Messages in this thread | | | From | David Howells <> | Subject | Re: [PATCH 3/3] IRQ: Maintain regs pointer globally rather than passing to IRQ handlers | Date | Tue, 03 Oct 2006 11:21:21 +0100 |
| |
Linus Torvalds <torvalds@osdl.org> wrote:
> 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.
What appears to make up the difference is the loop in handle_IRQ_event(). That has to resurrect the arguments for the IRQ handler after calling the previous IRQ handler.
FRV is just the easiest place for me to measure things like this. Trying to do so on i386 would be tricky, and Xen wouldn't help as it could affect the measurement of time - though it might permit me to count the intructions instead. I might be able to do so on my power5 box, I suppose, but again, like Xen, that's virtualised, and I'm not sure what affect that'd have.
So, I'm sure this will affect other archs, but it's much harder for me to measure those.
But, you're also right: this is a statistic, and I'm sure you know the old saying about those...
David - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |