thirdpartymirrors/lsfg-vk
1
0
Fork 0
mirror of https://github.com/PancakeTAS/lsfg-vk.git synced 2025-10-30 07:01:10 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

multi frame generation for backend

Browse source
This commit is contained in:
PancakeTAS 2025-07-02 13:30:45 +02:00
parent 3a6458ab7a
commit f3b5cfc131
No known key found for this signature in database
22 changed files with 402 additions and 239 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

16
lsfg-vk-gen/include/context.hpp
View file

@ -32,22 +32,24 @@ namespace LSFG {
/// @param height Height of the input images.
/// @param in0 File descriptor for the first input image.
/// @param in1 File descriptor for the second input image.
/// @param out File descriptor for the output image.
/// @param outN File descriptor for the output image.
///
/// @throws LSFG::vulkan_error if the generator fails to initialize.
///
Context(const Core::Device& device, uint32_t width, uint32_t height, int in0, int in1, int out);
Context(const Core::Device& device, uint32_t width, uint32_t height, int in0, int in1,
const std::vector<int>& outN);
///
/// Schedule the next generation.
///
/// @param device The Vulkan device to use.
/// @param inSem Semaphore to wait on before starting the generation.
/// @param outSem Semaphore to signal when the generation is complete.
/// @param outSem Semaphores to signal after each generation is done.
///
/// @throws LSFG::vulkan_error if the generator fails to present.
///
void present(const Core::Device& device, int inSem, int outSem);
void present(const Core::Device& device, int inSem,
const std::vector<int>& outSem);
// Trivially copyable, moveable and destructible
Context(const Context&) = default;
@ -61,8 +63,10 @@ namespace LSFG {
Core::Image inImg_0, inImg_1; // inImg_0 is next (inImg_1 prev) when fc % 2 == 0
std::array<Core::Semaphore, 8> inSemaphores;
std::array<Core::Semaphore, 8> outSemaphores;
std::array<Core::CommandBuffer, 8> cmdBuffers;
std::array<std::vector<Core::Semaphore>, 8> internalSemaphores;
std::vector<std::array<Core::Semaphore, 8>> outSemaphores;
std::array<Core::CommandBuffer, 8> cmdBuffers1;
std::vector<std::array<Core::CommandBuffer, 8>> cmdBuffers2;
uint64_t fc{0};
Shaderchains::Downsample downsampleChain;

12
lsfg-vk-gen/include/shaderchains/beta.hpp
View file

@ -32,23 +32,26 @@ namespace LSFG::Shaderchains {
/// @param inImgs_0 The next input images to process (when fc % 3 == 0)
/// @param inImgs_1 The prev input images to process (when fc % 3 == 0)
/// @param inImgs_2 The prev prev input images to process (when fc % 3 == 0)
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Beta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 4> inImgs_0,
std::array<Core::Image, 4> inImgs_1,
std::array<Core::Image, 4> inImgs_2);
std::array<Core::Image, 4> inImgs_2,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param fc The frame count, used to select the input images.
/// @param pass The pass number
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc);
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass);
/// Get the output images.
[[nodiscard]] const auto& getOutImages() const { return this->outImgs; }
@ -62,9 +65,10 @@ namespace LSFG::Shaderchains {
private:
std::array<Core::ShaderModule, 5> shaderModules;
std::array<Core::Pipeline, 5> pipelines;
std::array<Core::DescriptorSet, 4> descriptorSets; // first shader has special logic
std::array<Core::DescriptorSet, 3> descriptorSets; // first shader has special logic
std::array<Core::DescriptorSet, 3> specialDescriptorSets;
Core::Buffer buffer;
std::vector<Core::DescriptorSet> nDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 4> inImgs_0;
std::array<Core::Image, 4> inImgs_1;

12
lsfg-vk-gen/include/shaderchains/delta.hpp
View file

@ -31,21 +31,24 @@ namespace LSFG::Shaderchains {
/// @param pool The descriptor pool to allocate in.
/// @param inImgs The input images to process.
/// @param optImg An optional additional input from the previous pass.
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Delta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 2> inImgs,
std::optional<Core::Image> optImg);
std::optional<Core::Image> optImg,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf);
void Dispatch(const Core::CommandBuffer& buf, uint64_t pass);
/// Get the output image.
[[nodiscard]] const auto& getOutImage() const { return this->outImg; }
@ -59,8 +62,9 @@ namespace LSFG::Shaderchains {
private:
std::array<Core::ShaderModule, 4> shaderModules;
std::array<Core::Pipeline, 4> pipelines;
std::array<Core::DescriptorSet, 4> descriptorSets;
Core::Buffer buffer;
std::array<Core::DescriptorSet, 3> descriptorSets;
std::vector<Core::DescriptorSet> nDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 2> inImgs;
std::optional<Core::Image> optImg;

4
lsfg-vk-gen/include/shaderchains/downsample.hpp
View file

@ -30,11 +30,13 @@ namespace LSFG::Shaderchains {
/// @param pool The descriptor pool to allocate in.
/// @param inImg_0 The next full image to downsample (when fc % 2 == 0)
/// @param inImg_1 The next full image to downsample (when fc % 2 == 1)
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Downsample(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg_0, Core::Image inImg_1);
Core::Image inImg_0, Core::Image inImg_1,
size_t genc);
///
/// Dispatch the shaderchain.

12
lsfg-vk-gen/include/shaderchains/epsilon.hpp
View file

@ -32,22 +32,25 @@ namespace LSFG::Shaderchains {
/// @param inImgs1 The first set of input images to process.
/// @param inImg2 The second type image to process.
/// @param optImg An optional additional input from the previous pass.
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Epsilon(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 3> inImgs1,
Core::Image inImg2,
std::optional<Core::Image> optImg);
std::optional<Core::Image> optImg,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf);
void Dispatch(const Core::CommandBuffer& buf, uint64_t pass);
/// Get the output image.
[[nodiscard]] const auto& getOutImage() const { return this->outImg; }
@ -61,8 +64,9 @@ namespace LSFG::Shaderchains {
private:
std::array<Core::ShaderModule, 4> shaderModules;
std::array<Core::Pipeline, 4> pipelines;
std::array<Core::DescriptorSet, 4> descriptorSets;
Core::Buffer buffer;
std::array<Core::DescriptorSet, 3> descriptorSets;
std::vector<Core::DescriptorSet> nDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 3> inImgs1;
Core::Image inImg2;

11
lsfg-vk-gen/include/shaderchains/extract.hpp
View file

@ -30,22 +30,25 @@ namespace LSFG::Shaderchains {
/// @param inImg1 The first set of input images to process.
/// @param inImg2 The second type image to process.
/// @param outExtent The extent of the output image.
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Extract(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg1,
Core::Image inImg2,
VkExtent2D outExtent);
VkExtent2D outExtent,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf);
void Dispatch(const Core::CommandBuffer& buf, uint64_t pass);
/// Get the output image.
[[nodiscard]] const auto& getOutImage() const { return this->outImg; }
@ -59,8 +62,8 @@ namespace LSFG::Shaderchains {
private:
Core::ShaderModule shaderModule;
Core::Pipeline pipeline;
Core::DescriptorSet descriptorSet;
Core::Buffer buffer;
std::vector<Core::DescriptorSet> nDescriptorSets;
std::vector<Core::Buffer> buffers;
Core::Image inImg1;
Core::Image inImg2;

15
lsfg-vk-gen/include/shaderchains/gamma.hpp
View file

@ -38,6 +38,7 @@ namespace LSFG::Shaderchains {
/// @param optImg1 An optional additional input from the previous pass.
/// @param optImg2 An optional additional input image for processing non-first passes.
/// @param outExtent The extent of the output image.
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
@ -48,17 +49,19 @@ namespace LSFG::Shaderchains {
Core::Image inImg2,
std::optional<Core::Image> optImg1,
std::optional<Core::Image> optImg2,
VkExtent2D outExtent);
VkExtent2D outExtent,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param fc The frame count, used to select the input images.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc);
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass);
/// Get the first output image.
[[nodiscard]] const auto& getOutImage1() const { return this->outImg1; }
@ -74,9 +77,11 @@ namespace LSFG::Shaderchains {
private:
std::array<Core::ShaderModule, 6> shaderModules;
std::array<Core::Pipeline, 6> pipelines;
std::array<Core::DescriptorSet, 5> descriptorSets; // first shader has special logic
std::array<Core::DescriptorSet, 3> specialDescriptorSets;
Core::Buffer buffer;
std::array<Core::DescriptorSet, 3> descriptorSets; // first shader has special logic
std::vector<Core::DescriptorSet> n1DescriptorSets;
std::vector<Core::DescriptorSet> n2DescriptorSets;
std::vector<std::array<Core::DescriptorSet, 3>> nSpecialDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 4> inImgs1_0;
std::array<Core::Image, 4> inImgs1_1;

11
lsfg-vk-gen/include/shaderchains/magic.hpp
View file

@ -35,6 +35,7 @@ namespace LSFG::Shaderchains {
/// @param inImg2 The second input image to process.
/// @param inImg3 The third input image to process, next step up the resolution.
/// @param optImg An optional additional input from the previous pass.
/// @param genc Amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
@ -44,17 +45,19 @@ namespace LSFG::Shaderchains {
std::array<Core::Image, 4> inImgs1_2,
Core::Image inImg2,
Core::Image inImg3,
std::optional<Core::Image> optImg);
std::optional<Core::Image> optImg,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param fc The frame count, used to select the input images.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc);
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass);
/// Get the first set of output images
[[nodiscard]] const auto& getOutImages1() const { return this->outImgs1; }
@ -72,8 +75,8 @@ namespace LSFG::Shaderchains {
private:
Core::ShaderModule shaderModule;
Core::Pipeline pipeline;
std::array<Core::DescriptorSet, 3> descriptorSets;
Core::Buffer buffer;
std::vector<std::array<Core::DescriptorSet, 3>> nDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 4> inImgs1_0;
std::array<Core::Image, 4> inImgs1_1;

17
lsfg-vk-gen/include/shaderchains/merge.hpp
View file

@ -34,7 +34,8 @@ namespace LSFG::Shaderchains {
/// @param inImg3 The first related input texture
/// @param inImg4 The second related input texture
/// @param inImg5 The third related input texture
/// @param outFd File descriptor for the output image.
/// @param outFds File descriptors for the output images.
/// @param genc The amount of frames to generaten.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
@ -44,20 +45,22 @@ namespace LSFG::Shaderchains {
Core::Image inImg3,
Core::Image inImg4,
Core::Image inImg5,
int outFd);
const std::vector<int>& outFds,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param fc The frame count, used to select the input images.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc);
void Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass);
/// Get the output image
[[nodiscard]] const auto& getOutImage() const { return this->outImg; }
[[nodiscard]] const auto& getOutImage(size_t pass) const { return this->outImgs.at(pass); }
/// Trivially copyable, moveable and destructible
Merge(const Merge&) noexcept = default;
@ -68,8 +71,8 @@ namespace LSFG::Shaderchains {
private:
Core::ShaderModule shaderModule;
Core::Pipeline pipeline;
std::array<Core::DescriptorSet, 2> descriptorSets; // one for each input combination
Core::Buffer buffer;
std::vector<std::array<Core::DescriptorSet, 2>> nDescriptorSets; // per combo
std::vector<Core::Buffer> buffers;
Core::Image inImg1;
Core::Image inImg2;
@ -77,7 +80,7 @@ namespace LSFG::Shaderchains {
Core::Image inImg4;
Core::Image inImg5;
Core::Image outImg;
std::vector<Core::Image> outImgs;
};
}

12
lsfg-vk-gen/include/shaderchains/zeta.hpp
View file

@ -32,22 +32,25 @@ namespace LSFG::Shaderchains {
/// @param inImgs1 The first set of input images to process.
/// @param inImg2 The second type image to process.
/// @param inImg3 The third type image to process.
/// @param genc The amount of frames to generate.
///
/// @throws LSFG::vulkan_error if resource creation fails.
///
Zeta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 3> inImgs1,
Core::Image inImg2,
Core::Image inImg3);
Core::Image inImg3,
size_t genc);
///
/// Dispatch the shaderchain.
///
/// @param buf The command buffer to use for dispatching.
/// @param pass The pass number.
///
/// @throws std::logic_error if the command buffer is not recording.
///
void Dispatch(const Core::CommandBuffer& buf);
void Dispatch(const Core::CommandBuffer& buf, uint64_t pass);
/// Get the output image.
[[nodiscard]] const auto& getOutImage() const { return this->outImg; }
@ -61,8 +64,9 @@ namespace LSFG::Shaderchains {
private:
std::array<Core::ShaderModule, 4> shaderModules;
std::array<Core::Pipeline, 4> pipelines;
std::array<Core::DescriptorSet, 4> descriptorSets;
Core::Buffer buffer;
std::array<Core::DescriptorSet, 3> descriptorSets;
std::vector<Core::DescriptorSet> nDescriptorSets;
std::vector<Core::Buffer> buffers;
std::array<Core::Image, 3> inImgs1;
Core::Image inImg2;

10
lsfg-vk-gen/public/lsfg.hpp
View file

@ -3,6 +3,7 @@
#include <stdexcept>
#include <vector>
#include <vulkan/vulkan_core.h>
namespace LSFG {
@ -21,23 +22,24 @@ namespace LSFG {
/// @param height Height of the input images.
/// @param in0 File descriptor for the first input image.
/// @param in1 File descriptor for the second input image.
/// @param out File descriptor for the output image.
/// @param outN File descriptor for each output image. This defines the LSFG level.
/// @return A unique identifier for the created context.
///
/// @throws LSFG::vulkan_error if the context cannot be created.
///
int32_t createContext(uint32_t width, uint32_t height, int in0, int in1, int out);
int32_t createContext(uint32_t width, uint32_t height, int in0, int in1,
const std::vector<int>& outN);
///
/// Present a context.
///
/// @param id Unique identifier of the context to present.
/// @param inSem Semaphore to wait on before starting the generation.
/// @param outSem Semaphore to signal when the generation is complete.
/// @param outSem Semaphores to signal once each output image is ready.
///
/// @throws LSFG::vulkan_error if the context cannot be presented.
///
void presentContext(int32_t id, int inSem, int outSem);
void presentContext(int32_t id, int inSem, const std::vector<int>& outSem);
///
/// Delete an LSFG context.

101
lsfg-vk-gen/src/context.cpp
View file

@ -8,7 +8,8 @@
using namespace LSFG;
Context::Context(const Core::Device& device, uint32_t width, uint32_t height, int in0, int in1, int out) {
Context::Context(const Core::Device& device, uint32_t width, uint32_t height, int in0, int in1,
const std::vector<int>& outN) {
// import images
this->inImg_0 = Core::Image(device, { width, height },
VK_FORMAT_R8G8B8A8_UNORM,
@ -25,16 +26,24 @@ Context::Context(const Core::Device& device, uint32_t width, uint32_t height, in
this->descPool = Core::DescriptorPool(device);
this->cmdPool = Core::CommandPool(device);
// prepare vectors
for (size_t i = 0; i < 8; i++)
this->internalSemaphores.at(i).resize(outN.size());
for (size_t i = 0; i < outN.size(); i++) {
this->outSemaphores.emplace_back();
this->cmdBuffers2.emplace_back();
}
// create shader chains
this->downsampleChain = Shaderchains::Downsample(device, this->descPool,
this->inImg_0, this->inImg_1);
this->inImg_0, this->inImg_1, outN.size());
for (size_t i = 0; i < 7; i++)
this->alphaChains.at(i) = Shaderchains::Alpha(device, this->descPool,
this->downsampleChain.getOutImages().at(i));
this->betaChain = Shaderchains::Beta(device, this->descPool,
this->alphaChains.at(0).getOutImages0(),
this->alphaChains.at(0).getOutImages1(),
this->alphaChains.at(0).getOutImages2());
this->alphaChains.at(0).getOutImages2(), outN.size());
for (size_t i = 0; i < 7; i++) {
if (i < 4) {
this->gammaChains.at(i) = Shaderchains::Gamma(device, this->descPool,
@ -46,7 +55,8 @@ Context::Context(const Core::Device& device, uint32_t width, uint32_t height, in
: std::optional{this->gammaChains.at(i - 1).getOutImage2()},
i == 0 ? std::nullopt
: std::optional{this->gammaChains.at(i - 1).getOutImage1()},
this->alphaChains.at(6 - i - 1).getOutImages0().at(0).getExtent()
this->alphaChains.at(6 - i - 1).getOutImages0().at(0).getExtent(),
outN.size()
);
} else {
this->magicChains.at(i - 4) = Shaderchains::Magic(device, this->descPool,
@ -57,31 +67,36 @@ Context::Context(const Core::Device& device, uint32_t width, uint32_t height, in
: this->extractChains.at(i - 5).getOutImage(),
i == 4 ? this->gammaChains.at(i - 1).getOutImage1()
: this->zetaChains.at(i - 5).getOutImage(),
i == 4 ? std::nullopt : std::optional{this->epsilonChains.at(i - 5).getOutImage()}
i == 4 ? std::nullopt : std::optional{this->epsilonChains.at(i - 5).getOutImage()},
outN.size()
);
this->deltaChains.at(i - 4) = Shaderchains::Delta(device, this->descPool,
this->magicChains.at(i - 4).getOutImages1(),
i == 4 ? std::nullopt
: std::optional{this->deltaChains.at(i - 5).getOutImage()}
: std::optional{this->deltaChains.at(i - 5).getOutImage()},
outN.size()
);
this->epsilonChains.at(i - 4) = Shaderchains::Epsilon(device, this->descPool,
this->magicChains.at(i - 4).getOutImages2(),
this->betaChain.getOutImages().at(6 - i),
i == 4 ? std::nullopt
: std::optional{this->epsilonChains.at(i - 5).getOutImage()}
: std::optional{this->epsilonChains.at(i - 5).getOutImage()},
outN.size()
);
this->zetaChains.at(i - 4) = Shaderchains::Zeta(device, this->descPool,
this->magicChains.at(i - 4).getOutImages3(),
i == 4 ? this->gammaChains.at(i - 1).getOutImage1()
: this->zetaChains.at(i - 5).getOutImage(),
this->betaChain.getOutImages().at(6 - i)
this->betaChain.getOutImages().at(6 - i),
outN.size()
);
if (i >= 6)
continue; // no extract for i >= 6
this->extractChains.at(i - 4) = Shaderchains::Extract(device, this->descPool,
this->zetaChains.at(i - 4).getOutImage(),
this->epsilonChains.at(i - 4).getOutImage(),
this->alphaChains.at(6 - i - 1).getOutImages0().at(0).getExtent()
this->alphaChains.at(6 - i - 1).getOutImages0().at(0).getExtent(),
outN.size()
);
}
}
@ -91,41 +106,61 @@ Context::Context(const Core::Device& device, uint32_t width, uint32_t height, in
this->zetaChains.at(2).getOutImage(),
this->epsilonChains.at(2).getOutImage(),
this->deltaChains.at(2).getOutImage(),
out
outN,
outN.size()
);
}
void Context::present(const Core::Device& device, int inSem, int outSem) {
void Context::present(const Core::Device& device, int inSem,
const std::vector<int>& outSem) {
auto& inSemaphore = this->inSemaphores.at(this->fc % 8);
inSemaphore = Core::Semaphore(device, inSem);
auto& outSemaphore = this->outSemaphores.at(this->fc % 8);
outSemaphore = Core::Semaphore(device, outSem);
auto& internalSemaphores = this->internalSemaphores.at(this->fc % 8);
for (size_t i = 0; i < outSem.size(); i++)
internalSemaphores.at(i) = Core::Semaphore(device);
auto& cmdBuffer = this->cmdBuffers.at(this->fc % 8);
cmdBuffer = Core::CommandBuffer(device, this->cmdPool);
cmdBuffer.begin();
auto& cmdBuffer1 = this->cmdBuffers1.at(this->fc % 8);
cmdBuffer1 = Core::CommandBuffer(device, this->cmdPool);
cmdBuffer1.begin();
this->downsampleChain.Dispatch(cmdBuffer, fc);
this->downsampleChain.Dispatch(cmdBuffer1, fc);
for (size_t i = 0; i < 7; i++)
this->alphaChains.at(6 - i).Dispatch(cmdBuffer, fc);
this->betaChain.Dispatch(cmdBuffer, fc);
for (size_t i = 0; i < 4; i++)
this->gammaChains.at(i).Dispatch(cmdBuffer, fc);
for (size_t i = 0; i < 3; i++) {
this->magicChains.at(i).Dispatch(cmdBuffer, fc);
this->deltaChains.at(i).Dispatch(cmdBuffer);
this->epsilonChains.at(i).Dispatch(cmdBuffer);
this->zetaChains.at(i).Dispatch(cmdBuffer);
if (i < 2)
this->extractChains.at(i).Dispatch(cmdBuffer);
}
this->mergeChain.Dispatch(cmdBuffer, fc);
this->alphaChains.at(6 - i).Dispatch(cmdBuffer1, fc);
cmdBuffer.end();
cmdBuffer1.end();
cmdBuffer.submit(device.getComputeQueue(), std::nullopt,
cmdBuffer1.submit(device.getComputeQueue(), std::nullopt,
{ inSemaphore }, std::nullopt,
{ outSemaphore }, std::nullopt);
internalSemaphores, std::nullopt);
for (size_t pass = 0; pass < outSem.size(); pass++) {
auto& outSemaphore = this->outSemaphores.at(pass).at(this->fc % 8);
outSemaphore = Core::Semaphore(device, outSem.at(pass));
auto& cmdBuffer2 = this->cmdBuffers2.at(pass).at(this->fc % 8);
cmdBuffer2 = Core::CommandBuffer(device, this->cmdPool);
cmdBuffer2.begin();
this->betaChain.Dispatch(cmdBuffer2, fc, pass);
for (size_t i = 0; i < 4; i++)
this->gammaChains.at(i).Dispatch(cmdBuffer2, fc, pass);
for (size_t i = 0; i < 3; i++) {
this->magicChains.at(i).Dispatch(cmdBuffer2, fc, pass);
this->deltaChains.at(i).Dispatch(cmdBuffer2, pass);
this->epsilonChains.at(i).Dispatch(cmdBuffer2, pass);
this->zetaChains.at(i).Dispatch(cmdBuffer2, pass);
if (i < 2)
this->extractChains.at(i).Dispatch(cmdBuffer2, pass);
}
this->mergeChain.Dispatch(cmdBuffer2, fc, pass);
cmdBuffer2.end();
cmdBuffer2.submit(device.getComputeQueue(), std::nullopt,
{ internalSemaphores.at(pass) }, std::nullopt,
{ outSemaphore }, std::nullopt);
}
fc++;
}

7
lsfg-vk-gen/src/lsfg.cpp
View file

@ -28,16 +28,17 @@ void LSFG::initialize() {
std::srand(static_cast<uint32_t>(std::time(nullptr)));
}
int32_t LSFG::createContext(uint32_t width, uint32_t height, int in0, int in1, int out) {
int32_t LSFG::createContext(uint32_t width, uint32_t height, int in0, int in1,
const std::vector<int>& outN) {
if (!instance.has_value() || !device.has_value())
throw LSFG::vulkan_error(VK_ERROR_INITIALIZATION_FAILED, "LSFG not initialized");
auto id = std::rand();
contexts.emplace(id, Context(*device, width, height, in0, in1, out));
contexts.emplace(id, Context(*device, width, height, in0, in1, outN));
return id;
}
void LSFG::presentContext(int32_t id, int inSem, int outSem) {
void LSFG::presentContext(int32_t id, int inSem, const std::vector<int>& outSem) {
if (!instance.has_value() || !device.has_value())
throw LSFG::vulkan_error(VK_ERROR_INITIALIZATION_FAILED, "LSFG not initialized");

32
lsfg-vk-gen/src/shaderchains/beta.cpp
View file

@ -6,7 +6,8 @@ using namespace LSFG::Shaderchains;
Beta::Beta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 4> inImgs_0,
std::array<Core::Image, 4> inImgs_1,
std::array<Core::Image, 4> inImgs_2)
std::array<Core::Image, 4> inImgs_2,
size_t genc)
: inImgs_0(std::move(inImgs_0)),
inImgs_1(std::move(inImgs_1)),
inImgs_2(std::move(inImgs_2)) {
@ -36,14 +37,21 @@ Beta::Beta(const Core::Device& device, const Core::DescriptorPool& pool,
for (size_t i = 0; i < 5; i++) {
this->pipelines.at(i) = Core::Pipeline(device,
this->shaderModules.at(i));
if (i == 0) continue; // first shader has special logic
if (i == 0 || i == 4) continue; // first shader has special logic
this->descriptorSets.at(i - 1) = Core::DescriptorSet(device, pool,
this->shaderModules.at(i));
}
for (size_t i = 0; i < 3; i++)
this->specialDescriptorSets.at(i) = Core::DescriptorSet(device, pool,
this->shaderModules.at(0));
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->nDescriptorSets.emplace_back(device, pool,
this->shaderModules.at(4));
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
const auto extent = this->inImgs_0.at(0).getExtent();
@ -104,15 +112,17 @@ Beta::Beta(const Core::Device& device, const Core::DescriptorPool& pool,
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs2)
.build();
this->descriptorSets.at(3).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
void Beta::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
void Beta::Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass) {
const auto extent = this->tempImgs1.at(0).getExtent();
// first pass
@ -170,6 +180,6 @@ void Beta::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.build();
this->pipelines.at(4).bind(buf);
this->descriptorSets.at(3).bind(buf, this->pipelines.at(4));
this->nDescriptorSets.at(pass).bind(buf, this->pipelines.at(4));
buf.dispatch(threadsX, threadsY, 1);
}

39
lsfg-vk-gen/src/shaderchains/delta.cpp
View file

@ -5,7 +5,8 @@ using namespace LSFG::Shaderchains;
Delta::Delta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 2> inImgs,
std::optional<Core::Image> optImg)
std::optional<Core::Image> optImg,
size_t genc)
: inImgs(std::move(inImgs)),
optImg(std::move(optImg)) {
this->shaderModules = {{
@ -30,13 +31,19 @@ Delta::Delta(const Core::Device& device, const Core::DescriptorPool& pool,
for (size_t i = 0; i < 4; i++) {
this->pipelines.at(i) = Core::Pipeline(device,
this->shaderModules.at(i));
if (i == 3) continue;
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool,
this->shaderModules.at(i));
}
Globals::FgBuffer data = Globals::fgBuffer;
data.firstIterS = !this->optImg.has_value();
this->buffer = Core::Buffer(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->nDescriptorSets.emplace_back(device, pool,
this->shaderModules.at(3));
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
data.firstIterS = !this->optImg.has_value();
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
const auto extent = this->inImgs.at(0).getExtent();
@ -74,17 +81,19 @@ Delta::Delta(const Core::Device& device, const Core::DescriptorPool& pool,
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1)
.build();
this->descriptorSets.at(3).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
void Delta::Dispatch(const Core::CommandBuffer& buf) {
void Delta::Dispatch(const Core::CommandBuffer& buf, uint64_t pass) {
const auto extent = this->tempImgs1.at(0).getExtent();
// first pass
@ -128,6 +137,6 @@ void Delta::Dispatch(const Core::CommandBuffer& buf) {
.build();
this->pipelines.at(3).bind(buf);
this->descriptorSets.at(3).bind(buf, this->pipelines.at(3));
this->nDescriptorSets.at(pass).bind(buf, this->pipelines.at(3));
buf.dispatch(threadsX, threadsY, 1);
}

8
lsfg-vk-gen/src/shaderchains/downsample.cpp
View file

@ -4,7 +4,8 @@
using namespace LSFG::Shaderchains;
Downsample::Downsample(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg_0, Core::Image inImg_1)
Core::Image inImg_0, Core::Image inImg_1,
size_t genc)
: inImg_0(std::move(inImg_0)),
inImg_1(std::move(inImg_1)) {
this->shaderModule = Core::ShaderModule(device, "rsc/shaders/downsample.spv",
@ -15,7 +16,10 @@ Downsample::Downsample(const Core::Device& device, const Core::DescriptorPool& p
this->pipeline = Core::Pipeline(device, this->shaderModule);
for (size_t i = 0; i < 2; i++)
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool, this->shaderModule);
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
auto data = Globals::fgBuffer;
data.timestamp = 1.0F / static_cast<float>(genc + 1);
this->buffer = Core::Buffer(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
auto extent = this->inImg_0.getExtent();
for (size_t i = 0; i < 7; i++)

37
lsfg-vk-gen/src/shaderchains/epsilon.cpp
View file

@ -6,7 +6,8 @@ using namespace LSFG::Shaderchains;
Epsilon::Epsilon(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 3> inImgs1,
Core::Image inImg2,
std::optional<Core::Image> optImg)
std::optional<Core::Image> optImg,
size_t genc)
: inImgs1(std::move(inImgs1)),
inImg2(std::move(inImg2)),
optImg(std::move(optImg)) {
@ -32,10 +33,18 @@ Epsilon::Epsilon(const Core::Device& device, const Core::DescriptorPool& pool,
for (size_t i = 0; i < 4; i++) {
this->pipelines.at(i) = Core::Pipeline(device,
this->shaderModules.at(i));
if (i == 3) continue;
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool,
this->shaderModules.at(i));
}
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->nDescriptorSets.emplace_back(device, pool,
this->shaderModules.at(3));
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
const auto extent = this->inImgs1.at(0).getExtent();
@ -73,18 +82,20 @@ Epsilon::Epsilon(const Core::Device& device, const Core::DescriptorPool& pool,
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1)
.build();
this->descriptorSets.at(3).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
void Epsilon::Dispatch(const Core::CommandBuffer& buf) {
void Epsilon::Dispatch(const Core::CommandBuffer& buf, uint64_t pass) {
const auto extent = this->tempImgs1.at(0).getExtent();
// first pass
@ -129,6 +140,6 @@ void Epsilon::Dispatch(const Core::CommandBuffer& buf) {
.build();
this->pipelines.at(3).bind(buf);
this->descriptorSets.at(3).bind(buf, this->pipelines.at(3));
this->nDescriptorSets.at(pass).bind(buf, this->pipelines.at(3));
buf.dispatch(threadsX, threadsY, 1);
}

37
lsfg-vk-gen/src/shaderchains/extract.cpp
View file

@ -6,7 +6,8 @@ using namespace LSFG::Shaderchains;
Extract::Extract(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg1,
Core::Image inImg2,
VkExtent2D outExtent)
VkExtent2D outExtent,
size_t genc)
: inImg1(std::move(inImg1)),
inImg2(std::move(inImg2)) {
this->shaderModule = Core::ShaderModule(device, "rsc/shaders/extract.spv",
@ -15,8 +16,14 @@ Extract::Extract(const Core::Device& device, const Core::DescriptorPool& pool,
{ 1, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE },
{ 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER } });
this->pipeline = Core::Pipeline(device, this->shaderModule);
this->descriptorSet = Core::DescriptorSet(device, pool, this->shaderModule);
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->nDescriptorSets.emplace_back(device, pool,
this->shaderModule);
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
this->whiteImg = Core::Image(device,
outExtent,
@ -30,21 +37,23 @@ Extract::Extract(const Core::Device& device, const Core::DescriptorPool& pool,
VK_IMAGE_USAGE_STORAGE_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
VK_IMAGE_ASPECT_COLOR_BIT);
this->descriptorSet.update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->whiteImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->whiteImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
// clear white image
Utils::clearImage(device, this->whiteImg, true);
}
void Extract::Dispatch(const Core::CommandBuffer& buf) {
void Extract::Dispatch(const Core::CommandBuffer& buf, uint64_t pass) {
auto extent = this->whiteImg.getExtent();
// first pass
@ -59,6 +68,6 @@ void Extract::Dispatch(const Core::CommandBuffer& buf) {
.build();
this->pipeline.bind(buf);
this->descriptorSet.bind(buf, this->pipeline);
this->nDescriptorSets.at(pass).bind(buf, this->pipeline);
buf.dispatch(threadsX, threadsY, 1);
}

100
lsfg-vk-gen/src/shaderchains/gamma.cpp
View file

@ -10,7 +10,8 @@ Gamma::Gamma(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg2,
std::optional<Core::Image> optImg1, // NOLINT
std::optional<Core::Image> optImg2,
VkExtent2D outExtent)
VkExtent2D outExtent,
size_t genc)
: inImgs1_0(std::move(inImgs1_0)),
inImgs1_1(std::move(inImgs1_1)),
inImgs1_2(std::move(inImgs1_2)),
@ -48,17 +49,28 @@ Gamma::Gamma(const Core::Device& device, const Core::DescriptorPool& pool,
for (size_t i = 0; i < 6; i++) {
this->pipelines.at(i) = Core::Pipeline(device,
this->shaderModules.at(i));
if (i == 0) continue; // first shader has special logic
if (i == 0 || i >= 4) continue; // first shader has special logic
this->descriptorSets.at(i - 1) = Core::DescriptorSet(device, pool,
this->shaderModules.at(i));
}
for (size_t i = 0; i < 3; i++)
this->specialDescriptorSets.at(i) = Core::DescriptorSet(device, pool,
this->shaderModules.at(0));
Globals::FgBuffer data = Globals::fgBuffer;
data.firstIter = !optImg1.has_value();
this->buffer = Core::Buffer(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->n1DescriptorSets.emplace_back(device, pool,
this->shaderModules.at(4));
for (size_t i = 0; i < genc; i++)
this->n2DescriptorSets.emplace_back(device, pool,
this->shaderModules.at(5));
for (size_t i = 0; i < genc; i++) {
this->nSpecialDescriptorSets.emplace_back();
for (size_t j = 0; j < 3; j++)
this->nSpecialDescriptorSets.at(i).at(j) = Core::DescriptorSet(device, pool,
this->shaderModules.at(0));
}
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
data.firstIter = !optImg1.has_value();
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
const auto extent = this->inImgs1_0.at(0).getExtent();
@ -107,18 +119,20 @@ Gamma::Gamma(const Core::Device& device, const Core::DescriptorPool& pool,
nextImgs1 = &this->inImgs1_2;
prevImgs1 = &this->inImgs1_1;
}
this->specialDescriptorSets.at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *prevImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *nextImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(0))
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(1))
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(2))
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nSpecialDescriptorSets.at(i).at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *prevImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *nextImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(0))
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(1))
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1.at(2))
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
this->descriptorSets.at(0).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
@ -137,23 +151,25 @@ Gamma::Gamma(const Core::Device& device, const Core::DescriptorPool& pool,
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs2)
.build();
this->descriptorSets.at(3).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg1)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
this->descriptorSets.at(4).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->whiteImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->outImg1)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg2)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->n1DescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg1)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
this->n2DescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->whiteImg)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->outImg1)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg2)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
// clear white image and optImg1 if needed
Utils::clearImage(device, this->whiteImg, true);
@ -161,7 +177,7 @@ Gamma::Gamma(const Core::Device& device, const Core::DescriptorPool& pool,
Utils::clearImage(device, this->optImg1);
}
void Gamma::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
void Gamma::Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass) {
const auto extent = this->tempImgs1.at(0).getExtent();
// first pass
@ -188,7 +204,7 @@ void Gamma::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.build();
this->pipelines.at(0).bind(buf);
this->specialDescriptorSets.at(fc % 3).bind(buf, this->pipelines.at(0));
this->nSpecialDescriptorSets.at(pass).at(fc % 3).bind(buf, this->pipelines.at(0));
buf.dispatch(threadsX, threadsY, 1);
// second pass
@ -232,7 +248,7 @@ void Gamma::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.build();
this->pipelines.at(4).bind(buf);
this->descriptorSets.at(3).bind(buf, this->pipelines.at(4));
this->n1DescriptorSets.at(pass).bind(buf, this->pipelines.at(4));
buf.dispatch(threadsX, threadsY, 1);
// sixth pass
@ -246,6 +262,6 @@ void Gamma::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.build();
this->pipelines.at(5).bind(buf);
this->descriptorSets.at(4).bind(buf, this->pipelines.at(5));
this->n2DescriptorSets.at(pass).bind(buf, this->pipelines.at(5));
buf.dispatch(threadsX, threadsY, 1);
}

52
lsfg-vk-gen/src/shaderchains/magic.cpp
View file

@ -9,7 +9,8 @@ Magic::Magic(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 4> inImgs1_2,
Core::Image inImg2,
Core::Image inImg3,
std::optional<Core::Image> optImg)
std::optional<Core::Image> optImg,
size_t genc)
: inImgs1_0(std::move(inImgs1_0)),
inImgs1_1(std::move(inImgs1_1)),
inImgs1_2(std::move(inImgs1_2)),
@ -21,12 +22,17 @@ Magic::Magic(const Core::Device& device, const Core::DescriptorPool& pool,
{ 3+3+2, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE },
{ 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER } });
this->pipeline = Core::Pipeline(device, this->shaderModule);
for (size_t i = 0; i < 3; i++)
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool, this->shaderModule);
Globals::FgBuffer data = Globals::fgBuffer;
data.firstIterS = !this->optImg.has_value();
this->buffer = Core::Buffer(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.emplace_back();
for (size_t j = 0; j < 3; j++)
this->nDescriptorSets.at(i).at(j) = Core::DescriptorSet(device, pool, this->shaderModule);
}
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
data.firstIterS = !this->optImg.has_value();
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
auto extent = this->inImgs1_0.at(0).getExtent();
@ -59,23 +65,25 @@ Magic::Magic(const Core::Device& device, const Core::DescriptorPool& pool,
nextImgs1 = &this->inImgs1_2;
prevImgs1 = &this->inImgs1_1;
}
this->descriptorSets.at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *prevImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *nextImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs3)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs1)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *prevImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, *nextImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->optImg)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs3)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs1)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
}
void Magic::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
void Magic::Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass) {
auto extent = this->inImgs1_0.at(0).getExtent();
// first pass
@ -103,6 +111,6 @@ void Magic::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.build();
this->pipeline.bind(buf);
this->descriptorSets.at(fc % 3).bind(buf, this->pipeline);
this->nDescriptorSets.at(pass).at(fc % 3).bind(buf, this->pipeline);
buf.dispatch(threadsX, threadsY, 1);
}

59
lsfg-vk-gen/src/shaderchains/merge.cpp
View file

@ -9,7 +9,8 @@ Merge::Merge(const Core::Device& device, const Core::DescriptorPool& pool,
Core::Image inImg3,
Core::Image inImg4,
Core::Image inImg5,
int outFd)
const std::vector<int>& outFds,
size_t genc)
: inImg1(std::move(inImg1)),
inImg2(std::move(inImg2)),
inImg3(std::move(inImg3)),
@ -21,35 +22,45 @@ Merge::Merge(const Core::Device& device, const Core::DescriptorPool& pool,
{ 1, VK_DESCRIPTOR_TYPE_STORAGE_IMAGE },
{ 1, VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER } });
this->pipeline = Core::Pipeline(device, this->shaderModule);
for (size_t i = 0; i < 2; i++)
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool, this->shaderModule);
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.emplace_back();
for (size_t j = 0; j < 2; j++)
this->nDescriptorSets.at(i).at(j) = Core::DescriptorSet(device, pool, this->shaderModule);
}
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
auto extent = this->inImg1.getExtent();
this->outImg = Core::Image(device,
extent,
VK_FORMAT_R8G8B8A8_UNORM,
VK_IMAGE_USAGE_STORAGE_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
VK_IMAGE_ASPECT_COLOR_BIT,
outFd);
for (size_t i = 0; i < genc; i++)
this->outImgs.emplace_back(device,
extent,
VK_FORMAT_R8G8B8A8_UNORM,
VK_IMAGE_USAGE_STORAGE_BIT | VK_IMAGE_USAGE_SAMPLED_BIT,
VK_IMAGE_ASPECT_COLOR_BIT,
outFds.at(i));
for (size_t fc = 0; fc < 2; fc++) {
this->descriptorSets.at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, (fc % 2 == 0) ? this->inImg1 : this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, (fc % 2 == 0) ? this->inImg2 : this->inImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg4)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg5)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).at(fc).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, (fc % 2 == 0) ? this->inImg1 : this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, (fc % 2 == 0) ? this->inImg2 : this->inImg1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg4)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg5)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImgs.at(i))
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i))
.build();
}
}
}
void Merge::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
void Merge::Dispatch(const Core::CommandBuffer& buf, uint64_t fc, uint64_t pass) {
auto extent = this->inImg1.getExtent();
// first pass
@ -62,10 +73,10 @@ void Merge::Dispatch(const Core::CommandBuffer& buf, uint64_t fc) {
.addW2R(this->inImg3)
.addW2R(this->inImg4)
.addW2R(this->inImg5)
.addR2W(this->outImg)
.addR2W(this->outImgs.at(pass))
.build();
this->pipeline.bind(buf);
this->descriptorSets.at(fc % 2).bind(buf, this->pipeline);
this->nDescriptorSets.at(pass).at(fc % 2).bind(buf, this->pipeline);
buf.dispatch(threadsX, threadsY, 1);
}

37
lsfg-vk-gen/src/shaderchains/zeta.cpp
View file

@ -6,7 +6,8 @@ using namespace LSFG::Shaderchains;
Zeta::Zeta(const Core::Device& device, const Core::DescriptorPool& pool,
std::array<Core::Image, 3> inImgs1,
Core::Image inImg2,
Core::Image inImg3)
Core::Image inImg3,
size_t genc)
: inImgs1(std::move(inImgs1)),
inImg2(std::move(inImg2)),
inImg3(std::move(inImg3)) {
@ -32,10 +33,18 @@ Zeta::Zeta(const Core::Device& device, const Core::DescriptorPool& pool,
for (size_t i = 0; i < 4; i++) {
this->pipelines.at(i) = Core::Pipeline(device,
this->shaderModules.at(i));
if (i == 3) continue;
this->descriptorSets.at(i) = Core::DescriptorSet(device, pool,
this->shaderModules.at(i));
}
this->buffer = Core::Buffer(device, Globals::fgBuffer, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
for (size_t i = 0; i < genc; i++)
this->nDescriptorSets.emplace_back(device, pool,
this->shaderModules.at(3));
for (size_t i = 0; i < genc; i++) {
auto data = Globals::fgBuffer;
data.timestamp = static_cast<float>(i + 1) / static_cast<float>(genc + 1);
this->buffers.emplace_back(device, data, VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT);
}
const auto extent = this->inImgs1.at(0).getExtent();
@ -73,18 +82,20 @@ Zeta::Zeta(const Core::Device& device, const Core::DescriptorPool& pool,
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs2)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->tempImgs1)
.build();
this->descriptorSets.at(3).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffer)
.build();
for (size_t i = 0; i < genc; i++) {
this->nDescriptorSets.at(i).update(device)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampBorder)
.add(VK_DESCRIPTOR_TYPE_SAMPLER, Globals::samplerClampEdge)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->tempImgs1)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg2)
.add(VK_DESCRIPTOR_TYPE_SAMPLED_IMAGE, this->inImg3)
.add(VK_DESCRIPTOR_TYPE_STORAGE_IMAGE, this->outImg)
.add(VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER, this->buffers.at(i) )
.build();
}
}
void Zeta::Dispatch(const Core::CommandBuffer& buf) {
void Zeta::Dispatch(const Core::CommandBuffer& buf, uint64_t pass) {
const auto extent = this->tempImgs1.at(0).getExtent();
// first pass
@ -129,6 +140,6 @@ void Zeta::Dispatch(const Core::CommandBuffer& buf) {
.build();
this->pipelines.at(3).bind(buf);
this->descriptorSets.at(3).bind(buf, this->pipelines.at(3));
this->nDescriptorSets.at(pass).bind(buf, this->pipelines.at(3));
buf.dispatch(threadsX, threadsY, 1);
}