Used to manage synchronisation between tasks on async compute queues and the graphics queue.
Not all platforms support GPU fences. See SystemInfo.supportsGPUFence.
A GPUFence represents a point during GPU processing after a specific compute shader dispatch or draw call has completed. It can be used to achieve synchronisation between tasks running on the async compute queues or the graphics queue by having one or more queues wait until a given fence has passed. This is an important consideration when working with async compute as which tasks are running at the same time on the graphics queue and the async compute queues is key to improving GPU performance.
GPUFence's do not need to be used to synchronise where one GPU task is writing to a resource that will be read as an input by another. These resource dependencies are automatically handled by Unity.
GPUFences should be created via Graphics.CreateGPUFence or CommandBuffer.CreateGPUFence attempting to use a GPUFence that has not been created via one of these functions will result in an exception.
It is possible to create circular dependencies using GPUFences that, if executed, would deadlock the GPU. Unity will detect such circular dependencies in the Editor and raise exceptions if any exist after calls to Graphics.CreateGPUFence, Graphics.WaitOnGPUFence, Graphics.ExecuteCommandBuffer, Graphics.ExecuteCommandBufferAsync, ScriptableRenderContext.ExecuteCommandBuffer, ScriptableRenderContext.ExecuteCommandBufferAsync.
See Also: SystemInfo.supportsGPUFence, Graphics.CreateGPUFence, Graphics.WaitOnGPUFence, CommandBuffer.CreateGPUFence, CommandBuffer.WaitOnGPUFence, Graphics.ExecuteCommandBuffer, Graphics.ExecuteCommandBufferAsync, [ScriptableRenderContext.ExecuteCommandBuffer]], ScriptableRenderContext.ExecuteCommandBufferAsync.
| passed | Has the GPUFence passed?Allows for CPU determination of whether the GPU has passed the point in its processing represented by the GPUFence. |
Did you find this page useful? Please give it a rating: