On Sep 24, 2:10 pm, Robert Bradshaw <[EMAIL PROTECTED]>
wrote:
> On Sep 24, 2008, at 2:02 PM, mabshoff wrote:
>
>
>
> > On Sep 24, 9:48 am, Robert Bradshaw <[EMAIL PROTECTED]>
> > wrote:
> >> On Sep 24, 2008, at 3:42 AM, mabshoff wrote:
>
> > Hi Robert,
>
> >>> What is new in alpha1:
>
> >>> * Robert Bradshaw: more new coercion merges (causes speed
> >>> regression
> >>> in sr.py by about 40%, but since that sucks anyway we can bear
> >>> with it
> >>> a little longer until Robert fixes it)
>
> >> To my defense, that slowdown was all in a single #long doctest that
> >> created a huge number of examples of varying sizes--the rest of the
> >> file experienced a speedup. The new coercion model pays a slightly
> >> higher price for discovery for a lower price doing arithmetic
> >> thereafter.
>
> > I am not blaming you here (well, a little I guess :), I just wanted to
> > point out that there is a problem),
>
> Nothing was taken personally. I just wanted to dispel the impression
> that the new coercion was going to make everything slower...
Sure, I was not the most rested person when I wrote the above blurb
and the regression issue with sr.py and I did not mean to imply that
this was a generic issue. I guess since it is obvious to me that this
is one rather unusual use case I did not consider how other people who
shamelessly are not aware of every problem in Sage could misread
that :)
> > but I thought that the new
> > coercion model allows us to fix the above problem when coercion mv
> > polynomial rings with loads (thousands!) of indeterminates. IIRC the
> > problem is that when doing the coercion we end up creating a couple
> > thousand mv polynomial rings until we finally get to the one we want
> > and need. The fix discussed at SD6 was to first check some special
> > cases, but I am sure malb has a much better recollection of the issue
> > and the suggested fix here.
>
> Yes, there is the potential for massive (completely orthogonal)
> speedups there, it's just a matter of porting the polynomial changes
> over from the coercion branch back into Sage itself.
Nice. I have been pushing hard to get all your coercion patches into
Sage so that your and David Roe's work (and all the other contributors
to the coercion rewrite obviously) finally pays off. #4111 will be
merged in the next hour, so 3.1.3 should finally use the new
infrastructure in a bunch of places. I am currently valgrinding
3.1.3.a1 and so far I have not seen any funny business related to new
coercion, so I am hopeful that the transition will be smooth.
If you post the polynomial changes from new coercion I am sure someone
will be more than motivated to review this.
> - Robert
Cheers,
Michael
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-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/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
