MayaFlux::Core::VKComputePipeline Class Reference

Wrapper for Vulkan compute pipeline with simplified interface. More...

#include <VKComputePipeline.hpp>

Collaboration diagram for MayaFlux::Core::VKComputePipeline:

Public Member Functions
	VKComputePipeline ()=default
	~VKComputePipeline ()
	VKComputePipeline (const VKComputePipeline &)=delete
VKComputePipeline &	operator= (const VKComputePipeline &)=delete
	VKComputePipeline (VKComputePipeline &&) noexcept
VKComputePipeline &	operator= (VKComputePipeline &&) noexcept
bool	create (vk::Device device, const ComputePipelineConfig &config)
	Create compute pipeline from configuration.
void	cleanup (vk::Device device)
	Cleanup pipeline resources.
vk::Pipeline	get () const
	Get raw Vulkan pipeline handle.
vk::PipelineLayout	get_layout () const
	Get pipeline layout handle.
void	bind (vk::CommandBuffer cmd) const
	Bind pipeline to command buffer.
void	bind_descriptor_sets (vk::CommandBuffer cmd, const std::vector< vk::DescriptorSet > &descriptor_sets, uint32_t first_set=0, const std::vector< uint32_t > &dynamic_offsets={}) const
	Bind descriptor sets to pipeline.
void	push_constants (vk::CommandBuffer cmd, vk::ShaderStageFlags stage_flags, uint32_t offset, uint32_t size, const void *data) const
	Update push constants.
void	dispatch (vk::CommandBuffer cmd, uint32_t group_count_x, uint32_t group_count_y, uint32_t group_count_z) const
	Dispatch compute workgroups.
void	dispatch_1d (vk::CommandBuffer cmd, uint32_t element_count, uint32_t local_size_x) const
	Dispatch compute workgroups with automatic calculation.
void	dispatch_2d (vk::CommandBuffer cmd, uint32_t width_elements, uint32_t height_elements, uint32_t local_size_x, uint32_t local_size_y) const
	Dispatch compute workgroups in 2D with automatic calculation.
void	dispatch_3d (vk::CommandBuffer cmd, uint32_t width_elements, uint32_t height_elements, uint32_t depth_elements, uint32_t local_size_x, uint32_t local_size_y, uint32_t local_size_z) const
	Dispatch compute workgroups in 3D with automatic calculation.
std::optional< std::array< uint32_t, 3 > >	get_workgroup_size () const
	Get shader workgroup size from reflection.
bool	create_specialized (vk::Device device, const ComputePipelineConfig &config, const std::unordered_map< uint32_t, uint32_t > &specialization_data)
	Create pipeline with specialization constants.
template<typename T >
void	push_constants_typed (vk::CommandBuffer cmd, const T &data)
	Push constants with type safety.
void	dispatch_indirect (vk::CommandBuffer cmd, vk::Buffer buffer, vk::DeviceSize offset=0)
	Dispatch compute shader indirectly from GPU buffer.
bool	is_valid () const
	Get pipeline statistics (if available)
const ShaderReflection &	get_shader_reflection () const
	Get shader module reflection data.

Static Public Member Functions
static std::array< uint32_t, 3 >	calculate_dispatch_1d (uint32_t element_count, uint32_t workgroup_size)
	Calculate dispatch size from element count.
static std::array< uint32_t, 3 >	calculate_dispatch_2d (uint32_t width, uint32_t height, uint32_t workgroup_x, uint32_t workgroup_y)
	Calculate 2D dispatch size.
static std::array< uint32_t, 3 >	calculate_dispatch_3d (uint32_t width, uint32_t height, uint32_t depth, uint32_t workgroup_x, uint32_t workgroup_y, uint32_t workgroup_z)
	Calculate 3D dispatch size.

Private Member Functions
vk::PipelineLayout	create_pipeline_layout (vk::Device device, const ComputePipelineConfig &config)
	Create pipeline layout from configuration.

Static Private Member Functions
static uint32_t	calculate_workgroups (uint32_t element_count, uint32_t workgroup_size)
	Calculate number of workgroups needed.

Private Attributes
vk::Pipeline	m_pipeline = nullptr
vk::PipelineLayout	m_layout = nullptr
std::shared_ptr< VKShaderModule >	m_shader
std::optional< std::array< uint32_t, 3 > >	m_workgroup_size

Detailed Description

Wrapper for Vulkan compute pipeline with simplified interface.

Responsibilities:

• Create compute pipeline from shader and configuration
• Manage pipeline layout (descriptor sets + push constants)
• Bind pipeline to command buffer
• Bind descriptor sets
• Update push constants
• Dispatch compute workgroups
• Handle pipeline recreation (for hot-reload)

Does NOT handle:

• Shader compilation (that's VKShaderModule)
• Descriptor allocation (that's VKDescriptorManager)
• Command buffer management (that's VKCommandManager)
• Synchronization (that's the caller's responsibility)

Design:

• Immutable after creation (recreation required for changes)
• Lightweight wrapper around vk::Pipeline
• Can be used directly or via VKComputeProcessor

Thread Safety:

• NOT thread-safe - caller must synchronize access
• Safe to use same pipeline on different command buffers (sequentially)

Definition at line 85 of file VKComputePipeline.hpp.

---

The documentation for this class was generated from the following files:

• VKComputePipeline.hpp
• VKComputePipeline.cpp