Re: [PATCH -mm] ide_end_drive_cmd(): avoid instruction pipeline stall

[Al Boldi <a1426z@gawab.com>]

Alan Cox wrote:
> Ar Gwe, 2006-06-30 am 18:13 +0200, ysgrifennodd Andreas Mohr:
> > Use an independently-formatted "unsigned int" for data instead of a
> > restrictive "u16" to avoid instruction fetch pipeline stalls
> > probably caused by the byte calculations later.

I'm hitting this all the time on 2.6 more so than on 2.4, especially for CPU 
bound procs.

> 1. This is a gcc problem

Maybe, but on 2.6, echo 0 > /proc/sys/kernel/randomize_va_space seems to 
improve the situation most of the time, although not always.

> 2. Not everyone is using an intel x86-32 box which has such problems

Specifically, 3/4/586 don't show this problem, whereas 686+ show the problem 
to varying degrees with slowdown upto 333% on P4.

Can you confirm this?

Thanks!

--
Al

-
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/