This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH] Split mantissa calculation loop and add branchprediction to mp multiplication
On Mon, Dec 31, 2012 at 04:03:15PM +0530, Siddhesh Poyarekar wrote:
> On Mon, Dec 31, 2012 at 11:26:08AM +0100, Andreas Jaeger wrote:
> > >
> > >time ./powtest 1000000 1.0000000000000020 1.5000000000500000
> > >
> > >Taking into account the best times with and without the patch (with
> > >more than 10 runs each), the improvement is about a quarter of a
> > >second.
> >
> > What's the percentage? 0.25s can be a lot - or very little depending
> > on the total run time.
> >
>
> Ugh, sorry. The best time without the patch is 11.639 seconds for the
> above program and the best time with the patch is 11.409. So the
> improvement is 0.23 seconds, which is an improvement of about 1.97%.
>
Another note (sorry I'm doing this in instalments): with the
conversion of the mantissa to int[1], the improvement in performance
due to this patch jumps to about 0.4 seconds with the run time being
about 10.2 seconds without the patch, making it an improvement of
about 3.9%.
Siddhesh
[1] http://sourceware.org/ml/libc-alpha/2012-12/msg00443.html