renderer::sync::BarrierBatch Class Reference

Skips a barrier only when neither side writes and the layout and next read stages are already covered. More...

#include <gpu_sync.hpp>

Public Member Functions
void	image (ImageState &state, Usage next)
	No op for an image that does not exist. The widest usage until the next write avoids read to read barriers.
void	image (vk::Image image, const vk::ImageSubresourceRange &range, Usage previous, Usage next)
void	buffer (vk::Buffer buffer, Usage previous, Usage next, vk::DeviceSize offset=0, vk::DeviceSize size=VK_WHOLE_SIZE)
void	memory (Usage previous, Usage next)
void	release (vk::Image image, const vk::ImageSubresourceRange &range, Usage previous, Usage next, const QueueHandoff &handoff)
void	acquire (vk::Image image, const vk::ImageSubresourceRange &range, Usage previous, Usage next, const QueueHandoff &handoff)
void	release (vk::Buffer buffer, Usage previous, Usage next, const QueueHandoff &handoff, vk::DeviceSize offset=0, vk::DeviceSize size=VK_WHOLE_SIZE)
void	acquire (vk::Buffer buffer, Usage next, const QueueHandoff &handoff, vk::DeviceSize offset=0, vk::DeviceSize size=VK_WHOLE_SIZE)
void	flush (vk::CommandBuffer cmd)
auto	empty () const -> bool
auto	pendingImages () const -> std::span< const vk::ImageMemoryBarrier2 >
auto	pendingBuffers () const -> std::span< const vk::BufferMemoryBarrier2 >
auto	pendingMemory () const -> std::span< const vk::MemoryBarrier2 >

Detailed Description

Skips a barrier only when neither side writes and the layout and next read stages are already covered.

Member Function Documentation

acquire()

void renderer::sync::BarrierBatch::acquire	(	vk::Image	image,
		const vk::ImageSubresourceRange &	range,
		Usage	previous,
		Usage	next,
		const QueueHandoff &	handoff )

On the consuming queue after the release fence was waited on. previous and next must match the release exactly and nothing is recorded within one family

Here is the caller graph for this function:

release()

void renderer::sync::BarrierBatch::release	(	vk::Image	image,
		const vk::ImageSubresourceRange &	range,
		Usage	previous,
		Usage	next,
		const QueueHandoff &	handoff )

On the writing queue. A plain transition within one family. Across families next may name stages this queue lacks since the acquire waits for them

---

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

• engine/src/toast/renderer/gpu_sync.hpp
• engine/src/toast/renderer/gpu_sync.cpp