|From:||Linus Torvalds <torvalds-AT-osdl.org>|
|To:||Ingo Molnar <mingo-AT-elte.hu>|
|Subject:||Re: [PATCH 3/3] IRQ: Maintain regs pointer globally rather than passing to IRQ handlers|
|Date:||Mon, 2 Oct 2006 13:54:33 -0700 (PDT)|
|Cc:||Andrew Morton <akpm-AT-osdl.org>, David Howells <dhowells-AT-redhat.com>, Thomas Gleixner <tglx-AT-linutronix.de>, linux-kernel-AT-vger.kernel.org, linux-arch-AT-vger.kernel.org, Dmitry Torokhov <dtor-AT-mail.ru>, Greg KH <greg-AT-kroah.com>, David Brownell <david-b-AT-pacbell.net>, Alan Stern <stern-AT-rowland.harvard.edu>|
On Mon, 2 Oct 2006, Ingo Molnar wrote: > > i agree that we should do this in one go and in Linus' tree. I suspect > David has a script for this, so we can do it anytime for any tree, > right? > > the amount of code that truly relies on regs being present is very low. > Mostly only sysrq type of stuff and the timer interrupt is such. Yeah, well, it's been discussed before, and the real problem is not the patch itself, it's the damn drivers maintained outside the tree, and people who want to maintain the same driver for multiple different versions of the kernel. Things like the kernel graphics direct-rendering code, for example - mostly maintained in X.org trees that then want to compile with other kernels too. I don't personally mind the patch, I just wanted to bring that issue up. So far, when this has come up, the gains it gives have not been worth the pain. I don't quite see why FRV is so broken that it would matter 20% worth, and I suspect that number was somehow really not real, but more a matter of "this small code snippet that is part of the irq delivery and isn't really measurable improves by 20%", which is a different thing. That said, it's almost certainly worth it, and I don't think anybody really objects deep down. So if the patch works against my current tree, and nobody objects, I can certainly apply it. So speak up, people... Linus
Copyright © 2006, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds