On Nov 5, 6:18 pm, "Bill Hart" <[EMAIL PROTECTED]> wrote:
> I just compiled and linked again from a fresh session on sage.math and
> the result is the same.
>
> I can only guess that something is wrong on the amd64. Either that, or
> the most recent version in the copy of the svn repo that I have is not
> the most up-to-date. However I doubt the latter due to the fact that
> those warning messages come up during the compile. They should be
> harmless, as the linking seems to go ahead just fine.
>
> Bill.
>
> 2008/11/5 Cactus <[EMAIL PROTECTED]>:
>
>
>
> > On Nov 5, 5:03 pm, "Bill Hart" <[EMAIL PROTECTED]> wrote:
> >> Here are the times:
>
> >> Bit Size Magma eMPIRe+ eMPIRe-
> >> 2000000 1.500s 26.915s 35.497s
> >> 1000000 0.610s 5.794s 7.527s
> >> 500000 0.240s 1.440s 1.864s
> >> 250000 0.090s 0.312s 0.360s
> >> 100000 0.020s 0.040s 0.052s
> >> 50000 0.010s 0.012s 0.012s
>
> >> I begin to not believe the Magma times.
>
> >> SVN is not back up. Well it is locally for me, but I know nothing at
> >> all about apache, so I do not know how to start up the apache modules
> >> so everyone else can access it.
>
> >> Bill.
>
> >> 2008/11/5 Jason Martin <[EMAIL PROTECTED]>:
>
> >> > Also, is SVN back up yet?
>
> >> > Jason Worth Martin
> >> > Asst. Professor of Mathematics
> >> >http://www.math.jmu.edu/~martin
>
> >> > On Wed, Nov 5, 2008 at 11:56 AM, Bill Hart <[EMAIL PROTECTED]> wrote:
>
> >> >> Ok I'll give it a go. Perhaps it'll turn up something stupid that I did.
>
> >> >> Bill.
>
> >> >> 2008/11/5 Jason Martin <[EMAIL PROTECTED]>:
>
> >> >>> How about eMPIREe with Moller Patches versus eMPIREe without?
>
> >> >>> Jason Worth Martin
> >> >>> Asst. Professor of Mathematics
> >> >>>http://www.math.jmu.edu/~martin
>
> >> >>> On Wed, Nov 5, 2008 at 11:40 AM, Bill Hart <[EMAIL PROTECTED]> wrote:
>
> >> >>>> Yep, I timed Magma and eMPIRe with Moller's patches doing integer
> >> >>>> gcd's only on sage.math. Here is the table:
>
> >> >>>> Bit Size Magma eMPIRe
> >> >>>> 2000000 1.500s 26.915s
> >> >>>> 1000000 0.610s 5.794s
> >> >>>> 500000 0.240s 1.440s
> >> >>>> 250000 0.090s 0.312s
> >> >>>> 100000 0.020s 0.040s
> >> >>>> 50000 0.010s 0.012s
>
> >> >>>> Clearly the half-gcd is not working.
>
> >> >>>> Bill.
>
> >> >>>> 2008/11/5 Bill Hart <[EMAIL PROTECTED]>:
> >> >>>>> I plotted the poly gcd graph out to a higher bit size and now Magma
> >> >>>>> is
> >> >>>>> clearly winning.
>
> >> >>>>> The left most column is just integer GCD. Could it be that Magma
> >> >>>>> improves on Moller's gcd patches (I think I recall reading somewhere
> >> >>>>> that he originally wrote them for Magma), or did we do something
> >> >>>>> wrong
> >> >>>>> when merging them?
>
> >> >>>>>http://sage.math.washington.edu/home/wbhart/flint-trunk/graphing/gcd8...
>
> >> >>>>> Bill.
>
> >> >>>>> 2008/11/1 Bill Hart <[EMAIL PROTECTED]>:
> >> >>>>>> Sorry, I simply mean you commit to the repo. I agree Mollers
> >> >>>>>> algorithm
> >> >>>>>> is clever and it should be pointed out that his paper on the topic
> >> >>>>>> is
> >> >>>>>> new work. He didn't just implement the half-BCD algorithm.
>
> >> >>>>>> Bill.
>
> >> >>>>>> On 01/11/2008, Jason Martin <[EMAIL PROTECTED]> wrote:
>
> >> >>>>>>> Okay, I'll take a look at that when SVN is up again.
>
> >> >>>>>>> Also, don't call them "Jason's Patches". That makes it sound like
> >> >>>>>>> I
> >> >>>>>>> was clever... I wasn't. It's Moller's code, and his algorithm is
> >> >>>>>>> very
> >> >>>>>>> very clever.
>
> >> >>>>>>> --jason
>
> >> >>>>>>> Jason Worth Martin
> >> >>>>>>> Asst. Professor of Mathematics
> >> >>>>>>>http://www.math.jmu.edu/~martin
>
> >> >>>>>>> On Sat, Nov 1, 2008 at 9:55 AM, Bill Hart <[EMAIL PROTECTED]>
> >> >>>>>>> wrote:
>
> >> >>>>>>>> Jason, first bug report. When I build mpir with your patches (I
> >> >>>>>>>> have
> >> >>>>>>>> got the mpir repo working locally for me only - hopefully Michael
> >> >>>>>>>> will
> >> >>>>>>>> help me with the apache export so everyone else can get access
> >> >>>>>>>> again),
> >> >>>>>>>> I get warning messages in the compile on sage.math.
>
> >> >>>>>>>> gcd.c: In function 'mpz_rgcd':
> >> >>>>>>>> gcd.c:167: warning: implicit declaration of function 'mpn_rgcd'
> >> >>>>>>>> gcd.c: In function 'mpz_bgcd':
> >> >>>>>>>> gcd.c:171: warning: implicit declaration of function 'mpn_bgcd'
> >> >>>>>>>> gcd.c: In function 'mpz_sgcd':
> >> >>>>>>>> gcd.c:175: warning: implicit declaration of function 'mpn_sgcd'
> >> >>>>>>>> gcd.c: In function 'mpz_ngcd':
> >> >>>>>>>> gcd.c:179: warning: implicit declaration of function 'mpn_ngcd'
>
> >> >>>>>>>> It seems to build ok otherwise.
>
> >> >>>>>>>> Bill.
>
> >> >>>>>>>> 2008/11/1 Jason Martin <[EMAIL PROTECTED]>:
>
> >> >>>>>>>>> On Fri, Oct 31, 2008 at 9:11 PM, mabshoff
> >> >>>>>>>>> <[EMAIL PROTECTED]> wrote:
> >> >>>>>>>>>> [snip]
> >> >>>>>>>>>> For me the highest priority item is the Moller's gcd code which
> >> >>>>>>>>>> has
> >> >>>>>>>>>> already been merged. What is the status of the code, i.e.
> >> >>>>>>>>>> performance,
> >> >>>>>>>>>> any known bugs, etc?
>
> >> >>>>>>>>>> Cheers,
>
> >> >>>>>>>>>> Michael
>
> >> >>>>>>>>> Hi Michael,
>
> >> >>>>>>>>> I haven't tested Moller's code beyond the standard make check
> >> >>>>>>>>> test
> >> >>>>>>>>> suite (which passes all tests). I also haven't done any speed
> >> >>>>>>>>> tests
> >> >>>>>>>>> other than the most rudimentary checks. So if anyone has some
> >> >>>>>>>>> suggestions for test cases, I'd be happy to try them out.
> >> >>>>>>>>> Likewise,
> >> >>>>>>>>> if anyone has any bug reports for the gcd code, let me know and
> >> >>>>>>>>> I'll
> >> >>>>>>>>> attempt to figure it out.
>
> >> >>>>>>>>> --jason
>
> > I have just tried the equivalent of Bill's code in Mathematica 6 with
> > a, b and c containing 2,000,000 bits each and the GCD is computed in
> > 2.1 seconds on my dual core 2.2GHz Intel based laptop. Since
> > Mathematica uses my Windows GMP port without the new GCD code, Bill's
> > timings look to be a factor of 15 higher than I would expect from MPIR
> > (assuming a similar machine). I cannot quickly put the MPIR DLL under
> > Mathematica as I am not on my development machine right now but it
> > does seem that something has gone wrong - unless, of course, I am
> > wrongly interpreting Bill's timings.
>
> > Brian
I have just confirmed Bill's timings and it is clear that we are not
gaining full advantage from the new GCD code in eMPIRe.
I have checked out the changes that Jason made to add this code and I
have found one thing that he didn't do - our gmp_h.in does not include
the definitions of the interfaces to the new GCD routines that are in
Niels Möller's modified gmp_h.in.
Also, looking at Niels Möller's speed test code, I am inclined to
think that the new code either has to be called directly to get the
speed gain OR the existing GMP routines have to be modified to call
these new routines. I don't think that simply adding the new routines
alone is sufficient alone unless they are called directly (however I
am not sure about this as I have only just started to look at it)..
best regards,
Brian
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"mpir-devel" group.
To post to this group, send email to mpir-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/mpir-devel?hl=en
-~----------~----~----~----~------~----~------~--~---
