On Jan 3, 9:03 am, Bill Hart wrote:
> I figured I probably did misunderstand. What I think they are trying
> to get at is that one can use coefficients modulo 2^N+1 except at
> certain levels because the roots of unity are actually the same. But
> then the discussion becomes one of combining a
On Jan 2, 4:13 pm, Bill Hart wrote:
> 2) The MPIR FFT uses both a Fermat FFT and a Mersenne FFT, i.e. it
> does a multiplication of polynomials with coefficients mod p with p =
> 2^{aN}+1 then one with coefficients mod p = 2^N-1. Then it CRT's the
> results. By varying the parameter _a_, as well
Hi,
Bill encouraged me to mention some new approximate reciprocal code
I've written, based on the middle product, patch is available under
BSD license at
http://www.cims.nyu.edu/~harvey/mulmid/
I haven't had time to profile it properly yet, but it *should* be
pretty fast. I'll post timings
On Oct 19, 11:19 am, William Stein wrote:
> > William, it's difficult for me to take this comment seriously.
>
> Given that it is meant as a joke (which I hoped would be clear), I
> hope you will not take it seriously.
So, do you have a serious opinion on this issue? I happen to think
it's quit
On Oct 19, 10:42 am, William Stein wrote:
> On Mon, Oct 19, 2009 at 7:31 AM, David Harvey wrote:
>
> > On Oct 19, 4:47 am, Bill Hart wrote:
> >> I've just put up MPIR 1.3 release candidate 1. We are still waiting on some
> >> final testing from Jeff Gi
On Oct 19, 4:47 am, Bill Hart wrote:
> I've just put up MPIR 1.3 release candidate 1. We are still waiting on some
> final testing from Jeff Gilchrist, which will probably happen tomorrow. If
> everything is fine, the release candidate will become the final release.
Wait a sec you are actua
On Jun 12, 1:51 pm, Bill Hart wrote:
> Is there something in particular you are looking for? Maybe one of us can
> help?
No, nothing in particular just curious to see what's going on.
Something like
http://hg.sagemath.org/sage-main/
or
http://gmplib.org:8000/gmp/
It's nice to have a br
Hi,
Is there any web interface to the MPIR repository that shows recent
changes? All I can find is
http://modular.math.jmu.edu/svn/mpir/
which lets me browse the current files.
david
--~--~-~--~~~---~--~~
You received this message because you are subscribed to
ing from main memory (not even L2)
> L1 cache line size is 64 bytes which is 8 limbs so if this was affecting it we
> would have a n mod 8 pattern to the times not a n mod 4
>
> On Monday 04 May 2009 18:02:50 David Harvey wrote:
>
> > What happens if you remove the epilogue, i
op really running slower and not some
> delay after the loop. But I've put large alignments at the start *and* end of
> the main loop so we know it's not a mismatch between decode/execute loops. I
> can't think what else it could be.
>
> I've noticed this on some othe
Do you get consistent numbers if you run only for a single value of n?
i.e. it's not an artifact of the way the buffers are allocated or
something?
david
On May 4, 10:27 am, Jason Moxham wrote:
> Hi
>
> I've been playing with some assembler for the Intel Core2 chips and have come
> across this
On Apr 15, 7:11 pm, Bill Hart wrote:
> Clicking on the graph does make it bigger, but yeah, the critical
> region we are talking about is a bit compressed.
I just took a look at the toom4 code. It looks like there is a lot of
unnecessary sign/size management and data copying. I would suggest
On Apr 15, 6:08 pm, Bill Hart wrote:
> It's a bit hard to see the graph, but you see a similar phenomenon
> with the cutoffs on Bodrato's graph:
>
> http://bodrato.it/software/tc3-7-percent.png
>
> Toom 7 looks to be ahead of Toom 4 from a very early point.
Sorry mate, I can't quite see it, my
On Apr 15, 4:42 pm, William Stein wrote:
> You might find my notes from Bill's class right now interesting:
>
> http://wiki.wstein.org/09/583e
>
> Click on the "schedule" link.
Hmmm, the notes are a bit hard to decipher, but I see things like
"For Toom 7, cutoff is abut 190 limbs, and it alwa
On Apr 15, 2:21 pm, Bill Hart wrote:
> * Toom 4 and Toom 7 (balanced only)
Sorry, what do you mean by toom 7? Do you mean splitting the inputs
into 7 parts each? Wouldn't it make sense to tackle toom 5 and toom 6
first? Or am I misunderstanding something?
david
--~--~-~--~~--
g as essentially being
> > present in there. Sorry if I have conflated two different issues here.
>
> > If you say so, then I'm happy to credit Harvey, Hart, Gladman, Moxham.
>
> > Bill.
>
> > 2009/2/15 David Harvey :
>
> >> On Feb 15, 2:34 pm, Bill
On Feb 15, 2:34 pm, Bill Hart wrote:
> At some point we discussed where the original, original idea came
> from. At some point you read Granlund-Montgomery (after we had our
> discussion, I think) and at some point you may have decided that the
> idea was essentially already present in that work
what this branch was all about and thought it was
> just the divexactby3 stuff) and the mention of the improvements on the
> GMP list.
>
> Bill.
>
> 2009/2/15 David Harvey :
>
>
>
> > Is the algorithm based on the idea mentioned in Torbjorn's post from
&
Is the algorithm based on the idea mentioned in Torbjorn's post from
last year?
http://gmplib.org/list-archives/gmp-devel/2008-August/000816.html
david
On Feb 15, 10:25 am, ja...@njkfrudils.plus.com wrote:
> On Sunday 15 February 2009 14:13:23 David Harvey wrote:
>
> > Hi, I
Hi, I'm curious to try this new divide-by-3 code, but I can't find it
in the repo. Where do I look? How many c/l is it?
david
p.s. The wiki still says
svn co svn://sage.math.washington.edu/mpir/
but this is wrong, apparently
svn co http://modular.math.jmu.edu/svn/mpir/
is current?
On Feb 15
On Jan 25, 9:50 am, ja...@njkfrudils.plus.com wrote:
> Another way is to break the mul_basecase rectangle into columns rather than
> rows.Here we get a much better basic block of
>
> mov (src1),AX
> mul (src2)
> add AX,t1
> adc DX,t2
> adc 0,t3
>
> which is 6 macro-ops , hopefully leading to 2c
On Jan 19, 2:32 pm, ja...@njkfrudils.plus.com wrote:
> Presently for a 20x20 mul_basecase we have cycle counts of
> 2082 gmp-4.2.4
> 1461 mpir trunk
> 1153 mpir k8-branch
> 1103 pipeline
> 1033 perfect
I'm curious to try this code, but I can't seem to check out the repo
right now. Where is it lo
Hi,
Suppose that, in a month or two, MPIR and GMP 4.3 have been released,
and that the official Sage distribution now uses MPIR instead of GMP.
Suppose that a friend of mine wants to run Sage using GMP instead of
MPIR. Under these hypotheses, would it be technically feasible for my
friend
On Dec 22, 2008, at 9:35 PM, ja...@njkfrudils.plus.com wrote:
> On Tuesday 23 December 2008 02:03:25 Bill Hart wrote:
>> I don't see a fix on the gmp-discuss list.
>>
>
> Bet there be one tommorow from Torbjörn
I have reported it to gmp-bugs, but it has not passed through
moderation yet.
dav
On Dec 7, 2008, at 1:35 PM, [EMAIL PROTECTED] wrote:
>
> I been trying to account for each and every cycle in the asm
> routines , and I
> came across this oddity
>
> let mpn_fn be unrolled to X-way
> then run
> ./speed -cD -s X-X*20 -t X mpn_fn
>
> and you see the time differences per loop as
On Dec 4, 2008, at 10:59 AM, Bill Hart wrote:
> A quick update on ptlsim.
>
> Someone on the ptlsim list pointed out to me that when priming the
> cache, one needs to do this *during* the simulator run, not
> immediately before, as ptlsim also simulates the caches and has no way
> of accessing a
On Nov 26, 2008, at 1:12 PM, Bill Hart wrote:
> In another one of my projects (FLINT) there exists FFT integer
> multiplication code written by David Harvey and myself which will give
> up to a factor of 2 improvement on FFT sizes. But it needs to be
> rewritten to merge with eMPIRe
27 matches
Mail list logo