Re: Computer Algebra Algorithms lisp vs. C. BENCHMARKS?

parisse@xxxxxxxxxxxxxx wrote:

PPadilla wrote:

I don't understand why, when given the chance to prove your point you look for excuses. That usually tells folks that we are hearing (reading) a faith-based opinion rather than a data-based opinion. The proposed benchmarks are relatively simple and code probably available. When you want a more complex benchmark, it seems to me that you want to include things that have little relevance to your expressed concern. Adding things like GUI and other bit-twiddling code to a CAS benchmark is usually not done anyway.


I never spoke of GUI benchmarks. I spoke of more complex benchmarks
than just basic operations, some benchmarks who involve many CAS
algorithms like e.g. integration, factorization, polynomial gcd computation, etc. (<< That's why I suggested benchmarks
on relatively more complex tasks such as gcd or factorization
(to see if the library has a good implementation of "basic" services)>>)
If you think it has little relevance, it's your opinion... 

I quote from
http://www-fourier.ujf-grenoble.fr/~parisse/giac_us.html


"If you want to use NTL for some polynomial operations (currently factorization), get version >= 5.2 at http://www.shoup.net. "

Since NTL only factors univariate polynomials, am I correct in assuming
that any factorization problem that involved two or more variables
will not be possible in GIAC?

What limitations are there is the rest of your list of advanced functions,
e.g. "integration"  does this include just univariate polynomials?
univariate rational functions? Exponential, log, algebraic, combinations?

If we were to compare GIAC to (say) Axiom or Maxima, to mention two
other free systems, on selected problems, it seems likely we could find
a collection in which GIAC could not solve a single one, but Axiom and
Maxima could each solve them all.

As I mentioned a few days ago, NTL can be read into lisp (into Maxima,
for example). So comparing GIAC to MAxima on univariate polynomial
factorization could be simply comparing NTL to NTL.   Or GMP to GMP.
I don't know if this is good or bad :)
Of course in some tests NTL is slow. (sparse) and in some cases it
just doesn't work e.g. 2 variables.

RJF




There is ample evidence that I base my points on facts, since giac/xcas
is available to everybody who want to test and benchmark it. You can
also find in the examples/lewisw some benchmarks which are classical.

.

Relevant Pages

  • Re: Writing a Compiler: Lisp or Scheme or Haskell?
    ... incompetence and malignity of a certain spammer usually favorizing his ... your ignorance of the benchmarks game - ignorance achieved simply by not ... That should make you care about his opinion on things Lisp in which he ...
    (comp.lang.lisp)
  • Re: iTunes 7
    ... Do you have any benchmarks to back that up, ... Note that's with the old Safari - 1.3x, ... including benchmarks showing Gecko browsers ... I prefer my own opinion over those of "experts". ...
    (comp.sys.mac.advocacy)
  • Re: NexusDB database engine Crossgrade introduction
    ... > I did not see any benchmarks on the NexusDB site. ... Our official opinion about database "benchmarks" was made several times ...
    (borland.public.delphi.thirdpartytools.general)
  • Re: FastCode MM B&V 0.56
    ... > quality before 12/12. ... My opinion is that the validation needs a little ... Remember that the benchmarks are in effect also part of the validation. ...
    (borland.public.delphi.language.basm)