Compare commits

...

32 Commits

Author SHA1 Message Date
Fernando Sahmkow
5a3463bc2b Texture Cache: Secure insertions against deletions. 2021-11-28 10:53:27 +01:00
Fernando Sahmkow
ecefc932e6 Texture Cache: Redesigning the blitting system (again). 2021-11-27 11:22:16 +01:00
Fernando Sahmkow
1624f307d0 Texture Cache: Further fix regressions. 2021-11-26 17:03:48 +01:00
Fernando Sahmkow
08674aee87 Texture Cache: Fix issue with blitting 3D textures. 2021-11-22 06:07:21 +01:00
Fernando Sahmkow
1e474fb9d1 Texture Cache: Correct conversion shaders. 2021-11-22 00:21:42 +01:00
Fernando Sahmkow
0902119302 Texture Cache: Always copy on NVIDIA. 2021-11-22 00:06:56 +01:00
Fernando Sahmkow
8532849439 TextureCache: Simplify blitting of D24S8 formats and fix bugs. 2021-11-22 00:00:01 +01:00
Fernando Sahmkow
d7f4434bd5 VulkanTexturECache: Use reinterpret on D32_S8 formats. 2021-11-21 21:09:49 +01:00
Fernando Sahmkow
b96caf200d HostShaders: Fix D24S8 convertion shaders. 2021-11-21 21:04:04 +01:00
Fernando Sahmkow
779f4ac72d TextureCache: Eliminate format deduction as full depth conversion has been supported. 2021-11-21 05:37:01 +01:00
bunnei
ea6fa044f3 Merge pull request #7368 from FernandoS27/vulkan-conv
Fix ART Blit detection regression and add D24S8 <-> RGBA8 conv to Vulkan
2021-11-20 16:51:13 -08:00
Fernando Sahmkow
da2fe81905 TextureCache: Refactor and fix linux compiling. 2021-11-20 14:46:19 +01:00
Fernando Sahmkow
4ca6e9a9e2 TextureCache: Assure full conversions on depth/stencil write shaders. 2021-11-20 06:17:01 +01:00
Fernando Sahmkow
0857f82913 TextureCache: Implement buffer copies on Vulkan. 2021-11-20 06:15:29 +01:00
bunnei
c3e1ffc44b Merge pull request #7294 from vonchenplus/fix_image_update_error_when_width_too_small
Fix image update/download error when width too small
2021-11-19 15:56:27 -08:00
Fernando Sahmkow
e02cff2f69 TextureCache: Add R16G16 to D24S8 converter. 2021-11-20 00:02:12 +01:00
Fernando Sahmkow
1d5e6a51d7 TextureCache: Add B10G11R11 to D24S8 converter. 2021-11-19 23:22:44 +01:00
Fernando Sahmkow
6f896d1fae TextureCache: Further fixes on resolve algorithm. 2021-11-19 23:02:04 +01:00
bunnei
b44fbf6cdd Merge pull request #7369 from Morph1984/amd-fsr-statusbar
main: Shorten AMD FSR status bar text
2021-11-19 01:20:27 -08:00
bunnei
d2e009f355 Merge pull request #7273 from fpdotmonkey/submodule-checkout-command-on-new-line
Move the cmake submodule checkout command to a new line
2021-11-19 01:19:50 -08:00
bunnei
a69813948f Merge pull request #7342 from goldenx86/patch-3
Replace keys error pop up
2021-11-19 01:17:18 -08:00
bunnei
c45af76ea0 Merge pull request #7357 from Morph1984/s8_uint
video_core: Implement S8_UINT format
2021-11-19 01:16:49 -08:00
Fernando Sahmkow
b805c7bf05 TextureCache: Implement additional D24S8 convertions. 2021-11-19 06:27:44 +01:00
Fernando Sahmkow
0ff228405f TextureCache: force same image format when resolving an image. 2021-11-19 05:46:57 +01:00
Fernando Sahmkow
b130f648d7 TextureCache: Fix regression caused by ART and improve blit detection algorithm to be smarter. 2021-11-19 03:17:54 +01:00
Fernando Sahmkow
2ec7fcecb7 Vulkan: implement D24S8 <-> RGBA8 convertions. 2021-11-19 03:17:02 +01:00
Morph
dc61b7045b renderer_vulkan: Implement S8_UINT stencil format
It should be noted that on Windows, only nvidia gpus support this format natively as of this commit.
2021-11-18 00:05:51 -05:00
Morph
6dd6dc046c renderer_opengl: Implement S8_UINT stencil format 2021-11-17 15:05:07 -05:00
Morph
2348eb41f3 video_core: Add S8_UINT stencil format 2021-11-17 15:04:38 -05:00
Feng Chen
894cc9d876 Fix image update/download error when width too small 2021-11-17 12:21:17 +08:00
Matías Locatti
ffb79afd29 Replace keys error pop up
Fight me.
2021-11-16 03:53:28 -03:00
Fletcher Porter
2b3d66fe69 Move the cmake submodule checkout command to a new line
Presently, if you forget to initialize the git submodules before
running cmake, there'll be a helpful message that reminds you to do so.
However, on narrow terminals (e.g. 80 wide) there's a word wrap that
includes a new line in the middle of the git command, precluding easy
copy-paste.  This moves the entire git command to its own line to avoid
such tragedies.

Before:

```
CMake Error at CMakeLists.txt:59 (message):
  Git submodule externals/inih/inih not found.  Please run: git submodule
  update --init --recursive
```

After:

```
CMake Error at CMakeLists.txt:59 (message):
  Git submodule externals/inih/inih not found.  Please run:

  git submodule update --init --recursive
```
2021-11-02 00:01:39 -07:00
23 changed files with 637 additions and 57 deletions

View File

@@ -57,7 +57,7 @@ function(check_submodules_present)
string(REGEX REPLACE "path *= *" "" module ${module})
if (NOT EXISTS "${PROJECT_SOURCE_DIR}/${module}/.git")
message(FATAL_ERROR "Git submodule ${module} not found. "
"Please run: git submodule update --init --recursive")
"Please run: \ngit submodule update --init --recursive")
endif()
endforeach()
endfunction()

View File

@@ -7,6 +7,7 @@
#include <bit>
#include <climits>
#include <cstddef>
#include <type_traits>
#include "common/common_types.h"
@@ -44,4 +45,10 @@ template <typename T>
return static_cast<u32>(log2_f + static_cast<u64>((value ^ (1ULL << log2_f)) != 0ULL));
}
template <typename T>
requires std::is_integral_v<T>
[[nodiscard]] T NextPow2(T value) {
return static_cast<T>(1ULL << ((8U * sizeof(T)) - std::countl_zero(value - 1U)));
}
} // namespace Common

View File

@@ -83,6 +83,7 @@ enum class DepthFormat : u32 {
S8_UINT_Z24_UNORM = 0x14,
D24X8_UNORM = 0x15,
D24S8_UNORM = 0x16,
S8_UINT = 0x17,
D24C8_UNORM = 0x18,
D32_FLOAT_S8X24_UINT = 0x19,
};

View File

@@ -10,6 +10,8 @@ set(SHADER_FILES
astc_decoder.comp
block_linear_unswizzle_2d.comp
block_linear_unswizzle_3d.comp
convert_abgr8_to_d24s8.frag
convert_d24s8_to_abgr8.frag
convert_depth_to_float.frag
convert_float_to_depth.frag
full_screen_triangle.vert

View File

@@ -0,0 +1,18 @@
// Copyright 2021 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#version 450
#extension GL_ARB_shader_stencil_export : require
layout(binding = 0) uniform sampler2D color_texture;
void main() {
ivec2 coord = ivec2(gl_FragCoord.xy);
uvec4 color = uvec4(texelFetch(color_texture, coord, 0).abgr * (exp2(8) - 1.0f));
uvec4 bytes = color << uvec4(24, 16, 8, 0);
uint depth_stencil_unorm = bytes.x | bytes.y | bytes.z | bytes.w;
gl_FragDepth = float(depth_stencil_unorm & 0x00FFFFFFu) / (exp2(24.0) - 1.0f);
gl_FragStencilRefARB = int(depth_stencil_unorm >> 24);
}

View File

@@ -0,0 +1,23 @@
// Copyright 2021 yuzu Emulator Project
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
#version 450
layout(binding = 0) uniform sampler2D depth_tex;
layout(binding = 1) uniform isampler2D stencil_tex;
layout(location = 0) out vec4 color;
void main() {
ivec2 coord = ivec2(gl_FragCoord.xy);
uint depth = uint(textureLod(depth_tex, coord, 0).r * (exp2(24.0) - 1.0f));
uint stencil = uint(textureLod(stencil_tex, coord, 0).r);
highp uint depth_val =
uint(textureLod(depth_tex, coord, 0).r * (exp2(32.0) - 1.0));
lowp uint stencil_val = textureLod(stencil_tex, coord, 0).r;
highp uvec4 components =
uvec4(stencil_val, (uvec3(depth_val) >> uvec3(24u, 16u, 8u)) & 0x000000FFu);
color.abgr = vec4(components) / (exp2(8.0) - 1.0);
}

View File

@@ -9,6 +9,7 @@
#include <glad/glad.h>
#include "common/bit_util.h"
#include "common/literals.h"
#include "common/settings.h"
#include "video_core/renderer_opengl/gl_device.h"
@@ -148,6 +149,8 @@ GLenum AttachmentType(PixelFormat format) {
switch (const SurfaceType type = VideoCore::Surface::GetFormatType(format); type) {
case SurfaceType::Depth:
return GL_DEPTH_ATTACHMENT;
case SurfaceType::Stencil:
return GL_STENCIL_ATTACHMENT;
case SurfaceType::DepthStencil:
return GL_DEPTH_STENCIL_ATTACHMENT;
default:
@@ -317,13 +320,12 @@ void AttachTexture(GLuint fbo, GLenum attachment, const ImageView* image_view) {
}
}
OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_format) {
OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_format,
GLsizei gl_num_levels) {
const GLenum target = ImageTarget(info);
const GLsizei width = info.size.width;
const GLsizei height = info.size.height;
const GLsizei depth = info.size.depth;
const int max_host_mip_levels = std::bit_width(info.size.width);
const GLsizei num_levels = std::min(info.resources.levels, max_host_mip_levels);
const GLsizei num_layers = info.resources.layers;
const GLsizei num_samples = info.num_samples;
@@ -335,10 +337,10 @@ OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_form
}
switch (target) {
case GL_TEXTURE_1D_ARRAY:
glTextureStorage2D(handle, num_levels, gl_internal_format, width, num_layers);
glTextureStorage2D(handle, gl_num_levels, gl_internal_format, width, num_layers);
break;
case GL_TEXTURE_2D_ARRAY:
glTextureStorage3D(handle, num_levels, gl_internal_format, width, height, num_layers);
glTextureStorage3D(handle, gl_num_levels, gl_internal_format, width, height, num_layers);
break;
case GL_TEXTURE_2D_MULTISAMPLE_ARRAY: {
// TODO: Where should 'fixedsamplelocations' come from?
@@ -348,10 +350,10 @@ OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_form
break;
}
case GL_TEXTURE_RECTANGLE:
glTextureStorage2D(handle, num_levels, gl_internal_format, width, height);
glTextureStorage2D(handle, gl_num_levels, gl_internal_format, width, height);
break;
case GL_TEXTURE_3D:
glTextureStorage3D(handle, num_levels, gl_internal_format, width, height, depth);
glTextureStorage3D(handle, gl_num_levels, gl_internal_format, width, height, depth);
break;
case GL_TEXTURE_BUFFER:
UNREACHABLE();
@@ -397,9 +399,6 @@ OGLTexture MakeImage(const VideoCommon::ImageInfo& info, GLenum gl_internal_form
return GL_R32UI;
}
[[nodiscard]] u32 NextPow2(u32 value) {
return 1U << (32U - std::countl_zero(value - 1U));
}
} // Anonymous namespace
ImageBufferMap::~ImageBufferMap() {
@@ -526,8 +525,8 @@ void TextureCacheRuntime::CopyImage(Image& dst_image, Image& src_image,
}
}
void TextureCacheRuntime::ConvertImage(Image& dst, Image& src,
std::span<const VideoCommon::ImageCopy> copies) {
void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
std::span<const VideoCommon::ImageCopy> copies) {
LOG_DEBUG(Render_OpenGL, "Converting {} to {}", src.info.format, dst.info.format);
format_conversion_pass.ConvertImage(dst, src, copies);
}
@@ -696,7 +695,9 @@ Image::Image(TextureCacheRuntime& runtime_, const VideoCommon::ImageInfo& info_,
gl_format = tuple.format;
gl_type = tuple.type;
}
texture = MakeImage(info, gl_internal_format);
const int max_host_mip_levels = std::bit_width(info.size.width);
gl_num_levels = std::min(info.resources.levels, max_host_mip_levels);
texture = MakeImage(info, gl_internal_format, gl_num_levels);
current_texture = texture.handle;
if (runtime->device.HasDebuggingToolAttached()) {
const std::string name = VideoCommon::Name(*this);
@@ -724,6 +725,9 @@ void Image::UploadMemory(const ImageBufferMap& map,
u32 current_image_height = std::numeric_limits<u32>::max();
for (const VideoCommon::BufferImageCopy& copy : copies) {
if (copy.image_subresource.base_level >= gl_num_levels) {
continue;
}
if (current_row_length != copy.buffer_row_length) {
current_row_length = copy.buffer_row_length;
glPixelStorei(GL_UNPACK_ROW_LENGTH, current_row_length);
@@ -753,6 +757,9 @@ void Image::DownloadMemory(ImageBufferMap& map,
u32 current_image_height = std::numeric_limits<u32>::max();
for (const VideoCommon::BufferImageCopy& copy : copies) {
if (copy.image_subresource.base_level >= gl_num_levels) {
continue;
}
if (current_row_length != copy.buffer_row_length) {
current_row_length = copy.buffer_row_length;
glPixelStorei(GL_PACK_ROW_LENGTH, current_row_length);
@@ -792,7 +799,7 @@ GLuint Image::StorageHandle() noexcept {
}
store_view.Create();
glTextureView(store_view.handle, ImageTarget(info), current_texture, GL_RGBA8, 0,
info.resources.levels, 0, info.resources.layers);
gl_num_levels, 0, info.resources.layers);
return store_view.handle;
default:
return current_texture;
@@ -907,6 +914,8 @@ void Image::Scale(bool up_scale) {
return GL_COLOR_ATTACHMENT0;
case SurfaceType::Depth:
return GL_DEPTH_ATTACHMENT;
case SurfaceType::Stencil:
return GL_STENCIL_ATTACHMENT;
case SurfaceType::DepthStencil:
return GL_DEPTH_STENCIL_ATTACHMENT;
default:
@@ -920,8 +929,10 @@ void Image::Scale(bool up_scale) {
return GL_COLOR_BUFFER_BIT;
case SurfaceType::Depth:
return GL_DEPTH_BUFFER_BIT;
case SurfaceType::Stencil:
return GL_STENCIL_BUFFER_BIT;
case SurfaceType::DepthStencil:
return GL_STENCIL_BUFFER_BIT | GL_DEPTH_BUFFER_BIT;
return GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
default:
UNREACHABLE();
return GL_COLOR_BUFFER_BIT;
@@ -933,8 +944,10 @@ void Image::Scale(bool up_scale) {
return 0;
case SurfaceType::Depth:
return 1;
case SurfaceType::DepthStencil:
case SurfaceType::Stencil:
return 2;
case SurfaceType::DepthStencil:
return 3;
default:
UNREACHABLE();
return 0;
@@ -956,7 +969,7 @@ void Image::Scale(bool up_scale) {
auto dst_info = info;
dst_info.size.width = scaled_width;
dst_info.size.height = scaled_height;
upscaled_backup = MakeImage(dst_info, gl_internal_format);
upscaled_backup = MakeImage(dst_info, gl_internal_format, gl_num_levels);
}
const u32 src_width = up_scale ? original_width : scaled_width;
const u32 src_height = up_scale ? original_height : scaled_height;
@@ -1264,10 +1277,20 @@ Framebuffer::Framebuffer(TextureCacheRuntime& runtime, std::span<ImageView*, NUM
}
if (const ImageView* const image_view = depth_buffer; image_view) {
if (GetFormatType(image_view->format) == SurfaceType::DepthStencil) {
buffer_bits |= GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
} else {
switch (GetFormatType(image_view->format)) {
case SurfaceType::Depth:
buffer_bits |= GL_DEPTH_BUFFER_BIT;
break;
case SurfaceType::Stencil:
buffer_bits |= GL_STENCIL_BUFFER_BIT;
break;
case SurfaceType::DepthStencil:
buffer_bits |= GL_DEPTH_BUFFER_BIT | GL_STENCIL_BUFFER_BIT;
break;
default:
UNREACHABLE();
buffer_bits |= GL_DEPTH_BUFFER_BIT;
break;
}
const GLenum attachment = AttachmentType(image_view->format);
AttachTexture(handle, attachment, image_view);
@@ -1308,7 +1331,7 @@ void FormatConversionPass::ConvertImage(Image& dst_image, Image& src_image,
const u32 copy_size = region.width * region.height * region.depth * img_bpp;
if (pbo_size < copy_size) {
intermediate_pbo.Create();
pbo_size = NextPow2(copy_size);
pbo_size = Common::NextPow2(copy_size);
glNamedBufferData(intermediate_pbo.handle, pbo_size, nullptr, GL_STREAM_COPY);
}
// Copy from source to PBO

View File

@@ -84,9 +84,13 @@ public:
u64 GetDeviceLocalMemory() const;
bool ShouldReinterpret([[maybe_unused]] Image& dst, [[maybe_unused]] Image& src) {
return true;
}
void CopyImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
void ConvertImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
void ReinterpretImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
void ConvertImage(Framebuffer* dst, ImageView& dst_view, ImageView& src_view, bool rescaled) {
UNIMPLEMENTED();
@@ -164,8 +168,8 @@ private:
std::array<GLuint, Shader::NUM_TEXTURE_TYPES> null_image_views{};
std::array<OGLFramebuffer, 3> rescale_draw_fbos;
std::array<OGLFramebuffer, 3> rescale_read_fbos;
std::array<OGLFramebuffer, 4> rescale_draw_fbos;
std::array<OGLFramebuffer, 4> rescale_read_fbos;
const Settings::ResolutionScalingInfo& resolution;
};
@@ -221,6 +225,7 @@ private:
GLenum gl_internal_format = GL_NONE;
GLenum gl_format = GL_NONE;
GLenum gl_type = GL_NONE;
GLsizei gl_num_levels{};
TextureCacheRuntime* runtime{};
GLuint current_texture{};
};
@@ -338,7 +343,6 @@ struct TextureCacheParams {
static constexpr bool FRAMEBUFFER_BLITS = true;
static constexpr bool HAS_EMULATED_COPIES = true;
static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
static constexpr bool HAS_PIXEL_FORMAT_CONVERSIONS = true;
using Runtime = OpenGL::TextureCacheRuntime;
using Image = OpenGL::Image;

View File

@@ -108,6 +108,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> FORMAT_TAB
{GL_RGB9_E5, GL_RGB, GL_UNSIGNED_INT_5_9_9_9_REV}, // E5B9G9R9_FLOAT
{GL_DEPTH_COMPONENT32F, GL_DEPTH_COMPONENT, GL_FLOAT}, // D32_FLOAT
{GL_DEPTH_COMPONENT16, GL_DEPTH_COMPONENT, GL_UNSIGNED_SHORT}, // D16_UNORM
{GL_STENCIL_INDEX8, GL_STENCIL, GL_UNSIGNED_BYTE}, // S8_UINT
{GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // D24_UNORM_S8_UINT
{GL_DEPTH24_STENCIL8, GL_DEPTH_STENCIL, GL_UNSIGNED_INT_24_8}, // S8_UINT_D24_UNORM
{GL_DEPTH32F_STENCIL8, GL_DEPTH_STENCIL,

View File

@@ -4,6 +4,8 @@
#include <algorithm>
#include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
#include "video_core/host_shaders/convert_d24s8_to_abgr8_frag_spv.h"
#include "video_core/host_shaders/convert_depth_to_float_frag_spv.h"
#include "video_core/host_shaders/convert_float_to_depth_frag_spv.h"
#include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
@@ -354,6 +356,8 @@ BlitImageHelper::BlitImageHelper(const Device& device_, VKScheduler& scheduler_,
blit_color_to_color_frag(BuildShader(device, VULKAN_BLIT_COLOR_FLOAT_FRAG_SPV)),
convert_depth_to_float_frag(BuildShader(device, CONVERT_DEPTH_TO_FLOAT_FRAG_SPV)),
convert_float_to_depth_frag(BuildShader(device, CONVERT_FLOAT_TO_DEPTH_FRAG_SPV)),
convert_abgr8_to_d24s8_frag(BuildShader(device, CONVERT_ABGR8_TO_D24S8_FRAG_SPV)),
convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {
if (device.IsExtShaderStencilExportSupported()) {
@@ -448,6 +452,22 @@ void BlitImageHelper::ConvertR16ToD16(const Framebuffer* dst_framebuffer,
Convert(*convert_r16_to_d16_pipeline, dst_framebuffer, src_image_view, up_scale, down_shift);
}
void BlitImageHelper::ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift) {
ConvertPipelineDepthTargetEx(convert_abgr8_to_d24s8_pipeline, dst_framebuffer->RenderPass(),
convert_abgr8_to_d24s8_frag, true);
ConvertColor(*convert_abgr8_to_d24s8_pipeline, dst_framebuffer, src_image_view, up_scale,
down_shift);
}
void BlitImageHelper::ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift) {
ConvertPipelineColorTargetEx(convert_d24s8_to_abgr8_pipeline, dst_framebuffer->RenderPass(),
convert_d24s8_to_abgr8_frag, false);
ConvertDepthStencil(*convert_d24s8_to_abgr8_pipeline, dst_framebuffer, src_image_view, up_scale,
down_shift);
}
void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
const ImageView& src_image_view, u32 up_scale, u32 down_shift) {
const VkPipelineLayout layout = *one_texture_pipeline_layout;
@@ -495,6 +515,101 @@ void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_frameb
scheduler.InvalidateState();
}
void BlitImageHelper::ConvertColor(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift) {
const VkPipelineLayout layout = *one_texture_pipeline_layout;
const VkImageView src_view = src_image_view.ColorView();
const VkSampler sampler = *nearest_sampler;
const VkExtent2D extent{
.width = std::max((src_image_view.size.width * up_scale) >> down_shift, 1U),
.height = std::max((src_image_view.size.height * up_scale) >> down_shift, 1U),
};
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, sampler, src_view, extent, up_scale, down_shift,
this](vk::CommandBuffer cmdbuf) {
const VkOffset2D offset{
.x = 0,
.y = 0,
};
const VkViewport viewport{
.x = 0.0f,
.y = 0.0f,
.width = static_cast<float>(extent.width),
.height = static_cast<float>(extent.height),
.minDepth = 0.0f,
.maxDepth = 0.0f,
};
const VkRect2D scissor{
.offset = offset,
.extent = extent,
};
const PushConstants push_constants{
.tex_scale = {viewport.width, viewport.height},
.tex_offset = {0.0f, 0.0f},
};
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
// TODO: Barriers
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
cmdbuf.SetViewport(0, viewport);
cmdbuf.SetScissor(0, scissor);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
void BlitImageHelper::ConvertDepthStencil(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift) {
const VkPipelineLayout layout = *two_textures_pipeline_layout;
const VkImageView src_depth_view = src_image_view.DepthView();
const VkImageView src_stencil_view = src_image_view.StencilView();
const VkSampler sampler = *nearest_sampler;
const VkExtent2D extent{
.width = std::max((src_image_view.size.width * up_scale) >> down_shift, 1U),
.height = std::max((src_image_view.size.height * up_scale) >> down_shift, 1U),
};
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([pipeline, layout, sampler, src_depth_view, src_stencil_view, extent, up_scale,
down_shift, this](vk::CommandBuffer cmdbuf) {
const VkOffset2D offset{
.x = 0,
.y = 0,
};
const VkViewport viewport{
.x = 0.0f,
.y = 0.0f,
.width = static_cast<float>(extent.width),
.height = static_cast<float>(extent.height),
.minDepth = 0.0f,
.maxDepth = 0.0f,
};
const VkRect2D scissor{
.offset = offset,
.extent = extent,
};
const PushConstants push_constants{
.tex_scale = {viewport.width, viewport.height},
.tex_offset = {0.0f, 0.0f},
};
const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
src_stencil_view);
// TODO: Barriers
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
cmdbuf.SetViewport(0, viewport);
cmdbuf.SetScissor(0, scissor);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_VERTEX_BIT, push_constants);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
VkPipeline BlitImageHelper::FindOrEmplaceColorPipeline(const BlitImagePipelineKey& key) {
const auto it = std::ranges::find(blit_color_keys, key);
if (it != blit_color_keys.end()) {
@@ -636,4 +751,62 @@ void BlitImageHelper::ConvertColorToDepthPipeline(vk::Pipeline& pipeline, VkRend
});
}
void BlitImageHelper::ConvertPipelineColorTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
vk::ShaderModule& module, bool single_texture) {
if (pipeline) {
return;
}
const std::array stages = MakeStages(*full_screen_vert, *module);
pipeline = device.GetLogical().CreateGraphicsPipeline({
.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stageCount = static_cast<u32>(stages.size()),
.pStages = stages.data(),
.pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
.pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
.pTessellationState = nullptr,
.pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
.pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
.pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pDepthStencilState = nullptr,
.pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_GENERIC_CREATE_INFO,
.pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
.layout = single_texture ? *one_texture_pipeline_layout : *two_textures_pipeline_layout,
.renderPass = renderpass,
.subpass = 0,
.basePipelineHandle = VK_NULL_HANDLE,
.basePipelineIndex = 0,
});
}
void BlitImageHelper::ConvertPipelineDepthTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
vk::ShaderModule& module, bool single_texture) {
if (pipeline) {
return;
}
const std::array stages = MakeStages(*full_screen_vert, *module);
pipeline = device.GetLogical().CreateGraphicsPipeline({
.sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.stageCount = static_cast<u32>(stages.size()),
.pStages = stages.data(),
.pVertexInputState = &PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
.pInputAssemblyState = &PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
.pTessellationState = nullptr,
.pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
.pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
.pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pDepthStencilState = &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
.pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
.pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
.layout = single_texture ? *one_texture_pipeline_layout : *two_textures_pipeline_layout,
.renderPass = renderpass,
.subpass = 0,
.basePipelineHandle = VK_NULL_HANDLE,
.basePipelineIndex = 0,
});
}
} // namespace Vulkan

View File

@@ -56,10 +56,22 @@ public:
void ConvertR16ToD16(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
u32 up_scale, u32 down_shift);
void ConvertABGR8ToD24S8(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
u32 up_scale, u32 down_shift);
void ConvertD24S8ToABGR8(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
u32 up_scale, u32 down_shift);
private:
void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
const ImageView& src_image_view, u32 up_scale, u32 down_shift);
void ConvertColor(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift);
void ConvertDepthStencil(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
ImageView& src_image_view, u32 up_scale, u32 down_shift);
[[nodiscard]] VkPipeline FindOrEmplaceColorPipeline(const BlitImagePipelineKey& key);
[[nodiscard]] VkPipeline FindOrEmplaceDepthStencilPipeline(const BlitImagePipelineKey& key);
@@ -68,6 +80,12 @@ private:
void ConvertColorToDepthPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass);
void ConvertPipelineColorTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
vk::ShaderModule& module, bool single_texture);
void ConvertPipelineDepthTargetEx(vk::Pipeline& pipeline, VkRenderPass renderpass,
vk::ShaderModule& module, bool single_texture);
const Device& device;
VKScheduler& scheduler;
StateTracker& state_tracker;
@@ -83,6 +101,8 @@ private:
vk::ShaderModule blit_depth_stencil_frag;
vk::ShaderModule convert_depth_to_float_frag;
vk::ShaderModule convert_float_to_depth_frag;
vk::ShaderModule convert_abgr8_to_d24s8_frag;
vk::ShaderModule convert_d24s8_to_abgr8_frag;
vk::Sampler linear_sampler;
vk::Sampler nearest_sampler;
@@ -94,6 +114,8 @@ private:
vk::Pipeline convert_r32_to_d32_pipeline;
vk::Pipeline convert_d16_to_r16_pipeline;
vk::Pipeline convert_r16_to_d16_pipeline;
vk::Pipeline convert_abgr8_to_d24s8_pipeline;
vk::Pipeline convert_d24s8_to_abgr8_pipeline;
};
} // namespace Vulkan

View File

@@ -208,6 +208,9 @@ struct FormatTuple {
{VK_FORMAT_D32_SFLOAT, Attachable}, // D32_FLOAT
{VK_FORMAT_D16_UNORM, Attachable}, // D16_UNORM
// Stencil formats
{VK_FORMAT_S8_UINT, Attachable}, // S8_UINT
// DepthStencil formats
{VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // D24_UNORM_S8_UINT
{VK_FORMAT_D24_UNORM_S8_UINT, Attachable}, // S8_UINT_D24_UNORM (emulated)

View File

@@ -8,6 +8,7 @@
#include <vector>
#include "common/bit_cast.h"
#include "common/bit_util.h"
#include "common/settings.h"
#include "video_core/engines/fermi_2d.h"
@@ -102,6 +103,7 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
break;
case VideoCore::Surface::SurfaceType::Depth:
case VideoCore::Surface::SurfaceType::Stencil:
case VideoCore::Surface::SurfaceType::DepthStencil:
usage |= VK_IMAGE_USAGE_DEPTH_STENCIL_ATTACHMENT_BIT;
break;
@@ -173,6 +175,8 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
return VK_IMAGE_ASPECT_COLOR_BIT;
case VideoCore::Surface::SurfaceType::Depth:
return VK_IMAGE_ASPECT_DEPTH_BIT;
case VideoCore::Surface::SurfaceType::Stencil:
return VK_IMAGE_ASPECT_STENCIL_BIT;
case VideoCore::Surface::SurfaceType::DepthStencil:
return VK_IMAGE_ASPECT_DEPTH_BIT | VK_IMAGE_ASPECT_STENCIL_BIT;
default:
@@ -195,6 +199,8 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
case PixelFormat::D16_UNORM:
case PixelFormat::D32_FLOAT:
return VK_IMAGE_ASPECT_DEPTH_BIT;
case PixelFormat::S8_UINT:
return VK_IMAGE_ASPECT_STENCIL_BIT;
default:
return VK_IMAGE_ASPECT_COLOR_BIT;
}
@@ -308,6 +314,19 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
};
}
[[nodiscard]] VkBufferImageCopy MakeBufferImageCopy(const VideoCommon::ImageCopy& copy, bool is_src,
VkImageAspectFlags aspect_mask) noexcept {
return VkBufferImageCopy{
.bufferOffset = 0,
.bufferRowLength = 0,
.bufferImageHeight = 0,
.imageSubresource = MakeImageSubresourceLayers(
is_src ? copy.src_subresource : copy.dst_subresource, aspect_mask),
.imageOffset = MakeOffset3D(is_src ? copy.src_offset : copy.dst_offset),
.imageExtent = MakeExtent3D(copy.extent),
};
}
[[maybe_unused]] [[nodiscard]] std::vector<VkBufferCopy> TransformBufferCopies(
std::span<const VideoCommon::BufferCopy> copies, size_t buffer_offset) {
std::vector<VkBufferCopy> result(copies.size());
@@ -754,6 +773,173 @@ StagingBufferRef TextureCacheRuntime::DownloadStagingBuffer(size_t size) {
return staging_buffer_pool.Request(size, MemoryUsage::Download);
}
bool TextureCacheRuntime::ShouldReinterpret(Image& dst, Image& src) {
if (VideoCore::Surface::GetFormatType(dst.info.format) ==
VideoCore::Surface::SurfaceType::DepthStencil &&
!device.IsExtShaderStencilExportSupported()) {
return true;
}
if (VideoCore::Surface::GetFormatType(src.info.format) ==
VideoCore::Surface::SurfaceType::DepthStencil &&
!device.IsExtShaderStencilExportSupported()) {
return true;
}
if (dst.info.format == PixelFormat::D32_FLOAT_S8_UINT ||
src.info.format == PixelFormat::D32_FLOAT_S8_UINT) {
return true;
}
return false;
}
VkBuffer TextureCacheRuntime::GetTemporaryBuffer(size_t needed_size) {
const auto level = (8 * sizeof(size_t)) - std::countl_zero(needed_size - 1ULL);
if (buffer_commits[level]) {
return *buffers[level];
}
const auto new_size = Common::NextPow2(needed_size);
VkBufferUsageFlags flags = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
VK_BUFFER_USAGE_UNIFORM_TEXEL_BUFFER_BIT |
VK_BUFFER_USAGE_STORAGE_TEXEL_BUFFER_BIT;
buffers[level] = device.GetLogical().CreateBuffer({
.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.size = new_size,
.usage = flags,
.sharingMode = VK_SHARING_MODE_EXCLUSIVE,
.queueFamilyIndexCount = 0,
.pQueueFamilyIndices = nullptr,
});
buffer_commits[level] = std::make_unique<MemoryCommit>(
memory_allocator.Commit(buffers[level], MemoryUsage::DeviceLocal));
return *buffers[level];
}
void TextureCacheRuntime::ReinterpretImage(Image& dst, Image& src,
std::span<const VideoCommon::ImageCopy> copies) {
std::vector<VkBufferImageCopy> vk_in_copies(copies.size());
std::vector<VkBufferImageCopy> vk_out_copies(copies.size());
const VkImageAspectFlags src_aspect_mask = src.AspectMask();
const VkImageAspectFlags dst_aspect_mask = dst.AspectMask();
std::ranges::transform(copies, vk_in_copies.begin(), [src_aspect_mask](const auto& copy) {
return MakeBufferImageCopy(copy, true, src_aspect_mask);
});
std::ranges::transform(copies, vk_out_copies.begin(), [dst_aspect_mask](const auto& copy) {
return MakeBufferImageCopy(copy, false, dst_aspect_mask);
});
const u32 img_bpp = BytesPerBlock(src.info.format);
size_t total_size = 0;
for (const auto& copy : copies) {
total_size += copy.extent.width * copy.extent.height * copy.extent.depth * img_bpp;
}
const VkBuffer copy_buffer = GetTemporaryBuffer(total_size);
const VkImage dst_image = dst.Handle();
const VkImage src_image = src.Handle();
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([dst_image, src_image, copy_buffer, src_aspect_mask, dst_aspect_mask,
vk_in_copies, vk_out_copies](vk::CommandBuffer cmdbuf) {
RangedBarrierRange dst_range;
RangedBarrierRange src_range;
for (const VkBufferImageCopy& copy : vk_in_copies) {
src_range.AddLayers(copy.imageSubresource);
}
for (const VkBufferImageCopy& copy : vk_out_copies) {
dst_range.AddLayers(copy.imageSubresource);
}
static constexpr VkMemoryBarrier READ_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_MEMORY_WRITE_BIT,
.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
};
static constexpr VkMemoryBarrier WRITE_BARRIER{
.sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_MEMORY_READ_BIT | VK_ACCESS_MEMORY_WRITE_BIT,
};
const std::array pre_barriers{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_TRANSFER_READ_BIT,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src_image,
.subresourceRange = src_range.SubresourceRange(src_aspect_mask),
},
};
const std::array middle_in_barrier{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = 0,
.dstAccessMask = 0,
.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src_image,
.subresourceRange = src_range.SubresourceRange(src_aspect_mask),
},
};
const std::array middle_out_barrier{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst_image,
.subresourceRange = dst_range.SubresourceRange(dst_aspect_mask),
},
};
const std::array post_barriers{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT |
VK_ACCESS_TRANSFER_READ_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
.oldLayout = VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst_image,
.subresourceRange = dst_range.SubresourceRange(dst_aspect_mask),
},
};
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
0, {}, {}, pre_barriers);
cmdbuf.CopyImageToBuffer(src_image, VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL, copy_buffer,
vk_in_copies);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
0, WRITE_BARRIER, nullptr, middle_in_barrier);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_ALL_COMMANDS_BIT, VK_PIPELINE_STAGE_TRANSFER_BIT,
0, READ_BARRIER, {}, middle_out_barrier);
cmdbuf.CopyBufferToImage(copy_buffer, dst_image, VK_IMAGE_LAYOUT_GENERAL, vk_out_copies);
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_TRANSFER_BIT, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT,
0, {}, {}, post_barriers);
});
}
void TextureCacheRuntime::BlitImage(Framebuffer* dst_framebuffer, ImageView& dst, ImageView& src,
const Region2D& dst_region, const Region2D& src_region,
Tegra::Engines::Fermi2D::Filter filter,
@@ -881,6 +1067,11 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
return blit_image_helper.ConvertD16ToR16(dst, src_view, up_scale, down_shift);
}
break;
case PixelFormat::A8B8G8R8_UNORM:
if (src_view.format == PixelFormat::S8_UINT_D24_UNORM) {
return blit_image_helper.ConvertD24S8ToABGR8(dst, src_view, up_scale, down_shift);
}
break;
case PixelFormat::R32_FLOAT:
if (src_view.format == PixelFormat::D32_FLOAT) {
return blit_image_helper.ConvertD32ToR32(dst, src_view, up_scale, down_shift);
@@ -891,6 +1082,9 @@ void TextureCacheRuntime::ConvertImage(Framebuffer* dst, ImageView& dst_view, Im
return blit_image_helper.ConvertR16ToD16(dst, src_view, up_scale, down_shift);
}
break;
case PixelFormat::S8_UINT_D24_UNORM:
return blit_image_helper.ConvertABGR8ToD24S8(dst, src_view, up_scale, down_shift);
break;
case PixelFormat::D32_FLOAT:
if (src_view.format == PixelFormat::R32_FLOAT) {
return blit_image_helper.ConvertR32ToD32(dst, src_view, up_scale, down_shift);
@@ -1386,6 +1580,14 @@ VkImageView ImageView::StencilView() {
return *stencil_view;
}
VkImageView ImageView::ColorView() {
if (color_view) {
return *color_view;
}
color_view = MakeView(VK_FORMAT_R8G8B8A8_UNORM, VK_IMAGE_ASPECT_COLOR_BIT);
return *color_view;
}
VkImageView ImageView::StorageView(Shader::TextureType texture_type,
Shader::ImageFormat image_format) {
if (image_format == Shader::ImageFormat::Typeless) {

View File

@@ -61,6 +61,10 @@ public:
void CopyImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
bool ShouldReinterpret(Image& dst, Image& src);
void ReinterpretImage(Image& dst, Image& src, std::span<const VideoCommon::ImageCopy> copies);
void ConvertImage(Framebuffer* dst, ImageView& dst_view, ImageView& src_view, bool rescaled);
bool CanAccelerateImageUpload(Image&) const noexcept {
@@ -82,6 +86,8 @@ public:
return true;
}
[[nodiscard]] VkBuffer GetTemporaryBuffer(size_t needed_size);
const Device& device;
VKScheduler& scheduler;
MemoryAllocator& memory_allocator;
@@ -90,6 +96,10 @@ public:
ASTCDecoderPass& astc_decoder_pass;
RenderPassCache& render_pass_cache;
const Settings::ResolutionScalingInfo& resolution;
constexpr static size_t indexing_slots = 8 * sizeof(size_t);
std::array<vk::Buffer, indexing_slots> buffers{};
std::array<std::unique_ptr<MemoryCommit>, indexing_slots> buffer_commits{};
};
class Image : public VideoCommon::ImageBase {
@@ -174,6 +184,8 @@ public:
[[nodiscard]] VkImageView StencilView();
[[nodiscard]] VkImageView ColorView();
[[nodiscard]] VkImageView StorageView(Shader::TextureType texture_type,
Shader::ImageFormat image_format);
@@ -214,6 +226,7 @@ private:
std::unique_ptr<StorageViews> storage_views;
vk::ImageView depth_view;
vk::ImageView stencil_view;
vk::ImageView color_view;
VkImage image_handle = VK_NULL_HANDLE;
VkImageView render_target = VK_NULL_HANDLE;
VkSampleCountFlagBits samples = VK_SAMPLE_COUNT_1_BIT;
@@ -316,7 +329,6 @@ struct TextureCacheParams {
static constexpr bool FRAMEBUFFER_BLITS = false;
static constexpr bool HAS_EMULATED_COPIES = false;
static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
static constexpr bool HAS_PIXEL_FORMAT_CONVERSIONS = false;
using Runtime = Vulkan::TextureCacheRuntime;
using Image = Vulkan::Image;

View File

@@ -82,6 +82,8 @@ PixelFormat PixelFormatFromDepthFormat(Tegra::DepthFormat format) {
return PixelFormat::D32_FLOAT;
case Tegra::DepthFormat::D16_UNORM:
return PixelFormat::D16_UNORM;
case Tegra::DepthFormat::S8_UINT:
return PixelFormat::S8_UINT;
case Tegra::DepthFormat::D32_FLOAT_S8X24_UINT:
return PixelFormat::D32_FLOAT_S8_UINT;
default:
@@ -213,6 +215,11 @@ SurfaceType GetFormatType(PixelFormat pixel_format) {
return SurfaceType::Depth;
}
if (static_cast<std::size_t>(pixel_format) <
static_cast<std::size_t>(PixelFormat::MaxStencilFormat)) {
return SurfaceType::Stencil;
}
if (static_cast<std::size_t>(pixel_format) <
static_cast<std::size_t>(PixelFormat::MaxDepthStencilFormat)) {
return SurfaceType::DepthStencil;

View File

@@ -110,8 +110,12 @@ enum class PixelFormat {
MaxDepthFormat,
// Stencil formats
S8_UINT = MaxDepthFormat,
MaxStencilFormat,
// DepthStencil formats
D24_UNORM_S8_UINT = MaxDepthFormat,
D24_UNORM_S8_UINT = MaxStencilFormat,
S8_UINT_D24_UNORM,
D32_FLOAT_S8_UINT,
@@ -125,8 +129,9 @@ constexpr std::size_t MaxPixelFormat = static_cast<std::size_t>(PixelFormat::Max
enum class SurfaceType {
ColorTexture = 0,
Depth = 1,
DepthStencil = 2,
Invalid = 3,
Stencil = 2,
DepthStencil = 3,
Invalid = 4,
};
enum class SurfaceTarget {
@@ -229,6 +234,7 @@ constexpr std::array<u32, MaxPixelFormat> BLOCK_WIDTH_TABLE = {{
1, // E5B9G9R9_FLOAT
1, // D32_FLOAT
1, // D16_UNORM
1, // S8_UINT
1, // D24_UNORM_S8_UINT
1, // S8_UINT_D24_UNORM
1, // D32_FLOAT_S8_UINT
@@ -328,6 +334,7 @@ constexpr std::array<u32, MaxPixelFormat> BLOCK_HEIGHT_TABLE = {{
1, // E5B9G9R9_FLOAT
1, // D32_FLOAT
1, // D16_UNORM
1, // S8_UINT
1, // D24_UNORM_S8_UINT
1, // S8_UINT_D24_UNORM
1, // D32_FLOAT_S8_UINT
@@ -427,6 +434,7 @@ constexpr std::array<u32, MaxPixelFormat> BITS_PER_BLOCK_TABLE = {{
32, // E5B9G9R9_FLOAT
32, // D32_FLOAT
16, // D16_UNORM
8, // S8_UINT
32, // D24_UNORM_S8_UINT
32, // S8_UINT_D24_UNORM
64, // D32_FLOAT_S8_UINT

View File

@@ -194,6 +194,8 @@ struct fmt::formatter<VideoCore::Surface::PixelFormat> : fmt::formatter<fmt::str
return "D32_FLOAT";
case PixelFormat::D16_UNORM:
return "D16_UNORM";
case PixelFormat::S8_UINT:
return "S8_UINT";
case PixelFormat::D24_UNORM_S8_UINT:
return "D24_UNORM_S8_UINT";
case PixelFormat::S8_UINT_D24_UNORM:

View File

@@ -472,9 +472,10 @@ template <class P>
void TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
const Tegra::Engines::Fermi2D::Surface& src,
const Tegra::Engines::Fermi2D::Config& copy) {
const BlitImages images = GetBlitImages(dst, src);
const BlitImages images = GetBlitImages(dst, src, copy);
const ImageId dst_id = images.dst_id;
const ImageId src_id = images.src_id;
PrepareImage(src_id, false, false);
PrepareImage(dst_id, true, false);
@@ -758,14 +759,18 @@ ImageId TextureCache<P>::FindImage(const ImageInfo& info, GPUVAddr gpu_addr,
return ImageId{};
}
}
const bool broken_views = runtime.HasBrokenTextureViewFormats();
const bool broken_views =
runtime.HasBrokenTextureViewFormats() || True(options & RelaxedOptions::ForceBrokenViews);
const bool native_bgr = runtime.HasNativeBgr();
ImageId image_id;
const bool flexible_formats = True(options & RelaxedOptions::Format);
ImageId image_id{};
boost::container::small_vector<ImageId, 1> image_ids;
const auto lambda = [&](ImageId existing_image_id, ImageBase& existing_image) {
if (True(existing_image.flags & ImageFlagBits::Remapped)) {
return false;
}
if (info.type == ImageType::Linear || existing_image.info.type == ImageType::Linear) {
if (info.type == ImageType::Linear || existing_image.info.type == ImageType::Linear)
[[unlikely]] {
const bool strict_size = False(options & RelaxedOptions::Size) &&
True(existing_image.flags & ImageFlagBits::Strong);
const ImageInfo& existing = existing_image.info;
@@ -774,17 +779,27 @@ ImageId TextureCache<P>::FindImage(const ImageInfo& info, GPUVAddr gpu_addr,
IsPitchLinearSameSize(existing, info, strict_size) &&
IsViewCompatible(existing.format, info.format, broken_views, native_bgr)) {
image_id = existing_image_id;
return true;
image_ids.push_back(existing_image_id);
return !flexible_formats && existing.format == info.format;
}
} else if (IsSubresource(info, existing_image, gpu_addr, options, broken_views,
native_bgr)) {
image_id = existing_image_id;
return true;
image_ids.push_back(existing_image_id);
return !flexible_formats && existing_image.info.format == info.format;
}
return false;
};
ForEachImageInRegion(*cpu_addr, CalculateGuestSizeInBytes(info), lambda);
return image_id;
if (image_ids.size() <= 1) [[likely]] {
return image_id;
}
auto image_ids_compare = [this](ImageId a, ImageId b) {
auto& image_a = slot_images[a];
auto& image_b = slot_images[b];
return image_a.modification_tick < image_b.modification_tick;
};
return *std::ranges::max_element(image_ids, image_ids_compare);
}
template <class P>
@@ -1076,31 +1091,58 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
template <class P>
typename TextureCache<P>::BlitImages TextureCache<P>::GetBlitImages(
const Tegra::Engines::Fermi2D::Surface& dst, const Tegra::Engines::Fermi2D::Surface& src) {
static constexpr auto FIND_OPTIONS = RelaxedOptions::Format | RelaxedOptions::Samples;
const Tegra::Engines::Fermi2D::Surface& dst, const Tegra::Engines::Fermi2D::Surface& src,
const Tegra::Engines::Fermi2D::Config& copy) {
static constexpr auto FIND_OPTIONS = RelaxedOptions::Samples;
const GPUVAddr dst_addr = dst.Address();
const GPUVAddr src_addr = src.Address();
ImageInfo dst_info(dst);
ImageInfo src_info(src);
const bool can_be_depth_blit =
dst_info.format == src_info.format && copy.filter == Tegra::Engines::Fermi2D::Filter::Point;
ImageId dst_id;
ImageId src_id;
RelaxedOptions try_options = FIND_OPTIONS;
if (can_be_depth_blit) {
try_options |= RelaxedOptions::Format;
}
do {
has_deleted_images = false;
dst_id = FindImage(dst_info, dst_addr, FIND_OPTIONS);
src_id = FindImage(src_info, src_addr, FIND_OPTIONS);
const ImageBase* const dst_image = dst_id ? &slot_images[dst_id] : nullptr;
src_id = FindImage(src_info, src_addr, try_options);
dst_id = FindImage(dst_info, dst_addr, try_options);
const ImageBase* const src_image = src_id ? &slot_images[src_id] : nullptr;
DeduceBlitImages(dst_info, src_info, dst_image, src_image);
if (GetFormatType(dst_info.format) != GetFormatType(src_info.format)) {
continue;
if (src_image && src_image->info.num_samples > 1) {
RelaxedOptions find_options{FIND_OPTIONS | RelaxedOptions::ForceBrokenViews};
src_id = FindOrInsertImage(src_info, src_addr, find_options);
dst_id = FindOrInsertImage(dst_info, dst_addr, find_options);
if (has_deleted_images) {
continue;
}
break;
}
if (!dst_id) {
dst_id = InsertImage(dst_info, dst_addr, RelaxedOptions{});
if (can_be_depth_blit) {
const ImageBase* const dst_image = src_id ? &slot_images[src_id] : nullptr;
DeduceBlitImages(dst_info, src_info, dst_image, src_image);
if (GetFormatType(dst_info.format) != GetFormatType(src_info.format)) {
continue;
}
}
if (!src_id) {
src_id = InsertImage(src_info, src_addr, RelaxedOptions{});
}
if (!dst_id) {
dst_id = InsertImage(dst_info, dst_addr, RelaxedOptions{});
}
} while (has_deleted_images);
if (GetFormatType(dst_info.format) != SurfaceType::ColorTexture) {
// Make sure the images are depth and/or stencil textures.
do {
has_deleted_images = false;
src_id = FindOrInsertImage(src_info, src_addr, RelaxedOptions{});
dst_id = FindOrInsertImage(dst_info, dst_addr, RelaxedOptions{});
} while (has_deleted_images);
}
return BlitImages{
.dst_id = dst_id,
.src_id = src_id,
@@ -1157,7 +1199,14 @@ template <class P>
ImageViewId TextureCache<P>::FindRenderTargetView(const ImageInfo& info, GPUVAddr gpu_addr,
bool is_clear) {
const auto options = is_clear ? RelaxedOptions::Samples : RelaxedOptions{};
const ImageId image_id = FindOrInsertImage(info, gpu_addr, options);
ImageId image_id{};
bool delete_state = has_deleted_images;
do {
has_deleted_images = false;
image_id = FindOrInsertImage(info, gpu_addr, options);
delete_state |= has_deleted_images;
} while (has_deleted_images);
has_deleted_images = delete_state;
if (!image_id) {
return NULL_IMAGE_VIEW_ID;
}
@@ -1759,8 +1808,8 @@ void TextureCache<P>::CopyImage(ImageId dst_id, ImageId src_id, std::vector<Imag
}
UNIMPLEMENTED_IF(dst.info.type != ImageType::e2D);
UNIMPLEMENTED_IF(src.info.type != ImageType::e2D);
if constexpr (HAS_PIXEL_FORMAT_CONVERSIONS) {
return runtime.ConvertImage(dst, src, copies);
if (runtime.ShouldReinterpret(dst, src)) {
return runtime.ReinterpretImage(dst, src, copies);
}
for (const ImageCopy& copy : copies) {
UNIMPLEMENTED_IF(copy.dst_subresource.num_layers != 1);
@@ -1780,7 +1829,13 @@ void TextureCache<P>::CopyImage(ImageId dst_id, ImageId src_id, std::vector<Imag
const SubresourceExtent src_extent{.levels = 1, .layers = 1};
const SubresourceRange dst_range{.base = dst_base, .extent = dst_extent};
const SubresourceRange src_range{.base = src_base, .extent = src_extent};
const ImageViewInfo dst_view_info(ImageViewType::e2D, dst.info.format, dst_range);
PixelFormat dst_format = dst.info.format;
if (GetFormatType(src.info.format) == SurfaceType::DepthStencil &&
GetFormatType(dst_format) == SurfaceType::ColorTexture &&
BytesPerBlock(dst_format) == 4) {
dst_format = PixelFormat::A8B8G8R8_UNORM;
}
const ImageViewInfo dst_view_info(ImageViewType::e2D, dst_format, dst_range);
const ImageViewInfo src_view_info(ImageViewType::e2D, src.info.format, src_range);
const auto [dst_framebuffer_id, dst_view_id] = RenderTargetFromImage(dst_id, dst_view_info);
Framebuffer* const dst_framebuffer = &slot_framebuffers[dst_framebuffer_id];

View File

@@ -59,8 +59,6 @@ class TextureCache {
static constexpr bool HAS_EMULATED_COPIES = P::HAS_EMULATED_COPIES;
/// True when the API can provide info about the memory of the device.
static constexpr bool HAS_DEVICE_MEMORY_INFO = P::HAS_DEVICE_MEMORY_INFO;
/// True when the API provides utilities for pixel format conversions.
static constexpr bool HAS_PIXEL_FORMAT_CONVERSIONS = P::HAS_PIXEL_FORMAT_CONVERSIONS;
static constexpr u64 DEFAULT_EXPECTED_MEMORY = 1_GiB;
static constexpr u64 DEFAULT_CRITICAL_MEMORY = 2_GiB;
@@ -254,7 +252,8 @@ private:
/// Return a blit image pair from the given guest blit parameters
[[nodiscard]] BlitImages GetBlitImages(const Tegra::Engines::Fermi2D::Surface& dst,
const Tegra::Engines::Fermi2D::Surface& src);
const Tegra::Engines::Fermi2D::Surface& src,
const Tegra::Engines::Fermi2D::Config& copy);
/// Find or create a sampler from a guest descriptor sampler
[[nodiscard]] SamplerId FindSampler(const TSCEntry& config);

View File

@@ -54,6 +54,7 @@ enum class RelaxedOptions : u32 {
Size = 1 << 0,
Format = 1 << 1,
Samples = 1 << 2,
ForceBrokenViews = 1 << 3,
};
DECLARE_ENUM_FLAG_OPERATORS(RelaxedOptions)

View File

@@ -1151,6 +1151,7 @@ bool IsSubresource(const ImageInfo& candidate, const ImageBase& image, GPUVAddr
void DeduceBlitImages(ImageInfo& dst_info, ImageInfo& src_info, const ImageBase* dst,
const ImageBase* src) {
const auto original_dst_format = dst_info.format;
if (src && GetFormatType(src->info.format) != SurfaceType::ColorTexture) {
src_info.format = src->info.format;
}
@@ -1161,7 +1162,13 @@ void DeduceBlitImages(ImageInfo& dst_info, ImageInfo& src_info, const ImageBase*
dst_info.format = src->info.format;
}
if (dst && GetFormatType(dst->info.format) != SurfaceType::ColorTexture) {
src_info.format = dst->info.format;
if (src) {
if (GetFormatType(src->info.format) == SurfaceType::ColorTexture) {
dst_info.format = original_dst_format;
}
} else {
src_info.format = dst->info.format;
}
}
}

View File

@@ -21,6 +21,13 @@
namespace Vulkan {
namespace {
namespace Alternatives {
constexpr std::array STENCIL8_UINT{
VK_FORMAT_D16_UNORM_S8_UINT,
VK_FORMAT_D24_UNORM_S8_UINT,
VK_FORMAT_D32_SFLOAT_S8_UINT,
VK_FORMAT_UNDEFINED,
};
constexpr std::array DEPTH24_UNORM_STENCIL8_UINT{
VK_FORMAT_D32_SFLOAT_S8_UINT,
VK_FORMAT_D16_UNORM_S8_UINT,
@@ -74,6 +81,8 @@ void SetNext(void**& next, T& data) {
constexpr const VkFormat* GetFormatAlternatives(VkFormat format) {
switch (format) {
case VK_FORMAT_S8_UINT:
return Alternatives::STENCIL8_UINT.data();
case VK_FORMAT_D24_UNORM_S8_UINT:
return Alternatives::DEPTH24_UNORM_STENCIL8_UINT.data();
case VK_FORMAT_D16_UNORM_S8_UINT:
@@ -145,6 +154,7 @@ std::unordered_map<VkFormat, VkFormatProperties> GetFormatProperties(vk::Physica
VK_FORMAT_R4G4B4A4_UNORM_PACK16,
VK_FORMAT_D32_SFLOAT,
VK_FORMAT_D16_UNORM,
VK_FORMAT_S8_UINT,
VK_FORMAT_D16_UNORM_S8_UINT,
VK_FORMAT_D24_UNORM_S8_UINT,
VK_FORMAT_D32_SFLOAT_S8_UINT,

View File

@@ -3300,9 +3300,9 @@ void GMainWindow::OnReinitializeKeys(ReinitializeKeyBehavior behavior) {
if (!errors.isEmpty()) {
QMessageBox::warning(
this, tr("Derivation Components Missing"),
tr("Components are missing that may hinder key derivation from completing. "
tr("Encryption keys are missing. "
"<br>Please follow <a href='https://yuzu-emu.org/help/quickstart/'>the yuzu "
"quickstart guide</a> to get all your keys and "
"quickstart guide</a> to get all your keys, firmware and "
"games.<br><br><small>(%1)</small>")
.arg(errors));
}