Re: [PATCH] Additional compiler optimization options

From: Borislav Petkov
Date: Wed Dec 05 2012 - 18:08:27 EST


On Wed, Dec 05, 2012 at 03:00:12PM -0800, John wrote:
> > This only makes sense if you can show a noticeable performance
>
> > improvement using one of the those -march options with at least one
> > benchmark from those which we have now in the kernel.
>
> Thank you for the feedback, Boris. ÂCan you point me to the built in
> benchmarks to which you referred? ÂI am interested in probing this
> further.

Maybe the easiest thing to do is build the kernel:

make -j<number_of_cores+1>

in your kernel directory and time it. You could get more precise output
if you trace that workload with perf:

$ perf stat --repeat 10 make -j<number_of_cores+1>

And look at how the perf output looks like.

For all those runs, you need hardware supporting any of these options
and the running kernel should each time be built with a different
optimization option.

This will show us whether differently optimized kernels show any
performance difference.

And then you can go and try other benchmarks like kernbench, hackbench,
etc, whatever your favourite search engine gives you for "linux
benchmarks"

HTH.

--
Regards/Gruss,
Boris.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/