site_archiver@lists.apple.com Delivered-To: darwin-kernel@lists.apple.com On Mar 3, 2008, at 4:15 AM, Michael Smith wrote: Thanks, - Greg _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel@lists.apple.com) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.a... Can you explain why you think it would be more efficient, or that it would simplify your code? I ask because it's possible that it might, but without seeing your reasoning I can't comment on whether it might actually be, and it's common for folks to think that mapping would be cheaper when in fact it's not. I can only say that I need to be able to move relatively small chunks of data (around 90 bytes per chunk) in potentially large quantities as fast as is possible across the kernel/user boundary. You have to become an I/O Kit extension to use the IODataQueue. There are other ways to share mapped memory between the kernel and user space however. ... What would those ways be specifically? All I'd like is some example code on how to do this in C. Then at least I'd be able to experiment with both approaches and see which one yields the best results. Unless your data is large, you'll tend to find that copying it along with the notification (as you are currently doing) will be easier and more efficient than trying to manage a shared memory buffer. Again, it would help to be able to put this in perspective with your current workload. So yeah, like I said above, the data chunks are not large (I don't think), but I need to be able to move large quantities of them as quickly as is possible to the user-land program. This email sent to site_archiver@lists.apple.com