From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Helen Pang" Subject: Re: [RFC] iovec in ->aio_read/->aio_write Date: Wed, 16 Oct 2002 09:40:34 -0500 Sender: owner-linux-aio@kvack.org Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Benjamin LaHaise , Christoph Hellwig , akpm@digeo.com, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, linux-aio@kvack.org, janetinc@beaverton.ibm.com Return-path: To: Shailabh Nagar List-Id: linux-fsdevel.vger.kernel.org Shailabh Nagar wrote: > It would be interesting to see the performance boost when events > are retrieved at once, using the min_nr parameter of io_getevents My experience is that specified minimum number of events (min_nr) of io_getevents is not quite working yet in kernel 2.4.19. I haven't started to exercise this in 2.5.42, but if it is working, logically it will help the performance indeed. -Helen Shailabh Nagar @kvack.org on 10/16/2002 08:41:03 AM Sent by: owner-linux-aio@kvack.org To: janetinc@beaverton.ibm.com cc: Benjamin LaHaise , Christoph Hellwig , akpm@digeo.com, linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, linux-aio@kvack.org Subject: Re: [RFC] iovec in ->aio_read/->aio_write Janet Morgan wrote: > Here's a patch for aio readv/writev support. Basically it adds: > > - two new opcodes (IOCB_CMD_PREADV and IOCB_CMD_PWRITEV) > - a field to the iocb for the user vector > - aio_readv/writev methods to the file_operations structure I presume f_op->aio_readv could point to __generic_file_aio_read for most filesystems. Would f_op->aio_writev need a new wrapper function for 2.5.42 ? f_op->aio_write eventually calls generic_file_write which uses a different inode from generic_file_writev. So f_op->aio_writev might need to point to a function like generic_file_writev but using the same inode as generic_file_write. > - routine aio.c/io_readv_writev, which borrows heavily from do_readv_writev. > > I tested this using the aio dio patch that Badari submitted a while back. > I compared: > readv/writev io_submit for a vector of N iovecs > vs read/write io_submit for N iocbs. > > My performance data is only preliminary at this point, but aio readv/writev > appears to outperform aio read/write -- twice as fast in some cases. The > results generally make sense to me: while there is only one io_submit in both > cases, aio readv/writev shortens codepath (one instead of N calls to the > underlying filesystem routine) and should normally result in fewer Twice as fast looks good ! > bios/callbacks (at least for direct-io). As importantly, aio readv/writev > in my testing also reduces the number of (system) calls to io_getevents. It would be interesting to see the performance boost when events are retrieved at once, using the min_nr parameter of io_getevents. --Shailabh -- To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to majordomo@kvack.org. For more info on Linux AIO, see: http://www.kvack.org/aio/ -- To unsubscribe, send a message with 'unsubscribe linux-aio' in the body to majordomo@kvack.org. For more info on Linux AIO, see: http://www.kvack.org/aio/