DRQ什么时候调度Node去填写dependency
::Pipeline调度Node的sequenceId 0执行
Pipeline::ProcessRequest()
{
for (UINT nodeIndex = 0; nodeIndex < m_orderedNodeCount ; nodeIndex++)
m_pDeferredRequestQueue->AddDeferredNode(requestId, m_ppOrderedNodes[nodeIndex], NULL);//最后一个参数pDependencyUnit为NULL
m_pDeferredRequestQueue->DispatchReadyNodes();
}
TIPS:每个node seq id 0基本都是填写自己的depependency
添加deferred node流程图:
Node内部的ProcessSequenceId图
如何填写dependency
如何填写property dependency,以IPE NODE为例
如何填写fence dependency(camxnode),以IPE NODE为例
如何填写fence dependency(chi node),主要为chinodewarpper调用chinode
如何填写chi fence dependency图,同样主要为chinodewarpper调用chinode
Chi Fence(non-buffer) API调用详解图
Chi Fence(non-buffer) Create举例(EISV2)
Chi Fence(non-buffer) 注册async callback
Chi Fence(non-buffer) Signal举例(EISV2)
CHI Fence (non-buffer) callback flow
CSL Fence(buffer) async/callback详解
CSL Fence 注册async callback
CSL Fence callback
TIPS: SinkOutput :通知session直接返回给fwk targetBuffer
NonSinkOutput :通知DRQ 更新依赖, 继续连接到下一个node
Metadata/property update 详解
Property/metadata callback flow