type_a, instr.video.byte_height_a);
@@ -109,4 +117,54 @@ Node ShaderIR::GetVideoOperand(Node op, bool is_chunk, bool is_signed,
}
}
+void ShaderIR::DecodeVMNMX(NodeBlock& bb, Tegra::Shader::Instruction instr) {
+ UNIMPLEMENTED_IF(!instr.vmnmx.is_op_b_register);
+ UNIMPLEMENTED_IF(instr.vmnmx.SourceFormatA() != VmnmxType::Bits32);
+ UNIMPLEMENTED_IF(instr.vmnmx.SourceFormatB() != VmnmxType::Bits32);
+ UNIMPLEMENTED_IF(instr.vmnmx.is_src_a_signed != instr.vmnmx.is_src_b_signed);
+ UNIMPLEMENTED_IF(instr.vmnmx.sat);
+ UNIMPLEMENTED_IF(instr.generates_cc);
+
+ Node op_a = GetRegister(instr.gpr8);
+ Node op_b = GetRegister(instr.gpr20);
+ Node op_c = GetRegister(instr.gpr39);
+
+ const bool is_oper1_signed = instr.vmnmx.is_src_a_signed; // Stubbed
+ const bool is_oper2_signed = instr.vmnmx.is_dest_signed;
+
+ const auto operation_a = instr.vmnmx.mx ? OperationCode::IMax : OperationCode::IMin;
+ Node value = SignedOperation(operation_a, is_oper1_signed, move(op_a), move(op_b));
+
+ switch (instr.vmnmx.operation) {
+ case VmnmxOperation::Mrg_16H:
+ value = BitfieldInsert(move(op_c), move(value), 16, 16);
+ break;
+ case VmnmxOperation::Mrg_16L:
+ value = BitfieldInsert(move(op_c), move(value), 0, 16);
+ break;
+ case VmnmxOperation::Mrg_8B0:
+ value = BitfieldInsert(move(op_c), move(value), 0, 8);
+ break;
+ case VmnmxOperation::Mrg_8B2:
+ value = BitfieldInsert(move(op_c), move(value), 16, 8);
+ break;
+ case VmnmxOperation::Acc:
+ value = Operation(OperationCode::IAdd, move(value), move(op_c));
+ break;
+ case VmnmxOperation::Min:
+ value = SignedOperation(OperationCode::IMin, is_oper2_signed, move(value), move(op_c));
+ break;
+ case VmnmxOperation::Max:
+ value = SignedOperation(OperationCode::IMax, is_oper2_signed, move(value), move(op_c));
+ break;
+ case VmnmxOperation::Nop:
+ break;
+ default:
+ UNREACHABLE();
+ break;
+ }
+
+ SetRegister(bb, instr.gpr0, move(value));
+}
+
} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index ca6c976c9..c6e7bdf50 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -354,6 +354,9 @@ private:
/// Marks the usage of a input or output attribute.
void MarkAttributeUsage(Tegra::Shader::Attribute::Index index, u64 element);
+ /// Decodes VMNMX instruction and inserts its code into the passed basic block.
+ void DecodeVMNMX(NodeBlock& bb, Tegra::Shader::Instruction instr);
+
void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
const Node4& components);
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index 59b8a5e66..eba05aced 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -131,6 +131,20 @@ enum class SwizzleSource : u32 {
OneFloat = 7,
};
+enum class MsaaMode : u32 {
+ Msaa1x1 = 0,
+ Msaa2x1 = 1,
+ Msaa2x2 = 2,
+ Msaa4x2 = 3,
+ Msaa4x2_D3D = 4,
+ Msaa2x1_D3D = 5,
+ Msaa4x4 = 6,
+ Msaa2x2_VC4 = 8,
+ Msaa2x2_VC12 = 9,
+ Msaa4x2_VC8 = 10,
+ Msaa4x2_VC24 = 11,
+};
+
union TextureHandle {
TextureHandle(u32 raw) : raw{raw} {}
@@ -197,6 +211,7 @@ struct TICEntry {
union {
BitField<0, 4, u32> res_min_mip_level;
BitField<4, 4, u32> res_max_mip_level;
+ BitField<8, 4, MsaaMode> msaa_mode;
BitField<12, 12, u32> min_lod_clamp;
};
generated by cgit v1.2.3 (git 2.25.1) at 2024-08-13 22:14:19 +0200