On Fri, 25 Jan 2002, Andrew Morton wrote:
> >
> > Looking at the code, I suspect that 99.9% of this "improvement" comes from
> > one thing, and one thing only: you removed the "cli" in the system call
> > return path.
>
> Before the cli was in the stock kernel, I had added it in the
> low-latency patch. Careful testing showed that it added
> 13 machine cycles to a system call on a P3.
That sounds about right. The empty system call path is basically dominated
by the trap/iret costs, and is on the order of 200 cycles or so on most
CPU's. So 13 cycles would account for the roughly 5% improvement.
Linus
-
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/
This archive was generated by hypermail 2b29 : Thu Jan 31 2002 - 21:00:38 EST