[sage-devel] elliptic_e(0.5, 0.1) differs from Mathematica 7 by about 0.04%.

On Mon, Aug 10, 2009 at 9:08 AM, Raymond Toy <raymond.toy at stericsson.com>wrote:

> William Stein wrote:
> >
> >
> > On Mon, Aug 10, 2009 at 1:02 AM, Dr. David Kirkby
> > <david.kirkby at onetel.net <mailto:david.kirkby at onetel.net>> wrote:
> >
> >
> >     Valery Pipin wrote:
> >     > William Stein wrote:
> >     > <<sorry, it's too long> >
> >     >
> >     >>> I've no idea how the Sage group would feel about switches
> >     lisps. Given
> >     >>> they have just recently done that (I forgot what was used
> >     before), there
> >     >>> might not be too much enthusiasm for it.
> >     >> Since you have no idea, perhaps I should clarify:  There is no
> >     way in hell
> >     >> we are switching from ECL to anything else.
> >     >>
> >     >> ECL is massively better than CLISP, and is also the *only*
> >     other lisp that
> >     >> is currently supported and builds 100% from source code.   Both
> >     CMUCL and
> >     >> SBCL are immediately ruled out just because of that reason.
> >      This is one of
> >     >> the DoD requirements for Sage -- they absolutely will never
> >     consider using
> >     >> CMUCL or SBCL (I have asked).
> >     > Too bad :-). Sbcl builds from the source on the most Unixes at
> least
> >
> >
> > Just to be clear -- it does *not* build from source using the GCC
> > compiler anywhere. It requires an existing common lisp implementation
> > to get anywhere.  The first sentence of guide to installing SBCL from
> > source says: "To build SBCL you need a working toolchain and a Common
> > Lisp system  (see section 2.5 "Supported platforms")."   As a result,
> > SBCL can't be used for Sage, as mentioned above.   I should also
> > mention that according to
> > http://sbcl.sourceforge.net/platform-table.html, SBCL doesn't support
> > Microsoft Windows.
> >
> How do you build gcc from source?  Why is SBCL (or cmucl) disallowed on
> that ground but gcc is allowed?  What makes gcc special that it's not
> required to build gcc 100% from source?
>
> (Rhetorical questions.  I think I know the answer.)


It doesn't matter what the answer is, as it is dictated to me by the DoD
bureaucracy.     It's also relevant that Sage's main competitor in this
particular situation is Magma, which builds entirely using GCC.



> >
> >     >> Why do you think cmucl or sbcl would give high performance. Are
> you
> >     >>
> >     >>> suggesting ecl would give lower performance?
> >     > It is certainly true. Should I write a paper for sage journal
> >     about it?
>

Unfortunately, we don't really have a Sage journal yet...

>
> >
> I think this is debatable.  Someone here says cmucl runs his maxima code
> much faster than sbcl.  The testsuite runs fastest with gcl.
>
> The only way to know for sure is to try different ones on *your* code to
> see which is fastest.
>
> Ray


Thanks for your comments,
 William



-- 
William Stein
Associate Professor of Mathematics
University of Washington
http://wstein.org