jtuyls commented on a change in pull request #14: URL: https://github.com/apache/tvm-rfcs/pull/14#discussion_r683744744
########## File path: rfcs/0012-pipeline-executor.md ########## @@ -0,0 +1,367 @@ +<!--- Licensed to the Apache Software Foundation (ASF) under one --> +<!--- or more contributor license agreements. See the NOTICE file --> +<!--- distributed with this work for additional information --> +<!--- regarding copyright ownership. The ASF licenses this file --> +<!--- to you under the Apache License, Version 2.0 (the --> +<!--- "License"); you may not use this file except in compliance --> +<!--- with the License. You may obtain a copy of the License at --> + +<!--- http://www.apache.org/licenses/LICENSE-2.0 --> + +<!--- Unless required by applicable law or agreed to in writing, --> +<!--- software distributed under the License is distributed on an --> +<!--- "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY --> +<!--- KIND, either express or implied. See the License for the --> +<!--- specific language governing permissions and limitations --> +<!--- under the License. --> +- Feature Name: (fill me in with a unique identifier, `my_awesome_feature`) +- Start Date: (fill me in with today's date, YYYY-MM-DD) +- RFC PR: [apache/tvm-rfcs#0014](https://github.com/apache/tvm-rfcs/pull/0014) +- GitHub Issue: [apache/tvm#8596](https://github.com/apache/tvm/issues/8596) + +## 1. Summary + + +This proposal introduces Pipeline Executor: A runtime executor that by scheduling +splitted subgraph of relay graph in pipeline to implement task level parallism to +reduce compute latency. Review comment: With `not being processed together` I meant that both data inputs are not in the same stage at the same time, which is what is meant with the batch size. You could also execute in a pipelined manner with a batch size > 1, which would mean that you process multiple data inputs at every stage at the same time. So executing with a certain batch size is orthogonal to executing in a pipelined manner and latency refers to the time it takes to process a certain batch of data (typically size 1), irrespective of how you process it (pipelined or otherwise). It's indeed a terminology issue and I think it's best to use the common meaning to avoid confusion, so I think using `increase throughput` instead of `reduce latency` would be better here. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
