Compare commits

..

1 Commits

Author SHA1 Message Date
lizzie 424705d193 [opengl] prevent area from crashing nvlogl64.dll
Signed-off-by: lizzie <lizzie@eden-emu.dev>
2026-08-23 06:32:39 +02:00
31 changed files with 666 additions and 1644 deletions
+47 -3
View File
@@ -17,12 +17,56 @@
namespace Kernel::Svc {
constexpr auto MAX_MSG_TIME = std::chrono::milliseconds(250);
const auto MAX_MSG_SIZE = 0x1000;
/// Used to output a message on a debug hardware unit - does nothing on a retail unit
Result OutputDebugString(Core::System& system, u64 address, u64 len) {
static struct DebugFlusher {
std::string msg_buffer;
std::mutex msg_mutex;
std::condition_variable msg_cv;
std::chrono::steady_clock::time_point last_msg_time;
std::optional<std::jthread> thread;
} flusher_data;
R_SUCCEED_IF(len == 0);
std::string msg_buffer(len);
GetCurrentMemory(system.Kernel()).ReadBlock(address, msg_buffer.data(), len);
LOG_INFO(Debug_Emulated, "{}", msg_buffer);
// Only start the thread the very first time this function is called
if (!flusher_data.thread) {
flusher_data.thread.emplace([](std::stop_token stop_token) {
while (!stop_token.stop_requested()) {
std::unique_lock lock(flusher_data.msg_mutex);
flusher_data.msg_cv.wait(lock, [&stop_token] {
return !flusher_data.msg_buffer.empty() || stop_token.stop_requested();
});
if (stop_token.stop_requested() && flusher_data.msg_buffer.empty())
break;
auto timeout = flusher_data.last_msg_time + MAX_MSG_TIME;
bool woke_early = flusher_data.msg_cv.wait_until(lock, timeout, [&stop_token] {
return flusher_data.msg_buffer.size() >= MAX_MSG_SIZE || stop_token.stop_requested();
});
if (!woke_early || flusher_data.msg_buffer.size() >= MAX_MSG_SIZE || stop_token.stop_requested()) {
if (!flusher_data.msg_buffer.empty()) {
// Remove trailing newline as LOG_INFO adds that anyways
if (flusher_data.msg_buffer.back() == '\n')
flusher_data.msg_buffer.pop_back();
LOG_INFO(Debug_Emulated, "\n{}", flusher_data.msg_buffer);
flusher_data.msg_buffer.clear();
}
if (stop_token.stop_requested()) break;
}
}
flusher_data.msg_cv.notify_all();
});
}
{
std::lock_guard lock(flusher_data.msg_mutex);
const auto old_size = flusher_data.msg_buffer.size();
flusher_data.msg_buffer.resize(old_size + len);
GetCurrentMemory(system.Kernel()).ReadBlock(address, flusher_data.msg_buffer.data() + old_size, len);
flusher_data.last_msg_time = std::chrono::steady_clock::now();
}
flusher_data.msg_cv.notify_one();
R_SUCCEED();
}
@@ -18,7 +18,6 @@ set(SHADER_FILES
${CMAKE_CURRENT_SOURCE_DIR}/blit_color_float.frag
${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d.comp
${CMAKE_CURRENT_SOURCE_DIR}/blit_color_msaa.frag
${CMAKE_CURRENT_SOURCE_DIR}/blit_depth.frag
${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_msaa.frag
${CMAKE_CURRENT_SOURCE_DIR}/blit_depth_stencil_msaa.frag
${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_3d.comp
@@ -31,12 +30,8 @@ set(SHADER_FILES
${CMAKE_CURRENT_SOURCE_DIR}/convert_float_to_depth.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa.comp
${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_depth.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_msaa_to_non_msaa_depth_stencil.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.comp
${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_non_msaa_to_msaa_depth_stencil.frag
${CMAKE_CURRENT_SOURCE_DIR}/convert_s8d24_to_abgr8.frag
${CMAKE_CURRENT_SOURCE_DIR}/full_screen_triangle.vert
${CMAKE_CURRENT_SOURCE_DIR}/fxaa.frag
@@ -168,44 +163,6 @@ foreach(SOURCE_FILE IN ITEMS ${SHADER_FILES})
endif()
endforeach()
# Integer variants of the MSAA conversion shaders. They only differ from the float
# source in the sampler and output types, so they are generated from it via defines.
set(SHADER_TYPE_VARIANTS
"convert_msaa_to_non_msaa.frag|sint|isampler2DMS|ivec4"
"convert_msaa_to_non_msaa.frag|uint|usampler2DMS|uvec4"
"convert_non_msaa_to_msaa.frag|sint|isampler2D|ivec4"
"convert_non_msaa_to_msaa.frag|uint|usampler2D|uvec4"
)
foreach(VARIANT IN ITEMS ${SHADER_TYPE_VARIANTS})
string(REPLACE "|" ";" VARIANT_PARTS ${VARIANT})
list(GET VARIANT_PARTS 0 VARIANT_FILENAME)
list(GET VARIANT_PARTS 1 VARIANT_SUFFIX)
list(GET VARIANT_PARTS 2 VARIANT_SAMPLER)
list(GET VARIANT_PARTS 3 VARIANT_TEXEL)
set(VARIANT_SOURCE ${CMAKE_CURRENT_SOURCE_DIR}/${VARIANT_FILENAME})
get_filename_component(VARIANT_STEM ${VARIANT_FILENAME} NAME_WE)
get_filename_component(VARIANT_EXT ${VARIANT_FILENAME} EXT)
string(REPLACE "." "" VARIANT_EXT ${VARIANT_EXT})
set(VARIANT_NAME ${VARIANT_STEM}_${VARIANT_SUFFIX}_${VARIANT_EXT})
string(TOUPPER ${VARIANT_NAME}_SPV VARIANT_VARIABLE_NAME)
set(VARIANT_HEADER_FILE ${SHADER_DIR}/${VARIANT_NAME}_spv.h)
add_custom_command(
OUTPUT
${VARIANT_HEADER_FILE}
COMMAND
${GLSLANGVALIDATOR} -V ${QUIET_FLAG} -I"${FIDELITYFX_INCLUDE_DIR}" ${GLSL_FLAGS}
-DSAMPLER_TYPE=${VARIANT_SAMPLER} -DTEXEL_TYPE=${VARIANT_TEXEL}
--variable-name ${VARIANT_VARIABLE_NAME} -o ${VARIANT_HEADER_FILE} ${VARIANT_SOURCE}
--target-env ${SPIR_V_VERSION}
MAIN_DEPENDENCY
${VARIANT_SOURCE}
)
set(SHADER_HEADERS ${SHADER_HEADERS} ${VARIANT_HEADER_FILE})
endforeach()
foreach(FILEPATH IN ITEMS ${FIDELITYFX_FILES})
get_filename_component(FILENAME ${FILEPATH} NAME)
string(REPLACE "." "_" HEADER_NAME ${FILENAME})
@@ -1,12 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450 core
layout(binding = 0) uniform sampler2D depth_tex;
layout(location = 0) in vec2 texcoord;
void main() {
gl_FragDepth = textureLod(depth_tex, texcoord, 0).r;
}
@@ -8,5 +8,5 @@ layout(binding = 0) uniform sampler2DMS depth_tex;
layout(location = 0) in vec2 texcoord;
void main() {
gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), gl_SampleID).r;
gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), 0).r;
}
@@ -10,6 +10,6 @@ layout(binding = 1) uniform usampler2DMS stencil_tex;
layout(location = 0) in vec2 texcoord;
void main() {
gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), gl_SampleID).r;
gl_FragStencilRefARB = int(texelFetch(stencil_tex, ivec2(texcoord), gl_SampleID).r);
gl_FragDepth = texelFetch(depth_tex, ivec2(texcoord), 0).r;
gl_FragStencilRefARB = int(texelFetch(stencil_tex, ivec2(texcoord), 0).r);
}
@@ -3,14 +3,7 @@
#version 450 core
#ifndef SAMPLER_TYPE
#define SAMPLER_TYPE sampler2DMS
#endif
#ifndef TEXEL_TYPE
#define TEXEL_TYPE vec4
#endif
layout(binding = 0) uniform SAMPLER_TYPE msaa_in;
layout(binding = 0) uniform sampler2DMS msaa_in;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
@@ -18,7 +11,7 @@ layout(push_constant) uniform PushConstants {
ivec2 scale;
};
layout(location = 0) out TEXEL_TYPE frag_color;
layout(location = 0) out vec4 frag_color;
void main() {
const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
@@ -1,20 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450 core
layout(binding = 0) uniform sampler2DMS msaa_in;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
ivec2 src_offset;
ivec2 scale;
};
void main() {
const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
const ivec2 msaa_coord = coord / scale;
const ivec2 sample_offset = coord % scale;
const int sample_id = sample_offset.x + scale.x * sample_offset.y;
gl_FragDepth = texelFetch(msaa_in, msaa_coord, sample_id).r;
}
@@ -1,23 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450 core
#extension GL_ARB_shader_stencil_export : require
layout(binding = 0) uniform sampler2DMS depth_tex;
layout(binding = 1) uniform usampler2DMS stencil_tex;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
ivec2 src_offset;
ivec2 scale;
};
void main() {
const ivec2 coord = ivec2(gl_FragCoord.xy) - dst_offset + src_offset;
const ivec2 msaa_coord = coord / scale;
const ivec2 sample_offset = coord % scale;
const int sample_id = sample_offset.x + scale.x * sample_offset.y;
gl_FragDepth = texelFetch(depth_tex, msaa_coord, sample_id).r;
gl_FragStencilRefARB = int(texelFetch(stencil_tex, msaa_coord, sample_id).r);
}
@@ -3,14 +3,7 @@
#version 450 core
#ifndef SAMPLER_TYPE
#define SAMPLER_TYPE sampler2D
#endif
#ifndef TEXEL_TYPE
#define TEXEL_TYPE vec4
#endif
layout(binding = 0) uniform SAMPLER_TYPE img_in;
layout(binding = 0) uniform sampler2D img_in;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
@@ -18,7 +11,7 @@ layout(push_constant) uniform PushConstants {
ivec2 scale;
};
layout(location = 0) out TEXEL_TYPE frag_color;
layout(location = 0) out vec4 frag_color;
void main() {
const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
@@ -1,19 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450 core
layout(binding = 0) uniform sampler2D img_in;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
ivec2 src_offset;
ivec2 scale;
};
void main() {
const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
gl_FragDepth = texelFetch(img_in, coord, 0).r;
}
@@ -1,22 +0,0 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
#version 450 core
#extension GL_ARB_shader_stencil_export : require
layout(binding = 0) uniform sampler2D depth_tex;
layout(binding = 1) uniform usampler2D stencil_tex;
layout(push_constant) uniform PushConstants {
ivec2 dst_offset;
ivec2 src_offset;
ivec2 scale;
};
void main() {
const ivec2 msaa_coord = ivec2(gl_FragCoord.xy) - dst_offset;
const ivec2 sample_offset = ivec2(gl_SampleID % scale.x, gl_SampleID / scale.x);
const ivec2 coord = msaa_coord * scale + sample_offset + src_offset;
gl_FragDepth = texelFetch(depth_tex, coord, 0).r;
gl_FragStencilRefARB = int(texelFetch(stencil_tex, coord, 0).r);
}
+48 -64
View File
@@ -5,7 +5,6 @@ layout(location = 0) out vec4 color;
layout(binding = 0) uniform sampler2D color_texture;
#ifdef VULKAN
struct ScreenRectVertex {
vec2 position;
vec2 tex_coord;
@@ -14,94 +13,79 @@ layout (push_constant) uniform PushConstants {
mat4 modelview_matrix;
ScreenRectVertex vertices[4];
};
#else // OpenGL
layout(location = 1) uniform uvec2 screen_size;
#endif
/***** Area Sampling *****/
// By Sam Belliveau and Filippo Tarpini. Public Domain license.
// Effectively a more accurate sharp bilinear filter when upscaling,
// that also works as a mathematically perfect downscale filter.
// https://entropymine.com/imageworsener/pixelmixing/
// https://github.com/obsproject/obs-studio/pull/1715
// https://legacy.imagemagick.org/Usage/filter/
vec4 AreaSampling(sampler2D textureSampler, vec2 texCoords, vec2 source_size, vec2 target_size) {
// Determine the sizes of the source and target images.
vec2 inverted_target_size = vec2(1.0) / target_size;
// Determine the range of the source image that the target pixel will cover.
vec2 range = source_size * inverted_target_size;
vec2 beg = (texCoords.xy * source_size) - (range * 0.5);
vec2 end = beg + range;
vec4 AreaSampling(sampler2D s, vec2 tc, vec4 trans_bounds) {
// Compute the top-left and bottom-right corners of the pixel box.
ivec2 f_beg = ivec2(floor(beg));
ivec2 f_end = ivec2(floor(end));
ivec4 b = ivec4(floor(trans_bounds));
// Compute how much of the start and end pixels are covered horizontally & vertically.
float area_w = 1.0 - fract(beg.x);
float area_n = 1.0 - fract(beg.y);
float area_e = fract(end.x);
float area_s = fract(end.y);
// W,N,E,S
vec4 kb = vec4(1.0f - fract(trans_bounds.xy), fract(trans_bounds.zw));
// Compute the areas of the corner pixels in the pixel box.
float area_nw = area_n * area_w;
float area_ne = area_n * area_e;
float area_sw = area_s * area_w;
float area_se = area_s * area_e;
// Initialize the color accumulator.
vec4 avg_color = vec4(0.0, 0.0, 0.0, 0.0);
// Accumulate corner pixels.
avg_color += area_nw * texelFetch(textureSampler, ivec2(f_beg.x, f_beg.y), 0);
avg_color += area_ne * texelFetch(textureSampler, ivec2(f_end.x, f_beg.y), 0);
avg_color += area_sw * texelFetch(textureSampler, ivec2(f_beg.x, f_end.y), 0);
avg_color += area_se * texelFetch(textureSampler, ivec2(f_end.x, f_end.y), 0);
// NW,NE,SW,SE
vec4 kc = kb.yyww * kb.xzxz;
// Accumulate corner pixels by forming a corner matrix.
vec4 r = mat4x4(
texelFetch(s, ivec2(b.xy), 0),
texelFetch(s, ivec2(b.zy), 0),
texelFetch(s, ivec2(b.xw), 0),
texelFetch(s, ivec2(b.zw), 0)
) * kc;
// Determine the size of the pixel box.
int x_range = int(f_end.x - f_beg.x - 0.5);
int y_range = int(f_end.y - f_beg.y - 0.5);
ivec2 q = clamp(ivec2(
int(b.z - b.x - 0.5f),
int(b.w - b.y - 0.5f)
), ivec2(-16), ivec2(16));
vec2 qf = vec2(q);
// Accumulate top and bottom edge pixels.
for (int x = f_beg.x + 1; x <= f_beg.x + x_range; ++x) {
avg_color += area_n * texelFetch(textureSampler, ivec2(x, f_beg.y), 0);
avg_color += area_s * texelFetch(textureSampler, ivec2(x, f_end.y), 0);
for (int x = 0; x < q.x; ++x) {
r += kb.y * texelFetch(s, ivec2(x + b.x + 1, b.y), 0);
r += kb.w * texelFetch(s, ivec2(x + b.x + 1, b.w), 0);
}
// Accumulate left and right edge pixels and all the pixels in between.
for (int y = f_beg.y + 1; y <= f_beg.y + y_range; ++y) {
avg_color += area_w * texelFetch(textureSampler, ivec2(f_beg.x, y), 0);
avg_color += area_e * texelFetch(textureSampler, ivec2(f_end.x, y), 0);
for (int x = f_beg.x + 1; x <= f_beg.x + x_range; ++x) {
avg_color += texelFetch(textureSampler, ivec2(x, y), 0);
for (int y = 0; y < q.y; ++y) {
r += kb.x * texelFetch(s, ivec2(b.x, y + b.y + 1), 0);
r += kb.z * texelFetch(s, ivec2(b.z, y + b.y + 1), 0);
for (int x = 0; x < q.x; ++x) {
r += texelFetch(s, ivec2(x, y) + b.xy + 1, 0);
}
}
// Compute the area of the pixel box that was sampled.
float area_corners = area_nw + area_ne + area_sw + area_se;
float area_edges = float(x_range) * (area_n + area_s) + float(y_range) * (area_w + area_e);
float area_center = float(x_range) * float(y_range);
// Return the normalized average color.
return avg_color / (area_corners + area_edges + area_center);
return r / (
kc.x + kc.y + kc.z + kc.w //corners
+ qf.x * (kb.y + kb.w) + qf.y * (kb.x + kb.z) //edges
+ qf.x * qf.y // center
);
}
void main() {
vec2 source_image_size = textureSize(color_texture, 0);
vec2 window_size;
#ifdef VULKAN
vec2 dst_size = vec2(
vertices[1].position.x - vertices[0].position.x,
vertices[2].position.y - vertices[0].position.y
);
#else // OpenGL
vec2 dst_size = screen_size;
#endif
#ifdef VULKAN
window_size.x = vertices[1].position.x - vertices[0].position.x;
window_size.y = vertices[2].position.y - vertices[0].position.y;
#else // OpenGL
window_size = screen_size;
#endif
color = AreaSampling(color_texture, frag_tex_coord, source_image_size, window_size);
vec2 src_size = textureSize(color_texture, 0);
// Determine the range of the source image that the target pixel will cover.
vec2 scale = src_size * (1.0f / dst_size);
color = AreaSampling(color_texture, frag_tex_coord, vec4(
(frag_tex_coord * src_size) - (scale * 0.5f),
// (scale * 0.5f) + scale
// {A / 2 + A} ==> {(A + 2A) / 2} ==> {3A / 2} ==> {A * (3/2)}
(frag_tex_coord * src_size) + scale * 0.5f
));
}
@@ -75,8 +75,6 @@ public:
void Finish();
void FlushDeferredClear() {}
StagingBufferMap UploadStagingBuffer(size_t size, bool deferred = false);
StagingBufferMap DownloadStagingBuffer(size_t size, bool deferred = false);
@@ -372,7 +370,6 @@ struct TextureCacheParams {
static constexpr bool HAS_EMULATED_COPIES = true;
static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = true;
static constexpr bool HAS_MSAA_DOWNLOADS = false;
using Runtime = OpenGL::TextureCacheRuntime;
using Image = OpenGL::Image;
+235 -553
View File
@@ -12,7 +12,6 @@
#include "common/settings.h"
#include "video_core/host_shaders/blit_color_float_frag_spv.h"
#include "video_core/host_shaders/blit_color_msaa_frag_spv.h"
#include "video_core/host_shaders/blit_depth_frag_spv.h"
#include "video_core/host_shaders/blit_depth_msaa_frag_spv.h"
#include "video_core/host_shaders/blit_depth_stencil_msaa_frag_spv.h"
#include "video_core/host_shaders/convert_abgr8_to_d24s8_frag_spv.h"
@@ -22,15 +21,7 @@
#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/convert_msaa_to_non_msaa_frag_spv.h"
#include "video_core/host_shaders/convert_msaa_to_non_msaa_depth_frag_spv.h"
#include "video_core/host_shaders/convert_msaa_to_non_msaa_depth_stencil_frag_spv.h"
#include "video_core/host_shaders/convert_msaa_to_non_msaa_sint_frag_spv.h"
#include "video_core/host_shaders/convert_msaa_to_non_msaa_uint_frag_spv.h"
#include "video_core/host_shaders/convert_non_msaa_to_msaa_frag_spv.h"
#include "video_core/host_shaders/convert_non_msaa_to_msaa_sint_frag_spv.h"
#include "video_core/host_shaders/convert_non_msaa_to_msaa_uint_frag_spv.h"
#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_frag_spv.h"
#include "video_core/host_shaders/convert_non_msaa_to_msaa_depth_stencil_frag_spv.h"
#include "video_core/host_shaders/convert_s8d24_to_abgr8_frag_spv.h"
#include "video_core/host_shaders/full_screen_triangle_vert_spv.h"
#include "video_core/host_shaders/vulkan_blit_depth_stencil_frag_spv.h"
@@ -527,19 +518,8 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
}
}
[[nodiscard]] MSAACopyFormatClass FormatClass(VideoCore::Surface::PixelFormat format) {
if (!VideoCore::Surface::IsPixelFormatInteger(format)) {
return MSAACopyFormatClass::Float;
}
if (VideoCore::Surface::IsPixelFormatSignedInteger(format)) {
return MSAACopyFormatClass::SignedInteger;
}
return MSAACopyFormatClass::UnsignedInteger;
}
[[nodiscard]] vk::ImageView MakeMSAACopyView(const vk::Device& device, VkImage image,
VkFormat format, u32 base_level, u32 base_layer,
VkImageAspectFlags aspect_mask) {
VkFormat format, u32 base_level) {
return device.CreateImageView(VkImageViewCreateInfo{
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.pNext = nullptr,
@@ -554,10 +534,10 @@ void RecordShaderReadBarrier(Scheduler& scheduler, const ImageView& image_view)
.a = VK_COMPONENT_SWIZZLE_IDENTITY,
},
.subresourceRange{
.aspectMask = aspect_mask,
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.baseMipLevel = base_level,
.levelCount = 1,
.baseArrayLayer = base_layer,
.baseArrayLayer = 0,
.layerCount = 1,
},
});
@@ -606,17 +586,12 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
msaa_copy_pipeline_layout(device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
one_texture_set_layout.address(),
PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
msaa_copy_depth_stencil_pipeline_layout(
device.GetLogical().CreatePipelineLayout(PipelineLayoutCreateInfo(
two_textures_set_layout.address(),
PUSH_CONSTANT_RANGE<VK_SHADER_STAGE_FRAGMENT_BIT, sizeof(MSAACopyPushConstants)>))),
full_screen_vert(BuildShader(device, FULL_SCREEN_TRIANGLE_VERT_SPV)),
blit_color_to_color_frag(BuildShader(device, BLIT_COLOR_FLOAT_FRAG_SPV)),
blit_color_msaa_frag(BuildShader(device, BLIT_COLOR_MSAA_FRAG_SPV)),
blit_depth_stencil_frag(device.IsExtShaderStencilExportSupported()
? BuildShader(device, VULKAN_BLIT_DEPTH_STENCIL_FRAG_SPV)
: vk::ShaderModule{}),
blit_depth_frag(BuildShader(device, BLIT_DEPTH_FRAG_SPV)),
blit_depth_msaa_frag(BuildShader(device, BLIT_DEPTH_MSAA_FRAG_SPV)),
blit_depth_stencil_msaa_frag(device.IsExtShaderStencilExportSupported()
? BuildShader(device, BLIT_DEPTH_STENCIL_MSAA_FRAG_SPV)
@@ -634,25 +609,7 @@ BlitImageHelper::BlitImageHelper(const Device& device_, Scheduler& scheduler_,
convert_d24s8_to_abgr8_frag(BuildShader(device, CONVERT_D24S8_TO_ABGR8_FRAG_SPV)),
convert_s8d24_to_abgr8_frag(BuildShader(device, CONVERT_S8D24_TO_ABGR8_FRAG_SPV)),
convert_msaa_to_non_msaa_frag(BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_FRAG_SPV)),
convert_msaa_to_non_msaa_sint_frag(
BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_SINT_FRAG_SPV)),
convert_msaa_to_non_msaa_uint_frag(
BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_UINT_FRAG_SPV)),
convert_msaa_to_non_msaa_depth_frag(
BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_DEPTH_FRAG_SPV)),
convert_msaa_to_non_msaa_depth_stencil_frag(
BuildShader(device, CONVERT_MSAA_TO_NON_MSAA_DEPTH_STENCIL_FRAG_SPV)),
convert_non_msaa_to_msaa_frag(BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_FRAG_SPV)),
convert_non_msaa_to_msaa_sint_frag(
BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_SINT_FRAG_SPV)),
convert_non_msaa_to_msaa_uint_frag(
BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_UINT_FRAG_SPV)),
convert_non_msaa_to_msaa_depth_frag(
BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_FRAG_SPV)),
convert_non_msaa_to_msaa_depth_stencil_frag(
device.IsExtShaderStencilExportSupported()
? BuildShader(device, CONVERT_NON_MSAA_TO_MSAA_DEPTH_STENCIL_FRAG_SPV)
: vk::ShaderModule{}),
linear_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_LINEAR>)),
nearest_sampler(device.GetLogical().CreateSampler(SAMPLER_CREATE_INFO<VK_FILTER_NEAREST>)) {}
@@ -667,13 +624,24 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, const ImageV
.renderpass = dst_framebuffer->RenderPass(),
.operation = operation,
};
VkSampler sampler = *nearest_sampler;
if (is_linear) {
sampler = *linear_sampler;
}
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region,
FindOrEmplaceColorPipeline(key), sampler,
src_image_view.Handle(Shader::TextureType::Color2D), VK_NULL_HANDLE, false);
const VkPipelineLayout layout = *one_texture_pipeline_layout;
const VkSampler sampler = is_linear ? *linear_sampler : *nearest_sampler;
const VkPipeline pipeline = FindOrEmplaceColorPipeline(key);
const VkImageView src_view = src_image_view.Handle(Shader::TextureType::Color2D);
RecordShaderReadBarrier(scheduler, src_image_view);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler,
src_view](vk::CommandBuffer cmdbuf) {
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
BindBlitState(cmdbuf, layout, dst_region, src_region);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView src_image_view,
@@ -702,29 +670,24 @@ void BlitImageHelper::BlitColor(const Framebuffer* dst_framebuffer, VkImageView
});
}
void BlitImageHelper::BlitImpl(const Framebuffer* dst_framebuffer,
const ImageView& src_image_view, const Region2D& dst_region,
const Region2D& src_region, VkPipeline pipeline, VkSampler sampler,
VkImageView src_view, VkImageView src_stencil_view,
bool blit_stencil) {
VkPipelineLayout layout = *one_texture_pipeline_layout;
if (blit_stencil) {
layout = *two_textures_pipeline_layout;
}
void BlitImageHelper::BlitColorMSAA(const Framebuffer* dst_framebuffer,
const ImageView& src_image_view, const Region2D& dst_region,
const Region2D& src_region) {
const BlitMSAAPipelineKey key{
.renderpass = dst_framebuffer->RenderPass(),
.samples = dst_framebuffer->Samples(),
};
const VkPipelineLayout layout = *one_texture_pipeline_layout;
const VkSampler sampler = *nearest_sampler;
const VkPipeline pipeline = FindOrEmplaceBlitColorMSAAPipeline(key);
const VkImageView src_view = src_image_view.Handle(Shader::TextureType::Color2D);
RecordShaderReadBarrier(scheduler, src_image_view);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler, src_view,
src_stencil_view, blit_stencil](vk::CommandBuffer cmdbuf) {
VkDescriptorSet descriptor_set = VK_NULL_HANDLE;
if (blit_stencil) {
descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_view,
src_stencil_view);
} else {
descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
}
scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler,
src_view](vk::CommandBuffer cmdbuf) {
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
@@ -734,56 +697,41 @@ void BlitImageHelper::BlitImpl(const Framebuffer* dst_framebuffer,
scheduler.InvalidateState();
}
void BlitImageHelper::BlitColorMSAA(const Framebuffer* dst_framebuffer,
const ImageView& src_image_view, const Region2D& dst_region,
const Region2D& src_region) {
const BlitMSAAPipelineKey key{
.renderpass = dst_framebuffer->RenderPass(),
.samples = dst_framebuffer->Samples(),
};
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region,
FindOrEmplaceBlitColorMSAAPipeline(key), *nearest_sampler,
src_image_view.Handle(Shader::TextureType::Color2D), VK_NULL_HANDLE, false);
}
void BlitImageHelper::BlitDepthStencilMSAA(const Framebuffer* dst_framebuffer,
ImageView& src_image_view, const Region2D& dst_region,
const Region2D& src_region) {
const bool blit_stencil =
dst_framebuffer->HasAspectStencilBit() && device.IsExtShaderStencilExportSupported();
const BlitMSAAPipelineKey key{
.renderpass = dst_framebuffer->RenderPass(),
.samples = dst_framebuffer->Samples(),
};
VkImageView src_stencil_view = VK_NULL_HANDLE;
if (blit_stencil) {
src_stencil_view = src_image_view.StencilView();
}
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region,
FindOrEmplaceBlitDepthStencilMSAAPipeline(key, blit_stencil), *nearest_sampler,
src_image_view.DepthView(), src_stencil_view, blit_stencil);
}
void BlitImageHelper::BlitDepth(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region) {
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region,
FindOrEmplaceBlitDepthPipeline(dst_framebuffer->RenderPass()), *nearest_sampler,
src_image_view.DepthView(), VK_NULL_HANDLE, false);
}
void BlitImageHelper::ResolveDepthStencil(const Framebuffer* dst_framebuffer,
ImageView& src_image_view, const Region2D& dst_region,
const Region2D& src_region) {
const bool resolve_stencil =
dst_framebuffer->HasAspectStencilBit() && device.IsExtShaderStencilExportSupported();
VkImageView src_stencil_view = VK_NULL_HANDLE;
if (resolve_stencil) {
src_stencil_view = src_image_view.StencilView();
}
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region,
FindOrEmplaceResolveDepthStencilPipeline(dst_framebuffer->RenderPass(),
resolve_stencil),
*nearest_sampler, src_image_view.DepthView(), src_stencil_view, resolve_stencil);
const VkPipeline pipeline =
FindOrEmplaceResolveDepthStencilPipeline(dst_framebuffer->RenderPass(), resolve_stencil);
const VkPipelineLayout layout =
resolve_stencil ? *two_textures_pipeline_layout : *one_texture_pipeline_layout;
const VkSampler sampler = *nearest_sampler;
const VkImageView src_depth_view = src_image_view.DepthView();
const VkImageView src_stencil_view =
resolve_stencil ? src_image_view.StencilView() : VK_NULL_HANDLE;
RecordShaderReadBarrier(scheduler, src_image_view);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([this, dst_region, src_region, pipeline, layout, sampler, src_depth_view,
src_stencil_view, resolve_stencil](vk::CommandBuffer cmdbuf) {
if (resolve_stencil) {
const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
src_stencil_view);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
} else {
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_depth_view);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
}
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
BindBlitState(cmdbuf, layout, dst_region, src_region);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
@@ -791,23 +739,35 @@ void BlitImageHelper::BlitDepthStencil(const Framebuffer* dst_framebuffer,
const Region2D& dst_region, const Region2D& src_region,
Tegra::Engines::Fermi2D::Filter filter,
Tegra::Engines::Fermi2D::Operation operation) {
if (!device.IsExtShaderStencilExportSupported()) {
return;
}
ASSERT(filter == Tegra::Engines::Fermi2D::Filter::Point);
ASSERT(operation == Tegra::Engines::Fermi2D::Operation::SrcCopy);
const bool blit_stencil = device.IsExtShaderStencilExportSupported();
const BlitImagePipelineKey key{
.renderpass = dst_framebuffer->RenderPass(),
.operation = operation,
};
VkPipeline pipeline{};
VkImageView src_stencil_view = VK_NULL_HANDLE;
if (blit_stencil) {
pipeline = FindOrEmplaceDepthStencilPipeline(key);
src_stencil_view = src_image_view.StencilView();
} else {
pipeline = FindOrEmplaceBlitDepthPipeline(key.renderpass);
}
BlitImpl(dst_framebuffer, src_image_view, dst_region, src_region, pipeline, *nearest_sampler,
src_image_view.DepthView(), src_stencil_view, blit_stencil);
const VkPipelineLayout layout = *two_textures_pipeline_layout;
const VkSampler sampler = *nearest_sampler;
const VkPipeline pipeline = FindOrEmplaceDepthStencilPipeline(key);
const VkImageView src_depth_view = src_image_view.DepthView();
const VkImageView src_stencil_view = src_image_view.StencilView();
RecordShaderReadBarrier(scheduler, src_image_view);
scheduler.RequestRenderpass(dst_framebuffer);
scheduler.Record([dst_region, src_region, pipeline, layout, sampler, src_depth_view,
src_stencil_view, this](vk::CommandBuffer cmdbuf) {
const VkDescriptorSet descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_depth_view,
src_stencil_view);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
BindBlitState(cmdbuf, layout, dst_region, src_region);
cmdbuf.Draw(3, 1, 0, 0);
});
scheduler.InvalidateState();
}
void BlitImageHelper::ConvertD32ToR32(const Framebuffer* dst_framebuffer,
@@ -922,183 +882,19 @@ void BlitImageHelper::ClearDepthStencil(const Framebuffer* dst_framebuffer, bool
scheduler.InvalidateState();
}
void BlitImageHelper::CopyMSAAImpl(VkRenderPass renderpass, VkPipeline pipeline,
VkPipelineLayout layout, VkImage dst_image,
VkFormat dst_vk_format, VkImage src_image,
VkFormat src_vk_format, s32 scale_x, s32 scale_y,
std::span<const VideoCommon::ImageCopy> copies,
const MSAACopyAspectInfo& aspect_info, bool copy_stencil) {
while (!msaa_copy_resources.empty() && scheduler.IsFree(msaa_copy_resources.front().tick)) {
msaa_copy_resources.pop_front();
}
const VkSampler sampler = *nearest_sampler;
for (const VideoCommon::ImageCopy& copy : copies) {
const s32 num_layers = (std::min)(copy.src_subresource.num_layers,
copy.dst_subresource.num_layers);
for (s32 layer = 0; layer < num_layers; ++layer) {
const u32 src_level = static_cast<u32>(copy.src_subresource.base_level);
const u32 src_layer = static_cast<u32>(copy.src_subresource.base_layer + layer);
vk::ImageView src_view =
MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format, src_level,
src_layer, aspect_info.src_view_aspect);
vk::ImageView src_stencil_view;
if (copy_stencil) {
src_stencil_view =
MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format, src_level,
src_layer, VK_IMAGE_ASPECT_STENCIL_BIT);
}
vk::ImageView dst_view =
MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
static_cast<u32>(copy.dst_subresource.base_level),
static_cast<u32>(copy.dst_subresource.base_layer + layer),
aspect_info.attachment_aspect);
const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
const VkRect2D render_area{
.offset = dst_offset,
.extent = dst_extent,
};
vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.renderPass = renderpass,
.attachmentCount = 1,
.pAttachments = dst_view.address(),
.width = static_cast<u32>(dst_offset.x) + dst_extent.width,
.height = static_cast<u32>(dst_offset.y) + dst_extent.height,
.layers = 1,
});
const MSAACopyPushConstants push_constants{
.dst_offset = {dst_offset.x, dst_offset.y},
.src_offset = {copy.src_offset.x, copy.src_offset.y},
.scale = {scale_x, scale_y},
};
VkImageView src_stencil_handle = VK_NULL_HANDLE;
if (copy_stencil) {
src_stencil_handle = *src_stencil_view;
}
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, pipeline, layout, sampler, renderpass,
framebuffer_handle = *framebuffer, src_view_handle = *src_view,
src_stencil_handle, src = src_image, dst = dst_image, render_area,
aspect_info, push_constants](vk::CommandBuffer cmdbuf) {
const VkImageSubresourceRange barrier_range{
.aspectMask = aspect_info.barrier_aspect,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
};
const std::array pre_barriers{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = aspect_info.pre_src_access,
.dstAccessMask = aspect_info.pre_src_dst_access,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src,
.subresourceRange = barrier_range,
},
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = aspect_info.pre_src_access,
.dstAccessMask = aspect_info.pre_dst_dst_access,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst,
.subresourceRange = barrier_range,
},
};
cmdbuf.PipelineBarrier(aspect_info.pre_src_stages, aspect_info.pre_dst_stages, 0,
nullptr, nullptr, pre_barriers);
const VkRenderPassBeginInfo renderpass_bi{
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
.pNext = nullptr,
.renderPass = renderpass,
.framebuffer = framebuffer_handle,
.renderArea = render_area,
.clearValueCount = 0,
.pClearValues = nullptr,
};
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
VkDescriptorSet descriptor_set = VK_NULL_HANDLE;
if (src_stencil_handle != VK_NULL_HANDLE) {
descriptor_set = two_textures_descriptor_allocator.Commit();
UpdateTwoTexturesDescriptorSet(device, descriptor_set, sampler, src_view_handle,
src_stencil_handle);
} else {
descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
}
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
const VkViewport viewport{
.x = static_cast<float>(render_area.offset.x),
.y = static_cast<float>(render_area.offset.y),
.width = static_cast<float>(render_area.extent.width),
.height = static_cast<float>(render_area.extent.height),
.minDepth = 0.0f,
.maxDepth = 1.0f,
};
cmdbuf.SetViewport(0, viewport);
cmdbuf.SetScissor(0, render_area);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
cmdbuf.Draw(3, 1, 0, 0);
cmdbuf.EndRenderPass();
const VkImageMemoryBarrier post_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = aspect_info.post_src_access,
.dstAccessMask = aspect_info.post_dst_access,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst,
.subresourceRange = barrier_range,
};
cmdbuf.PipelineBarrier(aspect_info.post_src_stages, aspect_info.post_dst_stages, 0,
post_barrier);
});
msaa_copy_resources.push_back(MSAACopyResources{
.tick = scheduler.CurrentTick(),
.src_view = std::move(src_view),
.dst_view = std::move(dst_view),
.framebuffer = std::move(framebuffer),
});
if (copy_stencil) {
msaa_copy_resources.push_back(MSAACopyResources{
.tick = scheduler.CurrentTick(),
.src_view = std::move(src_stencil_view),
.dst_view = vk::ImageView{},
.framebuffer = vk::Framebuffer{},
});
}
}
}
scheduler.InvalidateState();
}
void BlitImageHelper::CopyMSAA(RenderPassCache& render_pass_cache, VkImage dst_image,
VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
VideoCore::Surface::PixelFormat src_format, u32 num_samples,
std::span<const VideoCommon::ImageCopy> copies,
bool msaa_to_non_msaa) {
while (!msaa_copy_resources.empty() && scheduler.IsFree(msaa_copy_resources.front().tick)) {
msaa_copy_resources.pop_front();
}
const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(static_cast<int>(num_samples));
const s32 scale_x = 1 << samples_x;
const s32 scale_y = 1 << samples_y;
VkSampleCountFlagBits samples = SampleCountFlag(num_samples);
if (msaa_to_non_msaa) {
samples = VK_SAMPLE_COUNT_1_BIT;
}
const VkSampleCountFlagBits samples =
msaa_to_non_msaa ? VK_SAMPLE_COUNT_1_BIT : SampleCountFlag(num_samples);
RenderPassKey renderpass_key{};
renderpass_key.color_formats.fill(VideoCore::Surface::PixelFormat::Invalid);
renderpass_key.color_formats[0] = dst_format;
@@ -1109,35 +905,149 @@ void BlitImageHelper::CopyMSAA(RenderPassCache& render_pass_cache, VkImage dst_i
.renderpass = renderpass,
.samples = samples,
.msaa_to_non_msaa = msaa_to_non_msaa,
.format_class = FormatClass(dst_format),
};
const MSAACopyAspectInfo aspect_info{
.src_view_aspect = VK_IMAGE_ASPECT_COLOR_BIT,
.attachment_aspect = VK_IMAGE_ASPECT_COLOR_BIT,
.barrier_aspect = VK_IMAGE_ASPECT_COLOR_BIT,
.pre_src_access = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT | VK_ACCESS_SHADER_WRITE_BIT |
VK_ACCESS_TRANSFER_WRITE_BIT,
.pre_src_dst_access = VK_ACCESS_SHADER_READ_BIT,
.pre_dst_dst_access =
VK_ACCESS_COLOR_ATTACHMENT_READ_BIT | VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
.pre_src_stages = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT,
.pre_dst_stages =
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
.post_src_access = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
.post_dst_access = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT,
.post_src_stages = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
.post_dst_stages = VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT | VK_PIPELINE_STAGE_TRANSFER_BIT,
};
const VkPipeline pipeline = FindOrEmplaceMSAACopyPipeline(key);
const VkPipelineLayout layout = *msaa_copy_pipeline_layout;
const VkSampler sampler = *nearest_sampler;
const VkFormat src_vk_format =
MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, src_format).format;
const VkFormat dst_vk_format =
MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, dst_format).format;
CopyMSAAImpl(renderpass, FindOrEmplaceMSAACopyPipeline(key), *msaa_copy_pipeline_layout,
dst_image, dst_vk_format, src_image, src_vk_format, scale_x, scale_y, copies,
aspect_info, false);
for (const VideoCommon::ImageCopy& copy : copies) {
ASSERT(copy.src_subresource.base_layer == 0);
ASSERT(copy.src_subresource.num_layers == 1);
ASSERT(copy.dst_subresource.base_layer == 0);
ASSERT(copy.dst_subresource.num_layers == 1);
vk::ImageView src_view =
MakeMSAACopyView(device.GetLogical(), src_image, src_vk_format,
static_cast<u32>(copy.src_subresource.base_level));
vk::ImageView dst_view =
MakeMSAACopyView(device.GetLogical(), dst_image, dst_vk_format,
static_cast<u32>(copy.dst_subresource.base_level));
const VkOffset2D dst_offset{copy.dst_offset.x, copy.dst_offset.y};
const VkExtent2D dst_extent{copy.extent.width, copy.extent.height};
const VkRect2D render_area{
.offset = dst_offset,
.extent = dst_extent,
};
vk::Framebuffer framebuffer = device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
.sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.renderPass = renderpass,
.attachmentCount = 1,
.pAttachments = dst_view.address(),
.width = static_cast<u32>(dst_offset.x) + dst_extent.width,
.height = static_cast<u32>(dst_offset.y) + dst_extent.height,
.layers = 1,
});
const MSAACopyPushConstants push_constants{
.dst_offset = {dst_offset.x, dst_offset.y},
.src_offset = {copy.src_offset.x, copy.src_offset.y},
.scale = {scale_x, scale_y},
};
scheduler.RequestOutsideRenderPassOperationContext();
scheduler.Record([this, pipeline, layout, sampler, renderpass,
framebuffer_handle = *framebuffer, src_view_handle = *src_view,
src = src_image, dst = dst_image, render_area,
push_constants](vk::CommandBuffer cmdbuf) {
constexpr VkImageSubresourceRange color_range{
.aspectMask = VK_IMAGE_ASPECT_COLOR_BIT,
.baseMipLevel = 0,
.levelCount = VK_REMAINING_MIP_LEVELS,
.baseArrayLayer = 0,
.layerCount = VK_REMAINING_ARRAY_LAYERS,
};
const std::array pre_barriers{
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = src,
.subresourceRange = color_range,
},
VkImageMemoryBarrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT |
VK_ACCESS_SHADER_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
.dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_READ_BIT |
VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst,
.subresourceRange = color_range,
},
};
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_TRANSFER_BIT,
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
0, nullptr, nullptr, pre_barriers);
const VkRenderPassBeginInfo renderpass_bi{
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
.pNext = nullptr,
.renderPass = renderpass,
.framebuffer = framebuffer_handle,
.renderArea = render_area,
.clearValueCount = 0,
.pClearValues = nullptr,
};
cmdbuf.BeginRenderPass(renderpass_bi, VK_SUBPASS_CONTENTS_INLINE);
const VkDescriptorSet descriptor_set = one_texture_descriptor_allocator.Commit();
UpdateOneTextureDescriptorSet(device, descriptor_set, sampler, src_view_handle);
cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, layout, 0, descriptor_set,
nullptr);
const VkViewport viewport{
.x = static_cast<float>(render_area.offset.x),
.y = static_cast<float>(render_area.offset.y),
.width = static_cast<float>(render_area.extent.width),
.height = static_cast<float>(render_area.extent.height),
.minDepth = 0.0f,
.maxDepth = 1.0f,
};
cmdbuf.SetViewport(0, viewport);
cmdbuf.SetScissor(0, render_area);
cmdbuf.PushConstants(layout, VK_SHADER_STAGE_FRAGMENT_BIT, push_constants);
cmdbuf.Draw(3, 1, 0, 0);
cmdbuf.EndRenderPass();
const VkImageMemoryBarrier post_barrier{
.sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
.pNext = nullptr,
.srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT,
.oldLayout = VK_IMAGE_LAYOUT_GENERAL,
.newLayout = VK_IMAGE_LAYOUT_GENERAL,
.srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
.image = dst,
.subresourceRange = color_range,
};
cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT |
VK_PIPELINE_STAGE_TRANSFER_BIT,
0, post_barrier);
});
msaa_copy_resources.push_back(MSAACopyResources{
.tick = scheduler.CurrentTick(),
.src_view = std::move(src_view),
.dst_view = std::move(dst_view),
.framebuffer = std::move(framebuffer),
});
}
scheduler.InvalidateState();
}
void BlitImageHelper::Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
@@ -1469,87 +1379,6 @@ VkPipeline BlitImageHelper::FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPip
return *blit_msaa_color_pipelines.back();
}
VkPipeline BlitImageHelper::FindOrEmplaceBlitDepthStencilMSAAPipeline(
const BlitMSAAPipelineKey& key, bool blit_stencil) {
auto& keys = blit_stencil ? blit_msaa_depth_stencil_keys : blit_msaa_depth_keys;
auto& pipelines = blit_stencil ? blit_msaa_depth_stencil_pipelines : blit_msaa_depth_pipelines;
const auto it = std::ranges::find(keys, key);
if (it != keys.end()) {
return *pipelines[std::distance(keys.begin(), it)];
}
keys.push_back(key);
const std::array stages =
MakeStages(*full_screen_vert,
blit_stencil ? *blit_depth_stencil_msaa_frag : *blit_depth_msaa_frag);
const VkPipelineMultisampleStateCreateInfo multisample_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.rasterizationSamples = key.samples,
.sampleShadingEnable = VK_TRUE,
.minSampleShading = 1.0f,
.pSampleMask = nullptr,
.alphaToCoverageEnable = VK_FALSE,
.alphaToOneEnable = VK_FALSE,
};
const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci = GetPipelineInputAssemblyStateCreateInfo(device);
pipelines.push_back(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 = &input_assembly_ci,
.pTessellationState = nullptr,
.pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
.pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
.pMultisampleState = &multisample_ci,
.pDepthStencilState = blit_stencil ? &PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO
: &PIPELINE_DEPTH_ONLY_STATE_CREATE_INFO,
.pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
.pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
.layout = blit_stencil ? *two_textures_pipeline_layout : *one_texture_pipeline_layout,
.renderPass = key.renderpass,
.subpass = 0,
.basePipelineHandle = VK_NULL_HANDLE,
.basePipelineIndex = 0,
}));
return *pipelines.back();
}
VkPipeline BlitImageHelper::FindOrEmplaceBlitDepthPipeline(VkRenderPass renderpass) {
const auto it = std::ranges::find(blit_depth_keys, renderpass);
if (it != blit_depth_keys.end()) {
return *blit_depth_pipelines[std::distance(blit_depth_keys.begin(), it)];
}
blit_depth_keys.push_back(renderpass);
const std::array stages = MakeStages(*full_screen_vert, *blit_depth_frag);
const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci = GetPipelineInputAssemblyStateCreateInfo(device);
blit_depth_pipelines.push_back(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 = &input_assembly_ci,
.pTessellationState = nullptr,
.pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
.pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
.pMultisampleState = &PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pDepthStencilState = &PIPELINE_DEPTH_ONLY_STATE_CREATE_INFO,
.pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
.pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
.layout = *one_texture_pipeline_layout,
.renderPass = renderpass,
.subpass = 0,
.basePipelineHandle = VK_NULL_HANDLE,
.basePipelineIndex = 0,
}));
return *blit_depth_pipelines.back();
}
VkPipeline BlitImageHelper::FindOrEmplaceResolveDepthStencilPipeline(VkRenderPass renderpass,
bool resolve_stencil) {
auto& keys = resolve_stencil ? resolve_depth_stencil_keys : resolve_depth_keys;
@@ -1588,83 +1417,15 @@ VkPipeline BlitImageHelper::FindOrEmplaceResolveDepthStencilPipeline(VkRenderPas
return *pipelines.back();
}
void BlitImageHelper::CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
VideoCore::Surface::PixelFormat src_format, u32 num_samples,
std::span<const VideoCommon::ImageCopy> copies,
bool copy_stencil, bool msaa_to_non_msaa) {
const auto [samples_x, samples_y] = VideoCommon::SamplesLog2(static_cast<int>(num_samples));
const s32 scale_x = 1 << samples_x;
const s32 scale_y = 1 << samples_y;
VkSampleCountFlagBits samples = SampleCountFlag(num_samples);
if (msaa_to_non_msaa) {
samples = VK_SAMPLE_COUNT_1_BIT;
}
RenderPassKey renderpass_key{};
renderpass_key.color_formats.fill(VideoCore::Surface::PixelFormat::Invalid);
renderpass_key.depth_format = dst_format;
renderpass_key.samples = samples;
const VkRenderPass renderpass = render_pass_cache.Get(renderpass_key);
const MSAACopyPipelineKey key{
.renderpass = renderpass,
.samples = samples,
.msaa_to_non_msaa = msaa_to_non_msaa,
.format_class = MSAACopyFormatClass::Float,
};
VkImageAspectFlags attachment_aspect = VK_IMAGE_ASPECT_DEPTH_BIT;
if (VideoCore::Surface::GetFormatType(dst_format) ==
VideoCore::Surface::SurfaceType::DepthStencil) {
attachment_aspect |= VK_IMAGE_ASPECT_STENCIL_BIT;
}
VkPipelineLayout layout = *msaa_copy_pipeline_layout;
if (copy_stencil) {
layout = *msaa_copy_depth_stencil_pipeline_layout;
}
const MSAACopyAspectInfo aspect_info{
.src_view_aspect = VK_IMAGE_ASPECT_DEPTH_BIT,
.attachment_aspect = attachment_aspect,
.barrier_aspect = attachment_aspect,
.pre_src_access =
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT | VK_ACCESS_TRANSFER_WRITE_BIT,
.pre_src_dst_access = VK_ACCESS_SHADER_READ_BIT,
.pre_dst_dst_access = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
.pre_src_stages = VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
VK_PIPELINE_STAGE_TRANSFER_BIT,
.pre_dst_stages =
VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT | VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT,
.post_src_access = VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT,
.post_dst_access = VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_TRANSFER_READ_BIT |
VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_READ_BIT,
.post_src_stages = VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT,
.post_dst_stages = vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER,
};
const VkFormat src_vk_format =
MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, src_format).format;
const VkFormat dst_vk_format =
MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, true, dst_format).format;
CopyMSAAImpl(renderpass, FindOrEmplaceMSAACopyDepthPipeline(key, copy_stencil), layout,
dst_image, dst_vk_format, src_image, src_vk_format, scale_x, scale_y, copies,
aspect_info, copy_stencil);
}
VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key) {
const auto it = std::ranges::find(msaa_copy_keys, key);
if (it != msaa_copy_keys.end()) {
return *msaa_copy_pipelines[std::distance(msaa_copy_keys.begin(), it)];
}
msaa_copy_keys.push_back(key);
VkShaderModule frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_frag
: *convert_non_msaa_to_msaa_frag;
if (key.format_class == MSAACopyFormatClass::SignedInteger) {
frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_sint_frag
: *convert_non_msaa_to_msaa_sint_frag;
} else if (key.format_class == MSAACopyFormatClass::UnsignedInteger) {
frag_module = key.msaa_to_non_msaa ? *convert_msaa_to_non_msaa_uint_frag
: *convert_non_msaa_to_msaa_uint_frag;
}
const std::array stages = MakeStages(*clear_color_vert, frag_module);
const std::array stages = MakeStages(*clear_color_vert, key.msaa_to_non_msaa
? *convert_msaa_to_non_msaa_frag
: *convert_non_msaa_to_msaa_frag);
const VkPipelineMultisampleStateCreateInfo multisample_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pNext = nullptr,
@@ -1701,85 +1462,6 @@ VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyPipeline(const MSAACopyPipeline
return *msaa_copy_pipelines.back();
}
VkPipeline BlitImageHelper::FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
bool copy_stencil) {
auto& keys = copy_stencil ? msaa_copy_depth_stencil_keys : msaa_copy_depth_keys;
auto& pipelines = copy_stencil ? msaa_copy_depth_stencil_pipelines : msaa_copy_depth_pipelines;
const auto it = std::ranges::find(keys, key);
if (it != keys.end()) {
return *pipelines[std::distance(keys.begin(), it)];
}
keys.push_back(key);
VkShaderModule frag_module;
if (key.msaa_to_non_msaa) {
frag_module = copy_stencil ? *convert_msaa_to_non_msaa_depth_stencil_frag
: *convert_msaa_to_non_msaa_depth_frag;
} else {
frag_module = copy_stencil ? *convert_non_msaa_to_msaa_depth_stencil_frag
: *convert_non_msaa_to_msaa_depth_frag;
}
const std::array stages = MakeStages(*clear_color_vert, frag_module);
const VkPipelineMultisampleStateCreateInfo multisample_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.rasterizationSamples = key.samples,
.sampleShadingEnable = key.msaa_to_non_msaa ? VK_FALSE : VK_TRUE,
.minSampleShading = key.msaa_to_non_msaa ? 0.0f : 1.0f,
.pSampleMask = nullptr,
.alphaToCoverageEnable = VK_FALSE,
.alphaToOneEnable = VK_FALSE,
};
static constexpr VkStencilOpState REPLACE_STENCIL_OP{
.failOp = VK_STENCIL_OP_REPLACE,
.passOp = VK_STENCIL_OP_REPLACE,
.depthFailOp = VK_STENCIL_OP_REPLACE,
.compareOp = VK_COMPARE_OP_ALWAYS,
.compareMask = 0xFF,
.writeMask = 0xFF,
.reference = 0,
};
const VkPipelineDepthStencilStateCreateInfo depth_stencil_ci{
.sType = VK_STRUCTURE_TYPE_PIPELINE_DEPTH_STENCIL_STATE_CREATE_INFO,
.pNext = nullptr,
.flags = 0,
.depthTestEnable = VK_TRUE,
.depthWriteEnable = VK_TRUE,
.depthCompareOp = VK_COMPARE_OP_ALWAYS,
.depthBoundsTestEnable = VK_FALSE,
.stencilTestEnable = copy_stencil ? VK_TRUE : VK_FALSE,
.front = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
.back = copy_stencil ? REPLACE_STENCIL_OP : VkStencilOpState{},
.minDepthBounds = 0.0f,
.maxDepthBounds = 0.0f,
};
const VkPipelineInputAssemblyStateCreateInfo input_assembly_ci =
GetPipelineInputAssemblyStateCreateInfo(device);
pipelines.push_back(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 = &input_assembly_ci,
.pTessellationState = nullptr,
.pViewportState = &PIPELINE_VIEWPORT_STATE_CREATE_INFO,
.pRasterizationState = &PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
.pMultisampleState = &multisample_ci,
.pDepthStencilState = &depth_stencil_ci,
.pColorBlendState = &PIPELINE_COLOR_BLEND_STATE_EMPTY_CREATE_INFO,
.pDynamicState = &PIPELINE_DYNAMIC_STATE_CREATE_INFO,
.layout = copy_stencil ? *msaa_copy_depth_stencil_pipeline_layout
: *msaa_copy_pipeline_layout,
.renderPass = key.renderpass,
.subpass = 0,
.basePipelineHandle = VK_NULL_HANDLE,
.basePipelineIndex = 0,
}));
return *pipelines.back();
}
void BlitImageHelper::ConvertDepthToColorPipeline(vk::Pipeline& pipeline, VkRenderPass renderpass) {
ConvertPipeline(pipeline, renderpass, false);
}
@@ -45,19 +45,12 @@ struct BlitDepthStencilPipelineKey {
u32 stencil_ref;
};
enum class MSAACopyFormatClass : u32 {
Float,
SignedInteger,
UnsignedInteger,
};
struct MSAACopyPipelineKey {
constexpr auto operator<=>(const MSAACopyPipelineKey&) const noexcept = default;
VkRenderPass renderpass;
VkSampleCountFlagBits samples;
bool msaa_to_non_msaa;
MSAACopyFormatClass format_class;
};
struct BlitMSAAPipelineKey {
@@ -85,12 +78,6 @@ public:
void BlitColorMSAA(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region);
void BlitDepthStencilMSAA(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region);
void BlitDepth(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region);
void ResolveDepthStencil(const Framebuffer* dst_framebuffer, ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region);
@@ -129,39 +116,7 @@ public:
VideoCore::Surface::PixelFormat src_format, u32 num_samples,
std::span<const VideoCommon::ImageCopy> copies, bool msaa_to_non_msaa);
void CopyMSAADepth(RenderPassCache& render_pass_cache, VkImage dst_image,
VideoCore::Surface::PixelFormat dst_format, VkImage src_image,
VideoCore::Surface::PixelFormat src_format, u32 num_samples,
std::span<const VideoCommon::ImageCopy> copies, bool copy_stencil,
bool msaa_to_non_msaa);
private:
struct MSAACopyAspectInfo {
VkImageAspectFlags src_view_aspect;
VkImageAspectFlags attachment_aspect;
VkImageAspectFlags barrier_aspect;
VkAccessFlags pre_src_access;
VkAccessFlags pre_src_dst_access;
VkAccessFlags pre_dst_dst_access;
VkPipelineStageFlags pre_src_stages;
VkPipelineStageFlags pre_dst_stages;
VkAccessFlags post_src_access;
VkAccessFlags post_dst_access;
VkPipelineStageFlags post_src_stages;
VkPipelineStageFlags post_dst_stages;
};
void BlitImpl(const Framebuffer* dst_framebuffer, const ImageView& src_image_view,
const Region2D& dst_region, const Region2D& src_region, VkPipeline pipeline,
VkSampler sampler, VkImageView src_view, VkImageView src_stencil_view,
bool blit_stencil);
void CopyMSAAImpl(VkRenderPass renderpass, VkPipeline pipeline, VkPipelineLayout layout,
VkImage dst_image, VkFormat dst_vk_format, VkImage src_image,
VkFormat src_vk_format, s32 scale_x, s32 scale_y,
std::span<const VideoCommon::ImageCopy> copies,
const MSAACopyAspectInfo& aspect_info, bool copy_stencil);
void Convert(VkPipeline pipeline, const Framebuffer* dst_framebuffer,
const ImageView& src_image_view);
@@ -176,13 +131,7 @@ private:
[[nodiscard]] VkPipeline FindOrEmplaceClearStencilPipeline(
const BlitDepthStencilPipelineKey& key);
[[nodiscard]] VkPipeline FindOrEmplaceMSAACopyPipeline(const MSAACopyPipelineKey& key);
[[nodiscard]] VkPipeline FindOrEmplaceMSAACopyDepthPipeline(const MSAACopyPipelineKey& key,
bool copy_stencil);
[[nodiscard]] VkPipeline FindOrEmplaceBlitColorMSAAPipeline(const BlitMSAAPipelineKey& key);
[[nodiscard]] VkPipeline FindOrEmplaceBlitDepthStencilMSAAPipeline(
const BlitMSAAPipelineKey& key, bool blit_stencil);
[[nodiscard]] VkPipeline FindOrEmplaceBlitDepthPipeline(VkRenderPass renderpass);
[[nodiscard]] VkPipeline FindOrEmplaceResolveDepthStencilPipeline(VkRenderPass renderpass,
bool resolve_stencil);
@@ -213,12 +162,10 @@ private:
vk::PipelineLayout two_textures_pipeline_layout;
vk::PipelineLayout clear_color_pipeline_layout;
vk::PipelineLayout msaa_copy_pipeline_layout;
vk::PipelineLayout msaa_copy_depth_stencil_pipeline_layout;
vk::ShaderModule full_screen_vert;
vk::ShaderModule blit_color_to_color_frag;
vk::ShaderModule blit_color_msaa_frag;
vk::ShaderModule blit_depth_stencil_frag;
vk::ShaderModule blit_depth_frag;
vk::ShaderModule blit_depth_msaa_frag;
vk::ShaderModule blit_depth_stencil_msaa_frag;
vk::ShaderModule clear_color_vert;
@@ -232,15 +179,7 @@ private:
vk::ShaderModule convert_d24s8_to_abgr8_frag;
vk::ShaderModule convert_s8d24_to_abgr8_frag;
vk::ShaderModule convert_msaa_to_non_msaa_frag;
vk::ShaderModule convert_msaa_to_non_msaa_sint_frag;
vk::ShaderModule convert_msaa_to_non_msaa_uint_frag;
vk::ShaderModule convert_msaa_to_non_msaa_depth_frag;
vk::ShaderModule convert_msaa_to_non_msaa_depth_stencil_frag;
vk::ShaderModule convert_non_msaa_to_msaa_frag;
vk::ShaderModule convert_non_msaa_to_msaa_sint_frag;
vk::ShaderModule convert_non_msaa_to_msaa_uint_frag;
vk::ShaderModule convert_non_msaa_to_msaa_depth_frag;
vk::ShaderModule convert_non_msaa_to_msaa_depth_stencil_frag;
vk::Sampler linear_sampler;
vk::Sampler nearest_sampler;
@@ -254,18 +193,8 @@ private:
std::vector<vk::Pipeline> clear_stencil_pipelines;
std::vector<MSAACopyPipelineKey> msaa_copy_keys;
std::vector<vk::Pipeline> msaa_copy_pipelines;
std::vector<MSAACopyPipelineKey> msaa_copy_depth_keys;
std::vector<vk::Pipeline> msaa_copy_depth_pipelines;
std::vector<MSAACopyPipelineKey> msaa_copy_depth_stencil_keys;
std::vector<vk::Pipeline> msaa_copy_depth_stencil_pipelines;
std::vector<BlitMSAAPipelineKey> blit_msaa_color_keys;
std::vector<vk::Pipeline> blit_msaa_color_pipelines;
std::vector<VkRenderPass> blit_depth_keys;
std::vector<vk::Pipeline> blit_depth_pipelines;
std::vector<BlitMSAAPipelineKey> blit_msaa_depth_keys;
std::vector<vk::Pipeline> blit_msaa_depth_pipelines;
std::vector<BlitMSAAPipelineKey> blit_msaa_depth_stencil_keys;
std::vector<vk::Pipeline> blit_msaa_depth_stencil_pipelines;
std::vector<VkRenderPass> resolve_depth_keys;
std::vector<vk::Pipeline> resolve_depth_pipelines;
std::vector<VkRenderPass> resolve_depth_stencil_keys;
@@ -424,12 +424,24 @@ void RasterizerVulkan::Clear(u32 layer_count) {
const bool ds_deferrable =
!ds_used || ((!framebuffer->HasAspectDepthBit() || use_depth) &&
(!framebuffer->HasAspectStencilBit() || use_stencil) && !stencil_partial);
const bool can_defer_clear = ENABLE_DEFERRED_CLEAR && !regs.clear_control.use_scissor &&
regs.clear_surface.layer == 0 &&
!scheduler.IsRenderPassActive() &&
(!use_color || color_full_channels) && ds_deferrable;
if (!can_defer_clear) {
scheduler.RequestRenderpass(framebuffer);
}
query_cache.NotifySegment(true);
query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64, maxwell3d->regs.zpass_pixel_count_enable);
u32 up_scale = 1;
u32 down_shift = 0;
if (texture_cache.IsRescaling()) {
up_scale = Settings::values.resolution_info.up_scale;
down_shift = Settings::values.resolution_info.down_shift;
}
UpdateViewportsState(regs);
VkRect2D default_scissor{};
default_scissor.offset.x = 0;
default_scissor.offset.y = 0;
@@ -485,22 +497,6 @@ void RasterizerVulkan::Clear(u32 layer_count) {
return;
}
const bool clear_covers_render_area =
clear_rect.rect.offset.x == 0 && clear_rect.rect.offset.y == 0 &&
clear_rect.rect.extent.width >= render_area.width &&
clear_rect.rect.extent.height >= render_area.height;
const bool can_defer_clear = ENABLE_DEFERRED_CLEAR && (!regs.clear_control.use_scissor || clear_covers_render_area) &&
regs.clear_surface.layer == 0 &&
!scheduler.IsRenderPassActive() &&
(!use_color || color_full_channels) && ds_deferrable;
if (!can_defer_clear) {
scheduler.RequestRenderpass(framebuffer);
}
query_cache.NotifySegment(true);
query_cache.CounterEnable(VideoCommon::QueryType::ZPassPixelCount64, maxwell3d->regs.zpass_pixel_count_enable);
UpdateViewportsState(regs);
const u32 color_attachment = regs.clear_surface.RT;
if (use_color && framebuffer->HasAspectColorBit(color_attachment)) {
const auto format = VideoCore::Surface::PixelFormatFromRenderTargetFormat(regs.rt[color_attachment].format);
@@ -65,72 +65,8 @@ using VideoCore::Surface::SurfaceType;
.finalLayout = VK_IMAGE_LAYOUT_GENERAL,
};
}
struct ResolveAspects {
bool depth;
bool stencil;
};
struct ResolveModes {
VkResolveModeFlagBits depth;
VkResolveModeFlagBits stencil;
};
constexpr ResolveAspects GetResolveAspects(PixelFormat format) {
const SurfaceType surface_type = GetSurfaceType(format);
return ResolveAspects{
.depth = surface_type == SurfaceType::Depth ||
surface_type == SurfaceType::DepthStencil,
.stencil = surface_type == SurfaceType::Stencil ||
surface_type == SurfaceType::DepthStencil,
};
}
ResolveModes PickResolveModes(const Device& device, PixelFormat format) {
constexpr VkResolveModeFlagBits mode = VK_RESOLVE_MODE_SAMPLE_ZERO_BIT;
const ResolveAspects aspects = GetResolveAspects(format);
const bool depth_mode_supported = (device.GetDepthResolveModes() & mode) != 0;
const bool stencil_mode_supported = (device.GetStencilResolveModes() & mode) != 0;
ResolveModes modes{
.depth = VK_RESOLVE_MODE_NONE,
.stencil = VK_RESOLVE_MODE_NONE,
};
if (aspects.depth && depth_mode_supported) {
modes.depth = mode;
}
if (aspects.stencil && stencil_mode_supported) {
modes.stencil = mode;
}
if (modes.depth == modes.stencil || device.SupportsIndependentResolveNone()) {
return modes;
}
if (modes.depth != VK_RESOLVE_MODE_NONE && stencil_mode_supported) {
modes.stencil = mode;
} else if (modes.stencil != VK_RESOLVE_MODE_NONE && depth_mode_supported) {
modes.depth = mode;
}
return modes;
}
} // Anonymous namespace
bool SupportsDepthStencilResolve(const Device& device, PixelFormat depth_format) {
if (depth_format == PixelFormat::Invalid || !device.IsKhrDepthStencilResolveSupported()) {
return false;
}
const ResolveAspects aspects = GetResolveAspects(depth_format);
if (!aspects.depth && !aspects.stencil) {
return false;
}
const ResolveModes modes = PickResolveModes(device, depth_format);
if ((aspects.depth && modes.depth == VK_RESOLVE_MODE_NONE) ||
(aspects.stencil && modes.stencil == VK_RESOLVE_MODE_NONE)) {
return false;
}
return modes.depth == modes.stencil || device.SupportsIndependentResolveNone();
}
RenderPassCache::RenderPassCache(const Device& device_) : device{&device_} {}
VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
@@ -139,9 +75,7 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
if (!is_new) {
return *pair->second;
}
static constexpr size_t MAX_ATTACHMENTS =
2 * std::tuple_size_v<decltype(RenderPassKey::color_formats)> + 2;
boost::container::static_vector<VkAttachmentDescription, MAX_ATTACHMENTS> descriptions;
boost::container::static_vector<VkAttachmentDescription, 9> descriptions;
std::array<VkAttachmentReference, 8> references{};
u32 num_attachments{};
u32 num_colors{};
@@ -175,11 +109,8 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
const VkAttachmentLoadOp depth_load_op = key.depth_stencil_clear
? VK_ATTACHMENT_LOAD_OP_CLEAR
: VK_ATTACHMENT_LOAD_OP_LOAD;
const VkAttachmentStoreOp depth_store_op = key.depth_stencil_discard
? VK_ATTACHMENT_STORE_OP_DONT_CARE
: VK_ATTACHMENT_STORE_OP_STORE;
descriptions.push_back(AttachmentDescription(*device, key.depth_format, key.samples,
depth_load_op, depth_store_op));
depth_load_op, VK_ATTACHMENT_STORE_OP_STORE));
}
std::array<VkAttachmentReference, 8> resolve_references{};
const bool do_resolve_color =
@@ -202,21 +133,6 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
}
}
}
const bool do_resolve_depth_stencil = key.resolve_depth_stencil && has_depth &&
key.samples != VK_SAMPLE_COUNT_1_BIT &&
SupportsDepthStencilResolve(*device, key.depth_format);
VkAttachmentReference depth_resolve_reference{};
if (do_resolve_depth_stencil) {
depth_resolve_reference = VkAttachmentReference{
.attachment = static_cast<u32>(descriptions.size()),
.layout = VK_IMAGE_LAYOUT_GENERAL,
};
VkAttachmentDescription resolve_desc =
AttachmentDescription(*device, key.depth_format, VK_SAMPLE_COUNT_1_BIT,
VK_ATTACHMENT_LOAD_OP_DONT_CARE, VK_ATTACHMENT_STORE_OP_STORE);
resolve_desc.initialLayout = VK_IMAGE_LAYOUT_UNDEFINED;
descriptions.push_back(resolve_desc);
}
const VkSubpassDescription subpass{
.flags = 0,
.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
@@ -241,92 +157,6 @@ VkRenderPass RenderPassCache::Get(const RenderPassKey& key) {
.dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
.dependencyFlags = VK_DEPENDENCY_BY_REGION_BIT
};
if (device->IsKhrCreateRenderPass2Supported()) {
boost::container::static_vector<VkAttachmentDescription2, MAX_ATTACHMENTS> descriptions2;
for (const VkAttachmentDescription& description : descriptions) {
descriptions2.push_back(VkAttachmentDescription2{
.sType = VK_STRUCTURE_TYPE_ATTACHMENT_DESCRIPTION_2,
.pNext = nullptr,
.flags = description.flags,
.format = description.format,
.samples = description.samples,
.loadOp = description.loadOp,
.storeOp = description.storeOp,
.stencilLoadOp = description.stencilLoadOp,
.stencilStoreOp = description.stencilStoreOp,
.initialLayout = description.initialLayout,
.finalLayout = description.finalLayout,
});
}
const auto promote = [](const VkAttachmentReference& reference) {
return VkAttachmentReference2{
.sType = VK_STRUCTURE_TYPE_ATTACHMENT_REFERENCE_2,
.pNext = nullptr,
.attachment = reference.attachment,
.layout = reference.layout,
.aspectMask = 0,
};
};
std::array<VkAttachmentReference2, 8> references2{};
std::array<VkAttachmentReference2, 8> resolve_references2{};
for (size_t index = 0; index < references.size(); ++index) {
references2[index] = promote(references[index]);
resolve_references2[index] = promote(resolve_references[index]);
}
const VkAttachmentReference2 depth_reference2 = promote(depth_reference);
const VkAttachmentReference2 depth_resolve_reference2 = promote(depth_resolve_reference);
const ResolveModes resolve_modes = PickResolveModes(*device, key.depth_format);
const VkSubpassDescriptionDepthStencilResolve depth_stencil_resolve{
.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_DEPTH_STENCIL_RESOLVE,
.pNext = nullptr,
.depthResolveMode = resolve_modes.depth,
.stencilResolveMode = resolve_modes.stencil,
.pDepthStencilResolveAttachment = &depth_resolve_reference2,
};
const VkSubpassDescription2 subpass2{
.sType = VK_STRUCTURE_TYPE_SUBPASS_DESCRIPTION_2,
.pNext = do_resolve_depth_stencil ? &depth_stencil_resolve : nullptr,
.flags = 0,
.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
.viewMask = 0,
.inputAttachmentCount = 0,
.pInputAttachments = nullptr,
.colorAttachmentCount = num_attachments,
.pColorAttachments = references2.data(),
.pResolveAttachments = do_resolve_color ? resolve_references2.data() : nullptr,
.pDepthStencilAttachment = has_depth ? &depth_reference2 : nullptr,
.preserveAttachmentCount = 0,
.pPreserveAttachments = nullptr,
};
const VkSubpassDependency2 dependency2{
.sType = VK_STRUCTURE_TYPE_SUBPASS_DEPENDENCY_2,
.pNext = nullptr,
.srcSubpass = dependency.srcSubpass,
.dstSubpass = dependency.dstSubpass,
.srcStageMask = dependency.srcStageMask,
.dstStageMask = dependency.dstStageMask,
.srcAccessMask = dependency.srcAccessMask,
.dstAccessMask = dependency.dstAccessMask,
.dependencyFlags = dependency.dependencyFlags,
.viewOffset = 0,
};
pair->second = device->GetLogical().CreateRenderPass2({
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO_2,
.pNext = nullptr,
.flags = 0,
.attachmentCount = static_cast<u32>(descriptions2.size()),
.pAttachments = descriptions2.empty() ? nullptr : descriptions2.data(),
.subpassCount = 1,
.pSubpasses = &subpass2,
.dependencyCount = 1,
.pDependencies = &dependency2,
.correlatedViewMaskCount = 0,
.pCorrelatedViewMasks = nullptr,
});
return *pair->second;
}
pair->second = device->GetLogical().CreateRenderPass({
.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO,
.pNext = nullptr,
@@ -9,7 +9,6 @@
#include <mutex>
#include <ankerl/unordered_dense.h>
#include "common/container_hash.h"
#include "video_core/surface.h"
#include "video_core/vulkan_common/vulkan_wrapper.h"
@@ -22,11 +21,9 @@ struct RenderPassKey {
VideoCore::Surface::PixelFormat depth_format;
VkSampleCountFlagBits samples;
bool resolve_color;
bool resolve_depth_stencil;
u32 color_clear_mask;
bool depth_stencil_clear;
u32 color_discard_mask;
bool depth_stencil_discard;
};
} // namespace Vulkan
@@ -34,28 +31,17 @@ struct RenderPassKey {
namespace std {
template <>
struct hash<Vulkan::RenderPassKey> {
static_assert(std::tuple_size_v<decltype(Vulkan::RenderPassKey::color_formats)> <= 8);
static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Invalid) <= 0xFF);
static_assert(static_cast<u32>(VideoCore::Surface::PixelFormat::Max) <= 0xFF);
static_assert(VK_SAMPLE_COUNT_64_BIT <= 0xFF);
[[nodiscard]] size_t operator()(const Vulkan::RenderPassKey& key) const noexcept {
u64 formats = 0;
for (size_t index = 0; index < key.color_formats.size(); ++index) {
formats |= static_cast<u64>(key.color_formats[index]) << (index * 8);
size_t value = static_cast<size_t>(key.depth_format) << 48;
value ^= static_cast<size_t>(key.samples) << 52;
value ^= static_cast<size_t>(key.resolve_color) << 63;
value ^= static_cast<size_t>(key.color_clear_mask) << 54;
value ^= static_cast<size_t>(key.depth_stencil_clear) << 62;
value ^= static_cast<size_t>(key.color_discard_mask) << 24;
for (size_t i = 0; i < key.color_formats.size(); ++i) {
value ^= static_cast<size_t>(key.color_formats[i]) << (i * 6);
}
const u64 state = static_cast<u64>(key.depth_format) |
(static_cast<u64>(key.samples) << 8) |
(static_cast<u64>(key.color_clear_mask) << 16) |
(static_cast<u64>(key.color_discard_mask) << 24) |
(static_cast<u64>(key.resolve_color) << 32) |
(static_cast<u64>(key.depth_stencil_clear) << 33) |
(static_cast<u64>(key.resolve_depth_stencil) << 34) |
(static_cast<u64>(key.depth_stencil_discard) << 35);
size_t seed = 0;
Common::HashCombine(seed, formats);
Common::HashCombine(seed, state);
return seed;
return value;
}
};
} // namespace std
@@ -64,9 +50,6 @@ namespace Vulkan {
class Device;
[[nodiscard]] bool SupportsDepthStencilResolve(const Device& device,
VideoCore::Surface::PixelFormat depth_format);
class RenderPassCache {
public:
explicit RenderPassCache(const Device& device_);
@@ -132,7 +132,6 @@ void Scheduler::BeginRenderPassImpl(const Framebuffer* framebuffer, VkRenderPass
num_renderpass_images = framebuffer->NumImages();
renderpass_images = framebuffer->Images();
renderpass_image_ranges = framebuffer->ImageRanges();
framebuffer->MarkResolveShadowsUpToDate();
}
void Scheduler::RealizeDeferredClear() {
@@ -156,10 +155,8 @@ void Scheduler::RealizeDeferredClear() {
}
const u32 color_discard_mask =
dc.framebuffer->DiscardsMsaaColor() ? dc.color_clear_mask : 0u;
const bool depth_stencil_discard =
dc.depth_stencil && dc.framebuffer->DiscardsMsaaDepthStencil();
const VkRenderPass renderpass = dc.framebuffer->RenderPassVariant(
dc.color_clear_mask, dc.depth_stencil, color_discard_mask, depth_stencil_discard);
dc.color_clear_mask, dc.depth_stencil, color_discard_mask);
EndRenderPass();
BeginRenderPassImpl(dc.framebuffer, renderpass, clear_values.data(), count);
}
@@ -193,14 +190,6 @@ bool Scheduler::DeferDepthStencilClear(const Framebuffer* framebuffer, const VkC
return true;
}
void Scheduler::FlushDeferredClear() {
if (deferred_clear.framebuffer == nullptr) {
return;
}
RealizeDeferredClear();
EndRenderPass();
}
void Scheduler::RequestRenderpass(const Framebuffer* framebuffer) {
if (deferred_clear.framebuffer == framebuffer) {
RealizeDeferredClear();
@@ -65,9 +65,6 @@ public:
/// Defers a full depth/stencil clear so it becomes the next render pass.
bool DeferDepthStencilClear(const Framebuffer* framebuffer, const VkClearValue& value);
/// Realizes any pending deferred clear before its framebuffer can be moved or freed.
void FlushDeferredClear();
/// Requests the current execution context to be able to execute operations only allowed outside
/// of a renderpass.
void RequestOutsideRenderPassOperationContext();
@@ -33,10 +33,10 @@ constexpr VkDeviceSize MAX_ALIGNMENT = 256;
// Windows ones however, can intake bigger buffers and generally do not OOM.
// - GTX 960 on Windows will not OOM with 256mib
// - GT 1030 on ^NIX will OOM with 256mib
#if defined(__FreeBSD__)
constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 128_MiB;
#else
#if defined(_WIN32) || defined(__ANDROID__)
constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 256_MiB;
#else
constexpr VkDeviceSize MAX_STREAM_BUFFER_SIZE = 128_MiB;
#endif
size_t GetStreamBufferSize(const Device& device) {
File diff suppressed because it is too large Load Diff
@@ -60,8 +60,6 @@ public:
void TickFrame();
void FlushDeferredClear();
u64 GetDeviceLocalMemory() const;
u64 GetDeviceMemoryUsage() const;
@@ -70,12 +68,6 @@ public:
std::optional<size_t> GetSamplerHeapBudget() const;
bool CanDownloadMsaa(const VideoCommon::ImageInfo& info) const;
[[nodiscard]] VkImage AcquireMsaaScratchImage(const VkImageCreateInfo& image_ci);
void ReleaseMsaaScratchImage(VkImage image);
void BlitImage(Framebuffer* dst_framebuffer, ImageView& dst, ImageView& src,
const Region2D& dst_region, const Region2D& src_region,
Tegra::Engines::Fermi2D::Filter filter,
@@ -125,20 +117,16 @@ public:
VkFormat format = VK_FORMAT_UNDEFINED;
VkExtent2D extent{};
u32 layers = 0;
VkImageAspectFlags aspect_mask = VK_IMAGE_ASPECT_COLOR_BIT;
bool up_to_date = false;
};
[[nodiscard]] VkImageView GetOrCreateResolveShadow(VkImage msaa_image, VkFormat format,
VkExtent2D extent, u32 layers,
VkImageAspectFlags aspect_mask);
VkExtent2D extent, u32 layers);
[[nodiscard]] const ResolveShadow* GetValidResolveShadow(VkImage msaa_image) const;
void InvalidateResolveShadow(VkImage msaa_image);
void MarkResolveShadowUpToDate(VkImage msaa_image);
void EraseResolveShadow(VkImage msaa_image);
std::span<const VkFormat> ViewFormats(PixelFormat format) {
@@ -166,30 +154,8 @@ public:
static constexpr size_t indexing_slots = 8 * sizeof(size_t);
std::array<vk::Buffer, indexing_slots> buffers{};
struct MsaaScratchKey {
VkFormat format;
VkImageType type;
u32 width;
u32 height;
u32 depth;
u32 levels;
u32 layers;
VkImageUsageFlags usage;
VkImageCreateFlags flags;
bool operator==(const MsaaScratchKey&) const noexcept = default;
};
struct MsaaScratchImage {
MsaaScratchKey key;
vk::Image image;
u64 tick;
u32 unused_frames;
};
std::vector<MsaaScratchImage> msaa_scratch_images;
std::vector<std::pair<u64, vk::Image>> pending_msaa_images;
ankerl::unordered_dense::map<VkImage, ResolveShadow> resolve_shadows;
std::vector<std::pair<u64, ResolveShadow>> pending_resolve_shadows;
};
class Framebuffer {
@@ -225,8 +191,7 @@ public:
}
[[nodiscard]] VkRenderPass RenderPassVariant(u32 color_clear_mask, bool depth_stencil_clear,
u32 color_discard_mask,
bool depth_stencil_discard) const;
u32 color_discard_mask) const;
[[nodiscard]] VkExtent2D RenderArea() const noexcept {
return render_area;
@@ -268,21 +233,19 @@ public:
return is_rescaled;
}
[[nodiscard]] bool HasResolveColor() const noexcept {
return !resolve_images.empty();
}
[[nodiscard]] VkImage ResolveColorImage(size_t index) const noexcept {
return index < resolve_images.size() ? *resolve_images[index] : VK_NULL_HANDLE;
}
[[nodiscard]] bool DiscardsMsaaColor() const noexcept {
return discard_msaa_color;
}
[[nodiscard]] bool DiscardsMsaaDepthStencil() const noexcept {
return discard_msaa_depth_stencil;
}
/// Records that a render pass has begun, so its resolve attachments will hold valid contents
/// once it ends.
void MarkResolveShadowsUpToDate() const;
private:
static constexpr size_t NUM_MEMOIZED_RENDER_PASS_VARIANTS = 8;
vk::Framebuffer framebuffer;
VkRenderPass renderpass{};
VkExtent2D render_area{};
@@ -295,16 +258,11 @@ private:
bool has_depth{};
bool has_stencil{};
bool is_rescaled{};
std::array<VkImage, 9> resolve_shadow_images{};
u32 num_resolve_shadows = 0;
TextureCacheRuntime* runtime_ptr{nullptr};
std::vector<vk::Image> resolve_images;
std::vector<vk::ImageView> resolve_image_views;
RenderPassKey render_pass_key{};
RenderPassCache* render_pass_cache{nullptr};
bool discard_msaa_color{};
bool discard_msaa_depth_stencil{};
mutable std::array<u32, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_keys{};
mutable std::array<VkRenderPass, NUM_MEMOIZED_RENDER_PASS_VARIANTS> variant_render_passes{};
mutable u32 num_memoized_variants{};
};
class Image : public VideoCommon::ImageBase {
@@ -528,7 +486,6 @@ struct TextureCacheParams {
static constexpr bool HAS_EMULATED_COPIES = false;
static constexpr bool HAS_DEVICE_MEMORY_INFO = true;
static constexpr bool IMPLEMENTS_ASYNC_DOWNLOADS = true;
static constexpr bool HAS_MSAA_DOWNLOADS = true;
using Runtime = Vulkan::TextureCacheRuntime;
using Image = Vulkan::Image;
+6 -2
View File
@@ -1,4 +1,4 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
@@ -121,7 +121,7 @@ void ImageBase::InsertView(const ImageViewInfo& view_info, ImageViewId image_vie
image_view_ids.push_back(image_view_id);
}
bool ImageBase::IsSafeGpuCopy() const noexcept {
bool ImageBase::IsSafeDownload() const noexcept {
// Skip images that were not modified from the GPU
if (False(flags & ImageFlagBits::GpuModified)) {
return false;
@@ -131,6 +131,10 @@ bool ImageBase::IsSafeGpuCopy() const noexcept {
if (True(flags & ImageFlagBits::CpuModified)) {
return false;
}
if (info.num_samples > 1) {
LOG_WARNING(HW_GPU, "MSAA image downloads are not implemented");
return false;
}
return true;
}
+1 -4
View File
@@ -1,6 +1,3 @@
// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
// SPDX-License-Identifier: GPL-3.0-or-later
// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
// SPDX-License-Identifier: GPL-2.0-or-later
@@ -67,7 +64,7 @@ struct ImageBase {
void InsertView(const ImageViewInfo& view_info, ImageViewId image_view_id);
[[nodiscard]] bool IsSafeGpuCopy() const noexcept;
[[nodiscard]] bool IsSafeDownload() const noexcept;
[[nodiscard]] bool Overlaps(VAddr overlap_cpu_addr, size_t overlap_size) const noexcept {
const VAddr overlap_end = overlap_cpu_addr + overlap_size;
+7 -27
View File
@@ -134,7 +134,7 @@ void TextureCache<P>::RunGarbageCollector() {
if (True(image.flags & ImageFlagBits::IsDecoding)) {
return false;
}
const bool must_download = IsDownloadable(image) && False(image.flags & ImageFlagBits::BadOverlap);
const bool must_download = image.IsSafeDownload() && False(image.flags & ImageFlagBits::BadOverlap);
if ((!aggressive_mode && True(image.flags & ImageFlagBits::CostlyLoad)) || (!high_priority_mode && must_download)) {
return false;
}
@@ -577,7 +577,6 @@ FramebufferId TextureCache<P>::GetFramebufferId(const RenderTargets& key) {
return id ? &slot_image_views[id] : nullptr;
});
ImageView* const depth_buffer = key.depth_buffer_id ? &slot_image_views[key.depth_buffer_id] : nullptr;
runtime.FlushDeferredClear();
framebuffer_id = slot_framebuffers.insert(runtime, color_buffers, depth_buffer, key);
return framebuffer_id;
}
@@ -595,26 +594,11 @@ void TextureCache<P>::WriteMemory(DAddr cpu_addr, size_t size) {
});
}
template <class P>
bool TextureCache<P>::IsDownloadable(const ImageBase& image) const noexcept {
if (!image.IsSafeGpuCopy()) {
return false;
}
if (image.info.num_samples == 1) {
return true;
}
if constexpr (P::HAS_MSAA_DOWNLOADS) {
return runtime.CanDownloadMsaa(image.info);
} else {
return false;
}
}
template <class P>
void TextureCache<P>::DownloadMemory(DAddr cpu_addr, size_t size) {
boost::container::small_vector<ImageId, 16> images;
ForEachImageInRegion(cpu_addr, size, [this, &images](ImageId image_id, ImageBase& image) {
if (!IsDownloadable(image)) {
ForEachImageInRegion(cpu_addr, size, [&images](ImageId image_id, ImageBase& image) {
if (!image.IsSafeDownload()) {
return;
}
image.flags &= ~ImageFlagBits::GpuModified;
@@ -1492,12 +1476,12 @@ template <class P>
bool TextureCache<P>::ScaleUp(Image& image) {
const bool has_copy = image.HasScaled();
const bool rescaled = image.ScaleUp();
if (!has_copy && image.HasScaled()) {
total_used_memory += GetScaledImageSizeBytes(image);
}
if (!rescaled) {
return false;
}
if (!has_copy) {
total_used_memory += GetScaledImageSizeBytes(image);
}
InvalidateScale(image);
return true;
}
@@ -1707,10 +1691,7 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, DA
for (const auto& copy_object : join_copies_to_do) {
Image& overlap = slot_images[copy_object.id];
if (copy_object.is_alias) {
if (!overlap.IsSafeGpuCopy()) {
continue;
}
if (overlap.info.num_samples != new_image.info.num_samples) {
if (!overlap.IsSafeDownload()) {
continue;
}
const auto alias_pointer = join_alias_indices.find(copy_object.id);
@@ -2480,7 +2461,6 @@ void TextureCache<P>::RemoveImageViewReferences(std::span<const ImageViewId> rem
template <class P>
void TextureCache<P>::RemoveFramebuffers(std::span<const ImageViewId> removed_views) {
runtime.FlushDeferredClear();
auto it = framebuffers.begin();
while (it != framebuffers.end()) {
if (it->first.Contains(removed_views)) {
@@ -303,8 +303,6 @@ private:
FramebufferId GetFramebufferId(const RenderTargets& key);
/// Refresh the contents (pixel data) of an image
[[nodiscard]] bool IsDownloadable(const ImageBase& image) const noexcept;
void RefreshContents(Image& image, ImageId image_id);
/// Upload data from guest to an image
+7 -11
View File
@@ -650,6 +650,13 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
features.shader_float16_int8.shaderFloat16 = false;
}
if (is_intel_windows) {
LOG_WARNING(Render_Vulkan,
"Intel proprietary drivers do not support MSAA->MSAA image blits. "
"MSAA scaling will use 3D helpers. MSAA resolves work normally.");
cant_blit_msaa = true;
}
has_broken_compute =
CheckBrokenCompute(properties.driver.driverID, properties.properties.driverVersion) &&
!Settings::values.enable_compute_pipelines.GetValue();
@@ -968,12 +975,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
FOR_EACH_VK_FEATURE_EXT(FEATURE_EXTENSION);
FOR_EACH_VK_EXTENSION(EXTENSION);
extensions.depth_stencil_resolve =
extensions.depth_stencil_resolve &&
(instance_version >= VK_API_VERSION_1_2 || extensions.create_renderpass2);
RemoveExtensionIfUnsuitable(extensions.depth_stencil_resolve,
VK_KHR_DEPTH_STENCIL_RESOLVE_EXTENSION_NAME);
if (supported_extensions.contains(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME)) {
loaded_extensions.erase(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME);
loaded_extensions.insert(VK_KHR_ROBUSTNESS_2_EXTENSION_NAME);
@@ -1121,11 +1122,6 @@ bool Device::GetSuitability(bool requires_swapchain) {
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR;
SetNext(next, properties.push_descriptor);
}
if (extensions.depth_stencil_resolve || instance_version >= VK_API_VERSION_1_2) {
properties.depth_stencil_resolve.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DEPTH_STENCIL_RESOLVE_PROPERTIES;
SetNext(next, properties.depth_stencil_resolve);
}
if (extensions.descriptor_buffer) {
properties.descriptor_buffer.sType =
VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_BUFFER_PROPERTIES_EXT;
+5 -29
View File
@@ -91,8 +91,6 @@ VK_DEFINE_HANDLE(VmaAllocator)
EXTENSION(EXT, SHADER_VIEWPORT_INDEX_LAYER, shader_viewport_index_layer) \
EXTENSION(EXT, TOOLING_INFO, tooling_info) \
EXTENSION(EXT, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor) \
EXTENSION(KHR, CREATE_RENDERPASS_2, create_renderpass2) \
EXTENSION(KHR, DEPTH_STENCIL_RESOLVE, depth_stencil_resolve) \
EXTENSION(KHR, DRAW_INDIRECT_COUNT, draw_indirect_count) \
EXTENSION(KHR, DRIVER_PROPERTIES, driver_properties) \
EXTENSION(KHR, PUSH_DESCRIPTOR, push_descriptor) \
@@ -615,32 +613,6 @@ FN_MAX_LIMIT_LIST
return extensions.shader_stencil_export;
}
/// Returns true if the device supports VK_KHR_create_renderpass2.
bool IsKhrCreateRenderPass2Supported() const {
return extensions.create_renderpass2 || instance_version >= VK_API_VERSION_1_2;
}
/// Returns true if the device supports VK_KHR_depth_stencil_resolve.
bool IsKhrDepthStencilResolveSupported() const {
return (extensions.depth_stencil_resolve || instance_version >= VK_API_VERSION_1_2) &&
IsKhrCreateRenderPass2Supported();
}
/// Returns the supported resolve modes for the depth aspect.
VkResolveModeFlags GetDepthResolveModes() const {
return properties.depth_stencil_resolve.supportedDepthResolveModes;
}
/// Returns the supported resolve modes for the stencil aspect.
VkResolveModeFlags GetStencilResolveModes() const {
return properties.depth_stencil_resolve.supportedStencilResolveModes;
}
/// Returns true if only one of the depth and stencil aspects may be resolved.
bool SupportsIndependentResolveNone() const {
return properties.depth_stencil_resolve.independentResolveNone == VK_TRUE;
}
/// Returns true if depth/stencil operations can be performed efficiently.
/// Either through shader export or hardware blits.
bool CanPerformDepthStencilOperations() const {
@@ -949,6 +921,10 @@ FN_MAX_LIMIT_LIST
return supports_d24_depth;
}
bool CantBlitMSAA() const {
return cant_blit_msaa;
}
bool MustEmulateScaledFormats() const {
return must_emulate_scaled_formats;
}
@@ -1172,7 +1148,6 @@ private:
VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{};
VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{};
VkPhysicalDeviceMaintenance5PropertiesKHR maintenance5{};
VkPhysicalDeviceDepthStencilResolveProperties depth_stencil_resolve{};
VkPhysicalDeviceProperties properties{};
};
@@ -1200,6 +1175,7 @@ private:
bool has_nsight_graphics{}; ///< Has Nsight Graphics attached
bool has_radeon_gpu_profiler{}; ///< Has Radeon GPU Profiler attached.
bool supports_d24_depth{}; ///< Supports D24 depth buffers.
bool cant_blit_msaa{}; ///< Does not support MSAA<->MSAA blitting.
bool must_emulate_scaled_formats{}; ///< Requires scaled vertex format emulation
bool dynamic_state3_blending{}; ///< Has blending features of dynamic_state3.
bool dynamic_state3_enables{}; ///< Has at least one enable feature of dynamic_state3.
@@ -184,7 +184,6 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
X(vkCreatePipelineLayout);
X(vkCreateQueryPool);
X(vkCreateRenderPass);
X(vkCreateRenderPass2);
X(vkCreateSampler);
X(vkCreateSemaphore);
X(vkCreateShaderModule);
@@ -271,10 +270,6 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
if (!dld.vkQueueSubmit2) {
Proc(dld.vkQueueSubmit2, dld, "vkQueueSubmit2KHR", device);
}
if (!dld.vkCreateRenderPass2) {
Proc(dld.vkCreateRenderPass2, dld, "vkCreateRenderPass2KHR", device);
}
#undef X
}
@@ -730,12 +725,6 @@ RenderPass Device::CreateRenderPass(const VkRenderPassCreateInfo& ci) const {
return RenderPass(object, handle, *dld);
}
RenderPass Device::CreateRenderPass2(const VkRenderPassCreateInfo2& ci) const {
VkRenderPass object;
Check(dld->vkCreateRenderPass2(handle, &ci, nullptr, &object));
return RenderPass(object, handle, *dld);
}
DescriptorSetLayout Device::CreateDescriptorSetLayout(
const VkDescriptorSetLayoutCreateInfo& ci) const {
VkDescriptorSetLayout object;
@@ -300,7 +300,6 @@ struct DeviceDispatch : InstanceDispatch {
PFN_vkCreatePipelineLayout vkCreatePipelineLayout{};
PFN_vkCreateQueryPool vkCreateQueryPool{};
PFN_vkCreateRenderPass vkCreateRenderPass{};
PFN_vkCreateRenderPass2 vkCreateRenderPass2{};
PFN_vkCreateSampler vkCreateSampler{};
PFN_vkCreateSemaphore vkCreateSemaphore{};
PFN_vkCreateShaderModule vkCreateShaderModule{};
@@ -1046,8 +1045,6 @@ public:
[[nodiscard]] RenderPass CreateRenderPass(const VkRenderPassCreateInfo& ci) const;
[[nodiscard]] RenderPass CreateRenderPass2(const VkRenderPassCreateInfo2& ci) const;
[[nodiscard]] DescriptorSetLayout CreateDescriptorSetLayout(
const VkDescriptorSetLayoutCreateInfo& ci) const;