citra/src/video_core/renderer_vulkan/vk_texture_runtime.cpp

1577 lines
64 KiB
C++
Raw Normal View History

Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
// Copyright 2023 Citra Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#include <boost/container/small_vector.hpp>
#include <boost/container/static_vector.hpp>
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
#include "common/literals.h"
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
#include "common/microprofile.h"
#include "common/scope_exit.h"
#include "video_core/custom_textures/material.h"
#include "video_core/rasterizer_cache/texture_codec.h"
#include "video_core/rasterizer_cache/utils.h"
#include "video_core/renderer_vulkan/pica_to_vk.h"
#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
#include "video_core/renderer_vulkan/vk_instance.h"
#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
#include "video_core/renderer_vulkan/vk_scheduler.h"
#include "video_core/renderer_vulkan/vk_texture_runtime.h"
#include <vk_mem_alloc.h>
#include <vulkan/vulkan_format_traits.hpp>
// Ignore the -Wclass-memaccess warning on memcpy for non-trivially default constructible objects.
#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wclass-memaccess"
#endif
MICROPROFILE_DEFINE(Vulkan_ImageAlloc, "Vulkan", "Texture Allocation", MP_RGB(192, 52, 235));
namespace Vulkan {
namespace {
using VideoCore::MapType;
using VideoCore::PixelFormat;
using VideoCore::SurfaceType;
using VideoCore::TextureType;
using namespace Common::Literals;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
struct RecordParams {
vk::ImageAspectFlags aspect;
vk::Filter filter;
vk::PipelineStageFlags pipeline_flags;
vk::AccessFlags src_access;
vk::AccessFlags dst_access;
vk::Image src_image;
vk::Image dst_image;
};
vk::Filter MakeFilter(VideoCore::PixelFormat pixel_format) {
switch (pixel_format) {
case VideoCore::PixelFormat::D16:
case VideoCore::PixelFormat::D24:
case VideoCore::PixelFormat::D24S8:
return vk::Filter::eNearest;
default:
return vk::Filter::eLinear;
}
}
[[nodiscard]] vk::ClearValue MakeClearValue(VideoCore::ClearValue clear) {
static_assert(sizeof(VideoCore::ClearValue) == sizeof(vk::ClearValue));
vk::ClearValue value{};
std::memcpy(&value, &clear, sizeof(vk::ClearValue));
return value;
}
[[nodiscard]] vk::ClearColorValue MakeClearColorValue(Common::Vec4f color) {
return vk::ClearColorValue{
.float32 = std::array{color[0], color[1], color[2], color[3]},
};
}
[[nodiscard]] vk::ClearDepthStencilValue MakeClearDepthStencilValue(VideoCore::ClearValue clear) {
return vk::ClearDepthStencilValue{
.depth = clear.depth,
.stencil = clear.stencil,
};
}
u32 UnpackDepthStencil(const VideoCore::StagingData& data, vk::Format dest) {
u32 depth_offset = 0;
u32 stencil_offset = 4 * data.size / 5;
const auto& mapped = data.mapped;
switch (dest) {
case vk::Format::eD24UnormS8Uint: {
for (; stencil_offset < data.size; depth_offset += 4) {
u8* ptr = mapped.data() + depth_offset;
const u32 d24s8 = VideoCore::MakeInt<u32>(ptr);
const u32 d24 = d24s8 >> 8;
mapped[stencil_offset] = d24s8 & 0xFF;
std::memcpy(ptr, &d24, 4);
stencil_offset++;
}
break;
}
case vk::Format::eD32SfloatS8Uint: {
for (; stencil_offset < data.size; depth_offset += 4) {
u8* ptr = mapped.data() + depth_offset;
const u32 d24s8 = VideoCore::MakeInt<u32>(ptr);
const float d32 = (d24s8 >> 8) / 16777215.f;
mapped[stencil_offset] = d24s8 & 0xFF;
std::memcpy(ptr, &d32, 4);
stencil_offset++;
}
break;
}
default:
LOG_ERROR(Render_Vulkan, "Unimplemented convertion for depth format {}",
vk::to_string(dest));
UNREACHABLE();
}
ASSERT(depth_offset == 4 * data.size / 5);
return depth_offset;
}
boost::container::small_vector<vk::ImageMemoryBarrier, 3> MakeInitBarriers(
vk::ImageAspectFlags aspect, std::span<const vk::Image> images) {
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
boost::container::small_vector<vk::ImageMemoryBarrier, 3> barriers;
for (const vk::Image& image : images) {
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
barriers.push_back(vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eNone,
.dstAccessMask = vk::AccessFlagBits::eNone,
.oldLayout = vk::ImageLayout::eUndefined,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = image,
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
.subresourceRange{
.aspectMask = aspect,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
});
}
return barriers;
}
Handle MakeHandle(const Instance* instance, u32 width, u32 height, u32 levels, TextureType type,
vk::Format format, vk::ImageUsageFlags usage, vk::ImageCreateFlags flags,
vk::ImageAspectFlags aspect, bool need_format_list,
std::string_view debug_name = {}) {
const u32 layers = type == TextureType::CubeMap ? 6 : 1;
const std::array format_list = {
vk::Format::eR8G8B8A8Unorm,
vk::Format::eR32Uint,
};
const vk::ImageFormatListCreateInfo image_format_list = {
.viewFormatCount = static_cast<u32>(format_list.size()),
.pViewFormats = format_list.data(),
};
const vk::ImageCreateInfo image_info = {
.pNext = need_format_list ? &image_format_list : nullptr,
.flags = flags,
.imageType = vk::ImageType::e2D,
.format = format,
.extent = {width, height, 1},
.mipLevels = levels,
.arrayLayers = layers,
.samples = vk::SampleCountFlagBits::e1,
.usage = usage,
};
const VmaAllocationCreateInfo alloc_info = {
.flags = VMA_ALLOCATION_CREATE_WITHIN_BUDGET_BIT,
.usage = VMA_MEMORY_USAGE_AUTO_PREFER_DEVICE,
.requiredFlags = 0,
.preferredFlags = 0,
.pool = VK_NULL_HANDLE,
.pUserData = nullptr,
};
VkImage unsafe_image{};
VkImageCreateInfo unsafe_image_info = static_cast<VkImageCreateInfo>(image_info);
VmaAllocation allocation{};
VkResult result = vmaCreateImage(instance->GetAllocator(), &unsafe_image_info, &alloc_info,
&unsafe_image, &allocation, nullptr);
if (result != VK_SUCCESS) [[unlikely]] {
LOG_CRITICAL(Render_Vulkan, "Failed allocating image with error {}", result);
UNREACHABLE();
}
const vk::Image image{unsafe_image};
const vk::ImageViewCreateInfo view_info = {
.image = image,
.viewType =
type == TextureType::CubeMap ? vk::ImageViewType::eCube : vk::ImageViewType::e2D,
.format = format,
.subresourceRange{
.aspectMask = aspect,
.baseMipLevel = 0,
.levelCount = levels,
.baseArrayLayer = 0,
.layerCount = layers,
},
};
vk::UniqueImageView image_view = instance->GetDevice().createImageViewUnique(view_info);
if (!debug_name.empty() && instance->HasDebuggingToolAttached()) {
Vulkan::SetObjectName(instance->GetDevice(), image, debug_name);
Vulkan::SetObjectName(instance->GetDevice(), image_view.get(), "{} View({})", debug_name,
vk::to_string(aspect));
}
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
return Handle{
.alloc = allocation,
.image = image,
.image_view = std::move(image_view),
};
}
vk::UniqueFramebuffer MakeFramebuffer(vk::Device device, vk::RenderPass render_pass, u32 width,
u32 height, std::span<const vk::ImageView> attachments) {
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
const vk::FramebufferCreateInfo framebuffer_info = {
.renderPass = render_pass,
.attachmentCount = static_cast<u32>(attachments.size()),
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
.pAttachments = attachments.data(),
.width = width,
.height = height,
.layers = 1,
};
return device.createFramebufferUnique(framebuffer_info);
}
vk::ImageSubresourceRange MakeSubresourceRange(vk::ImageAspectFlags aspect, u32 level = 0,
u32 levels = 1, u32 layer = 0) {
return vk::ImageSubresourceRange{
.aspectMask = aspect,
.baseMipLevel = level,
.levelCount = levels,
.baseArrayLayer = layer,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
};
}
constexpr u64 UPLOAD_BUFFER_SIZE = 512_MiB;
constexpr u64 DOWNLOAD_BUFFER_SIZE = 16_MiB;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
} // Anonymous namespace
TextureRuntime::TextureRuntime(const Instance& instance, Scheduler& scheduler,
RenderpassCache& renderpass_cache, DescriptorPool& pool,
DescriptorSetProvider& texture_provider_, u32 num_swapchain_images_)
: instance{instance}, scheduler{scheduler}, renderpass_cache{renderpass_cache},
texture_provider{texture_provider_}, blit_helper{instance, scheduler, pool, renderpass_cache},
upload_buffer{instance, scheduler, vk::BufferUsageFlagBits::eTransferSrc, UPLOAD_BUFFER_SIZE,
BufferType::Upload},
download_buffer{instance, scheduler,
vk::BufferUsageFlagBits::eTransferDst |
vk::BufferUsageFlagBits::eStorageBuffer,
DOWNLOAD_BUFFER_SIZE, BufferType::Download},
num_swapchain_images{num_swapchain_images_} {}
TextureRuntime::~TextureRuntime() = default;
VideoCore::StagingData TextureRuntime::FindStaging(u32 size, bool upload) {
StreamBuffer& buffer = upload ? upload_buffer : download_buffer;
const auto [data, offset, invalidate] = buffer.Map(size, 16);
return VideoCore::StagingData{
.size = size,
.offset = static_cast<u32>(offset),
.mapped = std::span{data, size},
};
}
u32 TextureRuntime::RemoveThreshold() {
return num_swapchain_images;
}
void TextureRuntime::Finish() {
scheduler.Finish();
}
bool TextureRuntime::Reinterpret(Surface& source, Surface& dest,
const VideoCore::TextureCopy& copy) {
const PixelFormat src_format = source.pixel_format;
const PixelFormat dst_format = dest.pixel_format;
ASSERT_MSG(src_format != dst_format, "Reinterpretation with the same format is invalid");
if (!source.traits.needs_conversion && !dest.traits.needs_conversion &&
source.type == dest.type) {
CopyTextures(source, dest, copy);
return true;
}
if (src_format == PixelFormat::D24S8 && dst_format == PixelFormat::RGBA8) {
blit_helper.ConvertDS24S8ToRGBA8(source, dest, copy);
} else {
LOG_WARNING(Render_Vulkan, "Unimplemented reinterpretation {} -> {}",
VideoCore::PixelFormatAsString(src_format),
VideoCore::PixelFormatAsString(dst_format));
return false;
}
return true;
}
bool TextureRuntime::ClearTexture(Surface& surface, const VideoCore::TextureClear& clear) {
renderpass_cache.EndRendering();
const RecordParams params = {
.aspect = surface.Aspect(),
.pipeline_flags = surface.PipelineStageFlags(),
.src_access = surface.AccessFlags(),
.src_image = surface.Image(),
};
if (clear.texture_rect == surface.GetScaledRect()) {
scheduler.Record([params, clear](vk::CommandBuffer cmdbuf) {
const vk::ImageSubresourceRange range = {
.aspectMask = params.aspect,
.baseMipLevel = clear.texture_level,
.levelCount = 1,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
};
const vk::ImageMemoryBarrier pre_barrier = {
.srcAccessMask = params.src_access,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = range,
};
const vk::ImageMemoryBarrier post_barrier = {
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = params.src_access,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = range,
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, pre_barrier);
const bool is_color =
static_cast<bool>(params.aspect & vk::ImageAspectFlagBits::eColor);
if (is_color) {
cmdbuf.clearColorImage(params.src_image, vk::ImageLayout::eTransferDstOptimal,
MakeClearColorValue(clear.value.color), range);
} else {
cmdbuf.clearDepthStencilImage(params.src_image,
vk::ImageLayout::eTransferDstOptimal,
MakeClearDepthStencilValue(clear.value), range);
}
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, post_barrier);
});
return true;
}
ClearTextureWithRenderpass(surface, clear);
return true;
}
void TextureRuntime::ClearTextureWithRenderpass(Surface& surface,
const VideoCore::TextureClear& clear) {
const bool is_color = surface.type != VideoCore::SurfaceType::Depth &&
surface.type != VideoCore::SurfaceType::DepthStencil;
const auto color_format = is_color ? surface.pixel_format : PixelFormat::Invalid;
const auto depth_format = is_color ? PixelFormat::Invalid : surface.pixel_format;
const auto render_pass = renderpass_cache.GetRenderpass(color_format, depth_format, true);
const RecordParams params = {
.aspect = surface.Aspect(),
.pipeline_flags = surface.PipelineStageFlags(),
.src_access = surface.AccessFlags(),
.src_image = surface.Image(),
};
scheduler.Record([params, is_color, clear, render_pass,
framebuffer = surface.Framebuffer()](vk::CommandBuffer cmdbuf) {
const vk::AccessFlags access_flag =
is_color ? vk::AccessFlagBits::eColorAttachmentRead |
vk::AccessFlagBits::eColorAttachmentWrite
: vk::AccessFlagBits::eDepthStencilAttachmentRead |
vk::AccessFlagBits::eDepthStencilAttachmentWrite;
const vk::PipelineStageFlags pipeline_flags =
is_color ? vk::PipelineStageFlagBits::eColorAttachmentOutput
: vk::PipelineStageFlagBits::eEarlyFragmentTests;
const vk::ImageMemoryBarrier pre_barrier = {
.srcAccessMask = params.src_access,
.dstAccessMask = access_flag,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, clear.texture_level),
};
const vk::ImageMemoryBarrier post_barrier = {
.srcAccessMask = access_flag,
.dstAccessMask = params.src_access,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, clear.texture_level),
};
const vk::Rect2D render_area = {
.offset{
.x = static_cast<s32>(clear.texture_rect.left),
.y = static_cast<s32>(clear.texture_rect.bottom),
},
.extent{
.width = clear.texture_rect.GetWidth(),
.height = clear.texture_rect.GetHeight(),
},
};
const auto clear_value = MakeClearValue(clear.value);
const vk::RenderPassBeginInfo renderpass_begin_info = {
.renderPass = render_pass,
.framebuffer = framebuffer,
.renderArea = render_area,
.clearValueCount = 1,
.pClearValues = &clear_value,
};
cmdbuf.pipelineBarrier(params.pipeline_flags, pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, pre_barrier);
cmdbuf.beginRenderPass(renderpass_begin_info, vk::SubpassContents::eInline);
cmdbuf.endRenderPass();
cmdbuf.pipelineBarrier(pipeline_flags, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, post_barrier);
});
}
bool TextureRuntime::CopyTextures(Surface& source, Surface& dest,
const VideoCore::TextureCopy& copy) {
renderpass_cache.EndRendering();
const RecordParams params = {
.aspect = source.Aspect(),
.filter = MakeFilter(source.pixel_format),
.pipeline_flags = source.PipelineStageFlags() | dest.PipelineStageFlags(),
.src_access = source.AccessFlags(),
.dst_access = dest.AccessFlags(),
.src_image = source.Image(),
.dst_image = dest.Image(),
};
scheduler.Record([params, copy](vk::CommandBuffer cmdbuf) {
const vk::ImageCopy image_copy = {
.srcSubresource{
.aspectMask = params.aspect,
.mipLevel = copy.src_level,
.baseArrayLayer = copy.src_layer,
.layerCount = 1,
},
.srcOffset = {static_cast<s32>(copy.src_offset.x), static_cast<s32>(copy.src_offset.y),
0},
.dstSubresource{
.aspectMask = params.aspect,
.mipLevel = copy.dst_level,
.baseArrayLayer = copy.dst_layer,
.layerCount = 1,
},
.dstOffset = {static_cast<s32>(copy.dst_offset.x), static_cast<s32>(copy.dst_offset.y),
0},
.extent = {copy.extent.width, copy.extent.height, 1},
};
const bool self_copy = params.src_image == params.dst_image;
const vk::ImageLayout new_src_layout =
self_copy ? vk::ImageLayout::eGeneral : vk::ImageLayout::eTransferSrcOptimal;
const vk::ImageLayout new_dst_layout =
self_copy ? vk::ImageLayout::eGeneral : vk::ImageLayout::eTransferDstOptimal;
const std::array pre_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = params.src_access,
.dstAccessMask = vk::AccessFlagBits::eTransferRead,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = new_src_layout,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, copy.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = params.dst_access,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = new_dst_layout,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, copy.dst_level),
},
};
const std::array post_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eNone,
.dstAccessMask = vk::AccessFlagBits::eNone,
.oldLayout = new_src_layout,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, copy.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = params.dst_access,
.oldLayout = new_dst_layout,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, copy.dst_level),
},
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, pre_barriers);
cmdbuf.copyImage(params.src_image, new_src_layout, params.dst_image, new_dst_layout,
image_copy);
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, post_barriers);
});
return true;
}
bool TextureRuntime::BlitTextures(Surface& source, Surface& dest,
const VideoCore::TextureBlit& blit) {
const bool is_depth_stencil = source.type == VideoCore::SurfaceType::DepthStencil;
const auto& depth_traits = instance.GetTraits(source.pixel_format);
if (is_depth_stencil && !depth_traits.blit_support) {
return blit_helper.BlitDepthStencil(source, dest, blit);
}
renderpass_cache.EndRendering();
const RecordParams params = {
.aspect = source.Aspect(),
.filter = MakeFilter(source.pixel_format),
.pipeline_flags = source.PipelineStageFlags() | dest.PipelineStageFlags(),
.src_access = source.AccessFlags(),
.dst_access = dest.AccessFlags(),
.src_image = source.Image(),
.dst_image = dest.Image(),
};
scheduler.Record([params, blit](vk::CommandBuffer cmdbuf) {
const std::array source_offsets = {
vk::Offset3D{static_cast<s32>(blit.src_rect.left),
static_cast<s32>(blit.src_rect.bottom), 0},
vk::Offset3D{static_cast<s32>(blit.src_rect.right), static_cast<s32>(blit.src_rect.top),
1},
};
const std::array dest_offsets = {
vk::Offset3D{static_cast<s32>(blit.dst_rect.left),
static_cast<s32>(blit.dst_rect.bottom), 0},
vk::Offset3D{static_cast<s32>(blit.dst_rect.right), static_cast<s32>(blit.dst_rect.top),
1},
};
const vk::ImageBlit blit_area = {
.srcSubresource{
.aspectMask = params.aspect,
.mipLevel = blit.src_level,
.baseArrayLayer = blit.src_layer,
.layerCount = 1,
},
.srcOffsets = source_offsets,
.dstSubresource{
.aspectMask = params.aspect,
.mipLevel = blit.dst_level,
.baseArrayLayer = blit.dst_layer,
.layerCount = 1,
},
.dstOffsets = dest_offsets,
};
const std::array read_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = params.src_access,
.dstAccessMask = vk::AccessFlagBits::eTransferRead,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferSrcOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, blit.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = params.dst_access,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, blit.dst_level),
},
};
const std::array write_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferRead,
.dstAccessMask = params.src_access,
.oldLayout = vk::ImageLayout::eTransferSrcOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, blit.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = params.dst_access,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, blit.dst_level),
},
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, read_barriers);
cmdbuf.blitImage(params.src_image, vk::ImageLayout::eTransferSrcOptimal, params.dst_image,
vk::ImageLayout::eTransferDstOptimal, blit_area, params.filter);
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, write_barriers);
});
return true;
}
void TextureRuntime::GenerateMipmaps(Surface& surface) {
if (VideoCore::IsCustomFormatCompressed(surface.custom_format)) {
LOG_ERROR(Render_Vulkan, "Generating mipmaps for compressed formats unsupported!");
return;
}
renderpass_cache.EndRendering();
auto [width, height] = surface.RealExtent();
const u32 levels = surface.levels;
for (u32 i = 1; i < levels; i++) {
const Common::Rectangle<u32> src_rect{0, height, width, 0};
width = width > 1 ? width >> 1 : 1;
height = height > 1 ? height >> 1 : 1;
const Common::Rectangle<u32> dst_rect{0, height, width, 0};
const VideoCore::TextureBlit blit = {
.src_level = i - 1,
.dst_level = i,
.src_rect = src_rect,
.dst_rect = dst_rect,
};
BlitTextures(surface, surface, blit);
}
}
bool TextureRuntime::NeedsConversion(VideoCore::PixelFormat format) const {
const FormatTraits traits = instance.GetTraits(format);
return traits.needs_conversion &&
// DepthStencil formats are handled elsewhere due to de-interleaving.
traits.aspect != (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil);
}
void TextureRuntime::FreeDescriptorSetsWithImage(vk::ImageView image_view) {
texture_provider.FreeWithImage(image_view);
blit_helper.compute_provider.FreeWithImage(image_view);
blit_helper.compute_buffer_provider.FreeWithImage(image_view);
blit_helper.two_textures_provider.FreeWithImage(image_view);
}
Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceParams& params)
: SurfaceBase{params}, runtime{&runtime_}, instance{&runtime_.GetInstance()},
scheduler{&runtime_.GetScheduler()}, traits{instance->GetTraits(pixel_format)} {
if (pixel_format == VideoCore::PixelFormat::Invalid) {
return;
}
const bool is_mutable = pixel_format == VideoCore::PixelFormat::RGBA8;
const vk::Format format = traits.native;
ASSERT_MSG(format != vk::Format::eUndefined && levels >= 1,
"Image allocation parameters are invalid");
boost::container::static_vector<vk::Image, 3> raw_images;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
vk::ImageCreateFlags flags{};
if (texture_type == VideoCore::TextureType::CubeMap) {
flags |= vk::ImageCreateFlagBits::eCubeCompatible;
}
if (is_mutable) {
flags |= vk::ImageCreateFlagBits::eMutableFormat;
}
const bool need_format_list = is_mutable && instance->IsImageFormatListSupported();
handles[0] = MakeHandle(instance, width, height, levels, texture_type, format, traits.usage,
flags, traits.aspect, need_format_list, DebugName(false));
raw_images.emplace_back(handles[0].image);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
if (res_scale != 1) {
handles[1] =
MakeHandle(instance, GetScaledWidth(), GetScaledHeight(), levels, texture_type, format,
traits.usage, flags, traits.aspect, need_format_list, DebugName(true));
raw_images.emplace_back(handles[1].image);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
runtime->renderpass_cache.EndRendering();
scheduler->Record([raw_images, aspect = traits.aspect](vk::CommandBuffer cmdbuf) {
const auto barriers = MakeInitBarriers(aspect, raw_images);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTopOfPipe,
vk::PipelineStageFlagBits::eTopOfPipe,
vk::DependencyFlagBits::eByRegion, {}, {}, barriers);
});
}
Surface::Surface(TextureRuntime& runtime_, const VideoCore::SurfaceBase& surface,
const VideoCore::Material* mat)
: SurfaceBase{surface}, runtime{&runtime_}, instance{&runtime_.GetInstance()},
scheduler{&runtime_.GetScheduler()}, traits{instance->GetTraits(mat->format)} {
if (!traits.transfer_support) {
return;
}
const bool has_normal = mat && mat->Map(MapType::Normal);
const vk::Format format = traits.native;
boost::container::static_vector<vk::Image, 2> raw_images;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
vk::ImageCreateFlags flags{};
if (texture_type == VideoCore::TextureType::CubeMap) {
flags |= vk::ImageCreateFlagBits::eCubeCompatible;
}
const std::string debug_name = DebugName(false, true);
handles[0] = MakeHandle(instance, mat->width, mat->height, levels, texture_type, format,
traits.usage, flags, traits.aspect, false, debug_name);
raw_images.emplace_back(handles[0].image);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
if (res_scale != 1) {
handles[1] = MakeHandle(instance, mat->width, mat->height, levels, texture_type,
vk::Format::eR8G8B8A8Unorm, traits.usage, flags, traits.aspect,
false, debug_name);
raw_images.emplace_back(handles[1].image);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
if (has_normal) {
handles[2] = MakeHandle(instance, mat->width, mat->height, levels, texture_type, format,
traits.usage, flags, traits.aspect, false, debug_name);
raw_images.emplace_back(handles[2].image);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
runtime->renderpass_cache.EndRendering();
scheduler->Record([raw_images, aspect = traits.aspect](vk::CommandBuffer cmdbuf) {
const auto barriers = MakeInitBarriers(aspect, raw_images);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTopOfPipe,
vk::PipelineStageFlagBits::eTopOfPipe,
vk::DependencyFlagBits::eByRegion, {}, {}, barriers);
});
custom_format = mat->format;
material = mat;
}
Surface::~Surface() {
if (!handles[0].image_view) {
return;
}
for (const auto& [alloc, image, image_view] : handles) {
if (image_view) {
runtime->FreeDescriptorSetsWithImage(*image_view);
}
if (image) {
vmaDestroyImage(instance->GetAllocator(), image, alloc);
}
}
if (copy_handle.image_view) {
vmaDestroyImage(instance->GetAllocator(), copy_handle.image, copy_handle.alloc);
}
}
void Surface::Upload(const VideoCore::BufferTextureCopy& upload,
const VideoCore::StagingData& staging) {
runtime->renderpass_cache.EndRendering();
const RecordParams params = {
.aspect = Aspect(),
.pipeline_flags = PipelineStageFlags(),
.src_access = AccessFlags(),
.src_image = Image(0),
};
scheduler->Record([buffer = runtime->upload_buffer.Handle(), format = traits.native, params,
staging, upload](vk::CommandBuffer cmdbuf) {
boost::container::static_vector<vk::BufferImageCopy, 2> buffer_image_copies;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
const auto rect = upload.texture_rect;
buffer_image_copies.emplace_back(vk::BufferImageCopy{
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
.bufferOffset = upload.buffer_offset,
.bufferRowLength = rect.GetWidth(),
.bufferImageHeight = rect.GetHeight(),
.imageSubresource{
.aspectMask = params.aspect,
.mipLevel = upload.texture_level,
.baseArrayLayer = 0,
.layerCount = 1,
},
.imageOffset = {static_cast<s32>(rect.left), static_cast<s32>(rect.bottom), 0},
.imageExtent = {rect.GetWidth(), rect.GetHeight(), 1},
});
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
if (params.aspect & vk::ImageAspectFlagBits::eStencil) {
buffer_image_copies[0].imageSubresource.aspectMask = vk::ImageAspectFlagBits::eDepth;
vk::BufferImageCopy& stencil_copy =
buffer_image_copies.emplace_back(buffer_image_copies[0]);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
stencil_copy = buffer_image_copies[0];
stencil_copy.bufferOffset += UnpackDepthStencil(staging, format);
stencil_copy.imageSubresource.aspectMask = vk::ImageAspectFlagBits::eStencil;
}
const vk::ImageMemoryBarrier read_barrier = {
.srcAccessMask = params.src_access,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, upload.texture_level),
};
const vk::ImageMemoryBarrier write_barrier = {
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = params.src_access,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, upload.texture_level),
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, read_barrier);
cmdbuf.copyBufferToImage(buffer, params.src_image, vk::ImageLayout::eTransferDstOptimal,
buffer_image_copies);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, write_barrier);
});
runtime->upload_buffer.Commit(staging.size);
if (res_scale != 1) {
const VideoCore::TextureBlit blit = {
.src_level = upload.texture_level,
.dst_level = upload.texture_level,
.src_rect = upload.texture_rect,
.dst_rect = upload.texture_rect * res_scale,
};
BlitScale(blit, true);
}
}
void Surface::UploadCustom(const VideoCore::Material* material, u32 level) {
const u32 width = material->width;
const u32 height = material->height;
const auto color = material->textures[0];
const Common::Rectangle rect{0U, height, width, 0U};
const auto upload = [&](u32 index, VideoCore::CustomTexture* texture) {
const u64 custom_size = texture->data.size();
const RecordParams params = {
.aspect = vk::ImageAspectFlagBits::eColor,
.pipeline_flags = PipelineStageFlags(),
.src_access = AccessFlags(),
.src_image = Image(index),
};
const auto [data, offset, invalidate] = runtime->upload_buffer.Map(custom_size, 0);
std::memcpy(data, texture->data.data(), custom_size);
runtime->upload_buffer.Commit(custom_size);
scheduler->Record([buffer = runtime->upload_buffer.Handle(), level, params, rect,
offset = offset](vk::CommandBuffer cmdbuf) {
const vk::BufferImageCopy buffer_image_copy = {
.bufferOffset = offset,
.bufferRowLength = 0,
.bufferImageHeight = rect.GetHeight(),
.imageSubresource{
.aspectMask = params.aspect,
.mipLevel = level,
.baseArrayLayer = 0,
.layerCount = 1,
},
.imageOffset = {static_cast<s32>(rect.left), static_cast<s32>(rect.bottom), 0},
.imageExtent = {rect.GetWidth(), rect.GetHeight(), 1},
};
const vk::ImageMemoryBarrier read_barrier = {
.srcAccessMask = params.src_access,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, level),
};
const vk::ImageMemoryBarrier write_barrier = {
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = params.src_access,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, level),
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, read_barrier);
cmdbuf.copyBufferToImage(buffer, params.src_image, vk::ImageLayout::eTransferDstOptimal,
buffer_image_copy);
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, write_barrier);
});
};
upload(0, color);
for (u32 i = 1; i < VideoCore::MAX_MAPS; i++) {
const auto texture = material->textures[i];
if (!texture) {
continue;
}
upload(i + 1, texture);
}
}
void Surface::Download(const VideoCore::BufferTextureCopy& download,
const VideoCore::StagingData& staging) {
SCOPE_EXIT({
scheduler->Finish();
runtime->download_buffer.Commit(staging.size);
});
runtime->renderpass_cache.EndRendering();
if (pixel_format == PixelFormat::D24S8) {
runtime->blit_helper.DepthToBuffer(*this, runtime->download_buffer.Handle(), download);
return;
}
if (res_scale != 1) {
const VideoCore::TextureBlit blit = {
.src_level = download.texture_level,
.dst_level = download.texture_level,
.src_rect = download.texture_rect * res_scale,
.dst_rect = download.texture_rect,
};
BlitScale(blit, false);
}
const RecordParams params = {
.aspect = Aspect(),
.pipeline_flags = PipelineStageFlags(),
.src_access = AccessFlags(),
.src_image = Image(0),
};
scheduler->Record(
[buffer = runtime->download_buffer.Handle(), params, download](vk::CommandBuffer cmdbuf) {
const auto rect = download.texture_rect;
const vk::BufferImageCopy buffer_image_copy = {
.bufferOffset = download.buffer_offset,
.bufferRowLength = rect.GetWidth(),
.bufferImageHeight = rect.GetHeight(),
.imageSubresource{
.aspectMask = params.aspect,
.mipLevel = download.texture_level,
.baseArrayLayer = 0,
.layerCount = 1,
},
.imageOffset = {static_cast<s32>(rect.left), static_cast<s32>(rect.bottom), 0},
.imageExtent = {rect.GetWidth(), rect.GetHeight(), 1},
};
const vk::ImageMemoryBarrier read_barrier = {
.srcAccessMask = vk::AccessFlagBits::eMemoryWrite,
.dstAccessMask = vk::AccessFlagBits::eTransferRead,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferSrcOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, download.texture_level),
};
const vk::ImageMemoryBarrier image_write_barrier = {
.srcAccessMask = vk::AccessFlagBits::eNone,
.dstAccessMask = vk::AccessFlagBits::eMemoryWrite,
.oldLayout = vk::ImageLayout::eTransferSrcOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, download.texture_level),
};
const vk::MemoryBarrier memory_write_barrier = {
.srcAccessMask = vk::AccessFlagBits::eMemoryWrite,
.dstAccessMask = vk::AccessFlagBits::eMemoryRead | vk::AccessFlagBits::eMemoryWrite,
};
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, read_barrier);
cmdbuf.copyImageToBuffer(params.src_image, vk::ImageLayout::eTransferSrcOptimal, buffer,
buffer_image_copy);
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, memory_write_barrier, {},
image_write_barrier);
});
}
void Surface::ScaleUp(u32 new_scale) {
if (res_scale == new_scale || new_scale == 1) {
return;
}
res_scale = new_scale;
const bool is_mutable = pixel_format == VideoCore::PixelFormat::RGBA8;
vk::ImageCreateFlags flags{};
if (texture_type == VideoCore::TextureType::CubeMap) {
flags |= vk::ImageCreateFlagBits::eCubeCompatible;
}
if (is_mutable) {
flags |= vk::ImageCreateFlagBits::eMutableFormat;
}
handles[1] =
MakeHandle(instance, GetScaledWidth(), GetScaledHeight(), levels, texture_type,
traits.native, traits.usage, flags, traits.aspect, false, DebugName(true));
runtime->renderpass_cache.EndRendering();
scheduler->Record(
[raw_images = std::array{Image()}, aspect = traits.aspect](vk::CommandBuffer cmdbuf) {
const auto barriers = MakeInitBarriers(aspect, raw_images);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTopOfPipe,
vk::PipelineStageFlagBits::eTopOfPipe,
vk::DependencyFlagBits::eByRegion, {}, {}, barriers);
});
LOG_INFO(HW_GPU, "Surface scale up!");
for (u32 level = 0; level < levels; level++) {
const VideoCore::TextureBlit blit = {
.src_level = level,
.dst_level = level,
.src_rect = GetRect(level),
.dst_rect = GetScaledRect(level),
};
BlitScale(blit, true);
}
}
u32 Surface::GetInternalBytesPerPixel() const {
// Request 5 bytes for D24S8 as well because we can use the
// extra space when deinterleaving the data during upload
if (traits.native == vk::Format::eD24UnormS8Uint) {
return 5;
}
return vk::blockSize(traits.native);
}
vk::AccessFlags Surface::AccessFlags() const noexcept {
const bool is_color = static_cast<bool>(Aspect() & vk::ImageAspectFlagBits::eColor);
const vk::AccessFlags attachment_flags =
is_color
? vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite
: vk::AccessFlagBits::eDepthStencilAttachmentRead |
vk::AccessFlagBits::eDepthStencilAttachmentWrite;
return vk::AccessFlagBits::eShaderRead | vk::AccessFlagBits::eTransferRead |
vk::AccessFlagBits::eTransferWrite |
(is_framebuffer ? attachment_flags : vk::AccessFlagBits::eNone) |
(is_storage ? vk::AccessFlagBits::eShaderWrite : vk::AccessFlagBits::eNone);
}
vk::PipelineStageFlags Surface::PipelineStageFlags() const noexcept {
const bool is_color = static_cast<bool>(Aspect() & vk::ImageAspectFlagBits::eColor);
const vk::PipelineStageFlags attachment_flags =
is_color ? vk::PipelineStageFlagBits::eColorAttachmentOutput
: vk::PipelineStageFlagBits::eEarlyFragmentTests |
vk::PipelineStageFlagBits::eLateFragmentTests;
return vk::PipelineStageFlagBits::eTransfer | vk::PipelineStageFlagBits::eFragmentShader |
(is_framebuffer ? attachment_flags : vk::PipelineStageFlagBits::eNone) |
(is_storage ? vk::PipelineStageFlagBits::eComputeShader
: vk::PipelineStageFlagBits::eNone);
}
vk::Image Surface::Image(u32 index) const noexcept {
const vk::Image image = handles[index].image;
if (!image) {
return handles[0].image;
}
return image;
}
vk::ImageView Surface::CopyImageView() noexcept {
vk::ImageLayout copy_layout = vk::ImageLayout::eGeneral;
if (!copy_handle.image) {
vk::ImageCreateFlags flags{};
if (texture_type == VideoCore::TextureType::CubeMap) {
flags |= vk::ImageCreateFlagBits::eCubeCompatible;
}
copy_handle =
MakeHandle(instance, GetScaledWidth(), GetScaledHeight(), levels, texture_type,
traits.native, traits.usage, flags, traits.aspect, false);
copy_layout = vk::ImageLayout::eUndefined;
}
runtime->renderpass_cache.EndRendering();
const RecordParams params = {
.aspect = Aspect(),
.pipeline_flags = PipelineStageFlags(),
.src_access = AccessFlags(),
.src_image = Image(),
.dst_image = copy_handle.image,
};
scheduler->Record([params, copy_layout, levels = this->levels, width = GetScaledWidth(),
height = GetScaledHeight()](vk::CommandBuffer cmdbuf) {
std::array pre_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eColorAttachmentWrite,
.dstAccessMask = vk::AccessFlagBits::eTransferRead,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferSrcOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, 0, levels),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eShaderRead,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = copy_layout,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, 0, levels),
},
};
std::array post_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferRead,
.dstAccessMask = vk::AccessFlagBits::eColorAttachmentWrite,
.oldLayout = vk::ImageLayout::eTransferSrcOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.src_image,
.subresourceRange = MakeSubresourceRange(params.aspect, 0, levels),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = vk::AccessFlagBits::eShaderRead,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = params.dst_image,
.subresourceRange = MakeSubresourceRange(params.aspect, 0, levels),
},
};
boost::container::small_vector<vk::ImageCopy, 3> image_copies;
for (u32 level = 0; level < levels; level++) {
image_copies.push_back(vk::ImageCopy{
.srcSubresource{
.aspectMask = vk::ImageAspectFlagBits::eColor,
.mipLevel = level,
.baseArrayLayer = 0,
.layerCount = 1,
},
.srcOffset = {0, 0, 0},
.dstSubresource{
.aspectMask = vk::ImageAspectFlagBits::eColor,
.mipLevel = level,
.baseArrayLayer = 0,
.layerCount = 1,
},
.dstOffset = {0, 0, 0},
.extent = {width >> level, height >> level, 1},
});
}
cmdbuf.pipelineBarrier(params.pipeline_flags, vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, pre_barriers);
cmdbuf.copyImage(params.src_image, vk::ImageLayout::eTransferSrcOptimal, params.dst_image,
vk::ImageLayout::eTransferDstOptimal, image_copies);
cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer, params.pipeline_flags,
vk::DependencyFlagBits::eByRegion, {}, {}, post_barriers);
});
return copy_handle.image_view.get();
}
vk::ImageView Surface::ImageView(u32 index) const noexcept {
const auto& image_view = handles[index].image_view.get();
if (!image_view) {
return handles[0].image_view.get();
}
return image_view;
}
vk::ImageView Surface::FramebufferView() noexcept {
is_framebuffer = true;
return ImageView();
}
vk::ImageView Surface::DepthView() noexcept {
if (depth_view) {
return depth_view.get();
}
const vk::ImageViewCreateInfo view_info = {
.image = Image(),
.viewType = vk::ImageViewType::e2D,
.format = instance->GetTraits(pixel_format).native,
.subresourceRange{
.aspectMask = vk::ImageAspectFlagBits::eDepth,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
depth_view = instance->GetDevice().createImageViewUnique(view_info);
return depth_view.get();
}
vk::ImageView Surface::StencilView() noexcept {
if (stencil_view) {
return stencil_view.get();
}
const vk::ImageViewCreateInfo view_info = {
.image = Image(),
.viewType = vk::ImageViewType::e2D,
.format = instance->GetTraits(pixel_format).native,
.subresourceRange{
.aspectMask = vk::ImageAspectFlagBits::eStencil,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
stencil_view = instance->GetDevice().createImageViewUnique(view_info);
return stencil_view.get();
}
vk::ImageView Surface::StorageView() noexcept {
if (storage_view) {
return storage_view.get();
}
if (pixel_format != VideoCore::PixelFormat::RGBA8) {
LOG_WARNING(Render_Vulkan,
"Attempted to retrieve storage view from unsupported surface with format {}",
VideoCore::PixelFormatAsString(pixel_format));
return ImageView();
}
is_storage = true;
const vk::ImageViewCreateInfo storage_view_info = {
.image = Image(),
.viewType = vk::ImageViewType::e2D,
.format = vk::Format::eR32Uint,
.subresourceRange{
.aspectMask = vk::ImageAspectFlagBits::eColor,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
},
};
storage_view = instance->GetDevice().createImageViewUnique(storage_view_info);
return storage_view.get();
}
vk::Framebuffer Surface::Framebuffer() noexcept {
const u32 index = res_scale == 1 ? 0u : 1u;
if (framebuffers[index]) {
return framebuffers[index].get();
}
const bool is_depth = type == SurfaceType::Depth || type == SurfaceType::DepthStencil;
const auto color_format = is_depth ? PixelFormat::Invalid : pixel_format;
const auto depth_format = is_depth ? pixel_format : PixelFormat::Invalid;
const auto render_pass =
runtime->renderpass_cache.GetRenderpass(color_format, depth_format, false);
const auto attachments = std::array{ImageView()};
framebuffers[index] = MakeFramebuffer(instance->GetDevice(), render_pass, GetScaledWidth(),
GetScaledHeight(), attachments);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
return framebuffers[index].get();
}
void Surface::BlitScale(const VideoCore::TextureBlit& blit, bool up_scale) {
const FormatTraits& depth_traits = instance->GetTraits(pixel_format);
const bool is_depth_stencil = pixel_format == PixelFormat::D24S8;
if (is_depth_stencil && !depth_traits.blit_support) {
LOG_WARNING(Render_Vulkan, "Depth scale unsupported by hardware");
return;
}
scheduler->Record([src_image = Image(!up_scale), aspect = Aspect(),
filter = MakeFilter(pixel_format), dst_image = Image(up_scale),
blit](vk::CommandBuffer render_cmdbuf) {
const std::array source_offsets = {
vk::Offset3D{static_cast<s32>(blit.src_rect.left),
static_cast<s32>(blit.src_rect.bottom), 0},
vk::Offset3D{static_cast<s32>(blit.src_rect.right), static_cast<s32>(blit.src_rect.top),
1},
};
const std::array dest_offsets = {
vk::Offset3D{static_cast<s32>(blit.dst_rect.left),
static_cast<s32>(blit.dst_rect.bottom), 0},
vk::Offset3D{static_cast<s32>(blit.dst_rect.right), static_cast<s32>(blit.dst_rect.top),
1},
};
const vk::ImageBlit blit_area = {
.srcSubresource{
.aspectMask = aspect,
.mipLevel = blit.src_level,
.baseArrayLayer = blit.src_layer,
.layerCount = 1,
},
.srcOffsets = source_offsets,
.dstSubresource{
.aspectMask = aspect,
.mipLevel = blit.dst_level,
.baseArrayLayer = blit.dst_layer,
.layerCount = 1,
},
.dstOffsets = dest_offsets,
};
const std::array read_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eMemoryWrite,
.dstAccessMask = vk::AccessFlagBits::eTransferRead,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferSrcOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src_image,
.subresourceRange = MakeSubresourceRange(aspect, blit.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eShaderRead |
vk::AccessFlagBits::eDepthStencilAttachmentRead |
vk::AccessFlagBits::eColorAttachmentRead |
vk::AccessFlagBits::eTransferRead,
.dstAccessMask = vk::AccessFlagBits::eTransferWrite,
.oldLayout = vk::ImageLayout::eGeneral,
.newLayout = vk::ImageLayout::eTransferDstOptimal,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst_image,
.subresourceRange = MakeSubresourceRange(aspect, blit.dst_level),
},
};
const std::array write_barriers = {
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eNone,
.dstAccessMask = vk::AccessFlagBits::eMemoryWrite | vk::AccessFlagBits::eMemoryRead,
.oldLayout = vk::ImageLayout::eTransferSrcOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src_image,
.subresourceRange = MakeSubresourceRange(aspect, blit.src_level),
},
vk::ImageMemoryBarrier{
.srcAccessMask = vk::AccessFlagBits::eTransferWrite,
.dstAccessMask = vk::AccessFlagBits::eMemoryWrite | vk::AccessFlagBits::eMemoryRead,
.oldLayout = vk::ImageLayout::eTransferDstOptimal,
.newLayout = vk::ImageLayout::eGeneral,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst_image,
.subresourceRange = MakeSubresourceRange(aspect, blit.dst_level),
},
};
render_cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eAllCommands,
vk::PipelineStageFlagBits::eTransfer,
vk::DependencyFlagBits::eByRegion, {}, {}, read_barriers);
render_cmdbuf.blitImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image,
vk::ImageLayout::eTransferDstOptimal, blit_area, filter);
render_cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eTransfer,
vk::PipelineStageFlagBits::eAllCommands,
vk::DependencyFlagBits::eByRegion, {}, {}, write_barriers);
});
}
Framebuffer::Framebuffer(TextureRuntime& runtime, const VideoCore::FramebufferParams& params,
Surface* color, Surface* depth)
: VideoCore::FramebufferParams{params}, res_scale{color ? color->res_scale
: (depth ? depth->res_scale : 1u)} {
auto& renderpass_cache = runtime.GetRenderpassCache();
if (shadow_rendering && !color) {
return;
}
width = height = std::numeric_limits<u32>::max();
const auto prepare = [&](u32 index, Surface* surface) {
const VideoCore::Extent extent = surface->RealExtent();
width = std::min(width, extent.width);
height = std::min(height, extent.height);
if (!shadow_rendering) {
formats[index] = surface->pixel_format;
}
images[index] = surface->Image();
aspects[index] = surface->Aspect();
image_views[index] = shadow_rendering ? surface->StorageView() : surface->FramebufferView();
};
boost::container::static_vector<vk::ImageView, 2> attachments;
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
if (color) {
prepare(0, color);
attachments.emplace_back(image_views[0]);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
if (depth) {
prepare(1, depth);
attachments.emplace_back(image_views[1]);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
const vk::Device device = runtime.GetInstance().GetDevice();
if (shadow_rendering) {
render_pass =
renderpass_cache.GetRenderpass(PixelFormat::Invalid, PixelFormat::Invalid, false);
framebuffer = MakeFramebuffer(device, render_pass, color->GetScaledWidth(),
color->GetScaledHeight(), {});
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
} else {
render_pass = renderpass_cache.GetRenderpass(formats[0], formats[1], false);
framebuffer = MakeFramebuffer(device, render_pass, width, height, attachments);
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
}
}
Framebuffer::~Framebuffer() = default;
Sampler::Sampler(TextureRuntime& runtime, const VideoCore::SamplerParams& params) {
using TextureConfig = VideoCore::SamplerParams::TextureConfig;
const Instance& instance = runtime.GetInstance();
const vk::PhysicalDeviceProperties properties = instance.GetPhysicalDevice().getProperties();
const bool use_border_color =
instance.IsCustomBorderColorSupported() && (params.wrap_s == TextureConfig::ClampToBorder ||
params.wrap_t == TextureConfig::ClampToBorder);
const Common::Vec4f color = PicaToVK::ColorRGBA8(params.border_color);
const vk::SamplerCustomBorderColorCreateInfoEXT border_color_info = {
.customBorderColor = MakeClearColorValue(color),
.format = vk::Format::eUndefined,
};
const vk::Filter mag_filter = PicaToVK::TextureFilterMode(params.mag_filter);
const vk::Filter min_filter = PicaToVK::TextureFilterMode(params.min_filter);
const vk::SamplerMipmapMode mipmap_mode = PicaToVK::TextureMipFilterMode(params.mip_filter);
const vk::SamplerAddressMode wrap_u = PicaToVK::WrapMode(params.wrap_s);
const vk::SamplerAddressMode wrap_v = PicaToVK::WrapMode(params.wrap_t);
const float lod_min = static_cast<float>(params.lod_min);
const float lod_max = static_cast<float>(params.lod_max);
const vk::SamplerCreateInfo sampler_info = {
.pNext = use_border_color ? &border_color_info : nullptr,
.magFilter = mag_filter,
.minFilter = min_filter,
.mipmapMode = mipmap_mode,
.addressModeU = wrap_u,
.addressModeV = wrap_v,
.mipLodBias = 0,
.anisotropyEnable = instance.IsAnisotropicFilteringSupported(),
.maxAnisotropy = properties.limits.maxSamplerAnisotropy,
.compareEnable = false,
.compareOp = vk::CompareOp::eAlways,
.minLod = lod_min,
.maxLod = lod_max,
.borderColor =
use_border_color ? vk::BorderColor::eFloatCustomEXT : vk::BorderColor::eIntOpaqueBlack,
.unnormalizedCoordinates = false,
};
sampler = instance.GetDevice().createSamplerUnique(sampler_info);
}
Sampler::~Sampler() = default;
DebugScope::DebugScope(TextureRuntime& runtime, Common::Vec4f color, std::string_view label)
: scheduler{runtime.GetScheduler()}, has_debug_tool{
runtime.GetInstance().HasDebuggingToolAttached()} {
if (!has_debug_tool) {
return;
}
scheduler.Record([color, label = std::string(label)](vk::CommandBuffer cmdbuf) {
Add vulkan backend (#6512) * code: Prepare frontend for vulkan support * citra_qt: Add vulkan options to the GUI * vk_instance: Collect tooling info * renderer_vulkan: Add vulkan backend * qt: Fix fullscreen and resize issues on macOS. (#47) * qt: Fix bugged macOS full screen transition. * renderer/vulkan: Fix swapchain recreation destroying in-use semaphore. * renderer/vulkan: Make gl_Position invariant. (#48) This fixes an issue with black artifacts in Pokemon games on Apple GPUs. If the vertex calculations differ slightly between render passes, it can cause parts of model faces to fail depth test. * vk_renderpass_cache: Bump pixel format count * android: Custom driver code * vk_instance: Set moltenvk configuration * rasterizer_cache: Proper surface unregister * citra_qt: Fix invalid characters * vk_rasterizer: Correct special unbind * android: Allow async presentation toggle * vk_graphics_pipeline: Fix async shader compilation * We were actually waiting for the pipelines regardless of the setting, oops * vk_rasterizer: More robust attribute loading * android: Move PollEvents to OpenGL window * Vulkan does not need this and it causes problems * vk_instance: Enable robust buffer access * Improves stability on mali devices * vk_renderpass_cache: Bring back renderpass flushing * externals: Update vulkan-headers * gl_rasterizer: Separable shaders for everyone * vk_blit_helper: Corect depth to color convertion * renderer_vulkan: Implement reinterpretation with copy * Allows reinterpreteration with simply copy on AMD * vk_graphics_pipeline: Only fast compile if no shaders are pending * With this shaders weren't being compiled in parallel * vk_swapchain: Ensure vsync doesn't lock framerate * vk_present_window: Match guest swapchain size to vulkan image count * Less latency and fixes crashes that were caused by images being deleted before free * vk_instance: Blacklist VK_EXT_pipeline_creation_cache_control with nvidia gpus * Resolves crashes when async shader compilation is enabled * vk_rasterizer: Bump async threshold to 6 * Many games have fullscreen quads with 6 vertices. Fixes pokemon textures missing with async shaders * android: More robust surface recreation * renderer_vulkan: Fix dynamic state being lost * vk_pipeline_cache: Skip cache save when no pipeline cache exists * This is the cache when loading a save state * sdl: Fix surface initialization on macOS. (#49) * sdl: Fix surface initialization on macOS. * sdl: Fix render window events not being handled under Vulkan. * renderer/vulkan: Fix binding/unbinding of shadow rendering buffer. * vk_stream_buffer: Respect non coherent access alignment * Required by nvidia GPUs on MacOS * renderer/vulkan: Support VK_EXT_fragment_shader_interlock for shadow rendering. (#51) * renderer_vulkan: Port some recent shader fixes * vk_pipeline_cache: Improve shadow detection * vk_swapchain: Add missing check * renderer_vulkan: Fix hybrid screen * Revert "gl_rasterizer: Separable shaders for everyone" Causes crashes on mali GPUs, will need separate PR This reverts commit d22d556d30ff641b62dfece85738c96b7fbf7061. * renderer_vulkan: Fix flipped screenshot --------- Co-authored-by: Steveice10 <1269164+Steveice10@users.noreply.github.com>
2023-09-13 00:28:50 +02:00
const vk::DebugUtilsLabelEXT debug_label = {
.pLabelName = label.data(),
.color = std::array{color[0], color[1], color[2], color[3]},
};
cmdbuf.beginDebugUtilsLabelEXT(debug_label);
});
}
DebugScope::~DebugScope() {
if (!has_debug_tool) {
return;
}
scheduler.Record([](vk::CommandBuffer cmdbuf) { cmdbuf.endDebugUtilsLabelEXT(); });
}
} // namespace Vulkan