Hi Bill again (sorry)
With 0.6.5, you can easily benchmark by just building and installing.
you'll find benchmarks in ${PREFIX}/share/mpack/bench/
in my case,
Raxpy.dd Raxpy.dd_ref ...
Thanks
Subject: [sage-devel] Re: multiprecision linear algebra
Date: Sat, 22 May 2010 00:04:26 -0700 (PDT)
> Hi Bill
>
> DD 136Mflops (250MFlops)
> means that 250MFlops using OpenMP parallelism.
>
> Note that above results are using 0.6.4. It would be bit different
> using 0.6.5.
> Also seems to depend which compiler you use.
>
> thanks
>
>
>
> On 5月22日, 午後3:58, Maho NAKATA <maho.nak...@gmail.com> wrote:
>> Hi Bill
>> thanks for your comments
>>
>> On 5月21日, 午後10:56, Bill Hart <goodwillh...@googlemail.com> wrote:
>>
>> > I just took a look at their website:
>>
>> >http://mplapack.sourceforge.net/
>>
>> > * I'm personally confused about the licensing (may be a reflection on
>> > me, not mplapack). The MBLAS part is BSD, but the MPLAPACK part is
>> > LGPL v3+, but the authors say they are going to replace this later.
>>
>> sorry, what i meant is I'll replace all the parts by mBSD style
>> license.
>>
>> > But with what? They list "Including in a package system like FreeBSD
>> > ports" in their future goals, though this doesn't necessarily imply
>> > they are moving towards BSD licensing.
>>
>> I'm also a FreeBSD ports committer, so I'd like to add it for FreeBSD
>> ports tree
>> for easier compilation.
>>
>> > * Some nonsense about GMP (and incidentally no mention of MPIR - note
>> > how many other bignum libraries *are* mentioned).
>>
>> > * From a code quality point of view, their code looks pretty good to
>> > me, commented well, in english. I didn't do an extensive survey
>> > though.
>>
>> > * They implement a lot of functionality, however I wouldn't personally
>> > use this as a metric. Sergey Bochkanov's automatic code generation is
>> > exceptionally clever and should allow him to be streets ahead of
>> > mplapack very quickly
>>
>> > * I would benchmark carefully
>>
>> You should, recently I wrote a paper in Japanese, and my result
>> is something like
>>
>> Raxpy
>> DD: 136MFlops (250MFlops)
>> QD: 14.2MFlops (41.2MFlops)
>> GMP 14.2MFlops (38.8MFlops)
>> GotoBlas 1.5GFlops
>>
>> Rgemm
>> DD: 141M
>> QD 13.9M
>> GMP 14.4M
>> GotoBLAS 42.5G :-O
>>
>>
>>
>> > I became acquainted with ALGLIB and Sergey through his interaction
>> > with the MPIR project. Even if at some point mplapack is better than
>> > ALGLIB, I reckon Sergey would be exceptionally helpful in addressing
>> > that. He's a smart guy and very supportive of what Sage is doing!
>>
>> > Bill.
>>
>> > On 21 May, 00:48, Jason Grout <jason-s...@creativetrax.com> wrote:
>>
>> > > A while ago someone talked about incorporating ALGLIB into Sage to
>> > > provide, among other things, multiprecision linear algebra. I just saw
>> > > another package that does this:
>>
>> > >http://mplapack.sourceforge.net/
>>
>> > > Does anyone have experience with mplapack? How does it compare to
>> > > ALGLIB in linear algebra (for algorithms that are implemented in both)?
>>
>> > > Thanks,
>>
>> > > Jason
>>
>> > > --
>> > > To post to this group, send an email to sage-devel@googlegroups.com
>> > > To unsubscribe from this group, send an email to
>> > > sage-devel+unsubscr...@googlegroups.com
>> > > For more options, visit this group
>> > > athttp://groups.google.com/group/sage-devel
>> > > URL:http://www.sagemath.org
>>
>> > --
>> > To post to this group, send an email to sage-devel@googlegroups.com
>> > To unsubscribe from this group, send an email to
>> > sage-devel+unsubscr...@googlegroups.com
>> > For more options, visit this group
>> > athttp://groups.google.com/group/sage-devel
>> > URL:http://www.sagemath.org
>>
>> --
>> To post to this group, send an email to sage-devel@googlegroups.com
>> To unsubscribe from this group, send an email to
>> sage-devel+unsubscr...@googlegroups.com
>> For more options, visit this group
>> athttp://groups.google.com/group/sage-devel
>> URL:http://www.sagemath.org
>
> --
> To post to this group, send an email to sage-devel@googlegroups.com
> To unsubscribe from this group, send an email to
> sage-devel+unsubscr...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/sage-devel
> URL: http://www.sagemath.org
>
--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org
