Navigation
API > API/Runtime > API/Runtime/RHI > API/Runtime/RHI/FRHIGPUMemoryReadback
Syntax
uint32 LastLockGPUIndex
Remarks
We need to separately track which GPU buffer was last locked. It's possible for a new copy operation to be enqueued (writing to LastCopyGPUMask) while the buffer is technically locked, with the unlock and enqueued copy on the GPU itself happening later, during pass execution in FRDGBuilder::Execute (for example, this happens with Nanite streaming). It's not unsafe, because the operations are occurring in order on both the render thread and later pass Execute, but our locking logic needs to handle that scenario.