cudnn Backend API注意事项

一、在包含多个节点的图中，不支持in-place node。（如果图只包含一个节点，支持in-place node）

Note that graphs with more than one operation node do not support in-place operations (that is, where any of the input UIDs matches any of the output UIDs). Such in-place operations are considered cyclic in later graph analysis and deemed unsupported. In-place operations are supported for single-node graphs.

二、op的存放顺序无关紧要，因为cudnn可以推理出图的拓扑结构。

Also note that the operation descriptors can be created and passed into cuDNN in any order, as the tensor UIDs are enough to determine the dependencies in the graph.

几类融合算子的限制

https://docs.nvidia.com/deeplearning/cudnn/developer-guide/index.html#op-specific-contraints-runtime-fusion-engine