Tuvie commented on issue #1995: URL: https://github.com/apache/incubator-brpc/issues/1995#issuecomment-1318269408
> @Tuvie We use `append_user_data` in tensorflow's `AsyncOpKernel::ComputeAsync`. The memory comes from tensor in TensorFlow OP. It's allocated by the framework on each run. We could use `IOBuf::append`, but `append` introduces an extra copy. So what do you think if we add a new parameter for append_user_data function to specify the lkey directly? In this case we don't need to lookup the lkey in an inefficient way. But you have to manage the lkey as a corresponding buffer id by yourself. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
