On Jan 3, 2009, at 2:55 PM, mabshoff wrote:
> > > > On Jan 3, 11:48 am, Tim Lahey <tim.la...@gmail.com> wrote: >> On Jan 3, 2009, at 2:43 PM, mabshoff wrote: > > Hi Tim, > >>> Given a choice between Pan-Axiom, Maxima and Reduce for the role of >>> symbolic calculus in Sage these days I personally would pick FriCAS >>> over the other choices. But since all of the above do not solve the >>> fundamental problem of using pexpect I think that in the long term >>> we >>> will come up with something on our own, i.e. partially sympy and >>> partially our own code. That doesn't mean the interface to any >>> existing system will be removed, just that Maxima would >>> potentially at >>> some point be made optional - should it ever come to the point where >>> we have something better. >> >> FriCAS has the problem that in some cases it is very slow. > > Ok, can you report them to the FriCAS people so they can potentially > fix this? Part of the problem might also be clisp here, so let's see > how for example ecl fares. > However, Maxima is running under the same lisp, so it isn't just that. Anybody can run the suite and get the timings so it's fairly easy to see which integrals are slow and which aren't. I'll post complete timings once I've finished all the integrals. >> It turns out >> that there are many of these slow cases based upon my tests. If >> Reduce >> can support multiple solutions directly and is faster than FriCAS >> then >> it is a viable interim solution. > > Well, interim is a bad, bad word :). Seriously, switching from one to > the other and sorting out all the build issues and even writing the > code takes a considerable amount of man power. If that is going to be > replaced in a year's time (just to throw around a figure) it is much > better in the mean time to "hang in there" with Maxima and concentrate > on our own code and shifting functionality to SymPy in the cases where > it is appropriate. > I understand that. I was speaking more from a perspective of FriCAS vs. Maxima vs. Reduce. If we were to switch away from Maxima to another Lisp package, I think it should be based upon correctness and performance. Right now, based upon the tests I've done, aside from the questions, Maxima's correctness appears to match FriCAS and on average has better performance. > And re the speed concern: Correctness should come first. As is non of > the solutions seem to be very fast and/or as powerful as MMA in the > average case (I am sure you can find problem classes when Axiom and or > Maxima beat the pants of MMA, but that is likely on average not the > case). Maxima and FriCAS aren't too bad at integration, but I don't know how Mathematica compares. I have Maple 11 and 12, so once the Maple interface is fixed, I'll definitely test Maple against the suite. > > > One thing that reduce did very well last time we checked was mv > factorization. Maybe the mv gcd is also very fast, so I think there is > potential there. But unless someone actually goes out, builds a spkg, > tests it on many platforms this is all an academic exercise :) > True. I'm hoping that Reduce is an option, because I definitely want to try some of the packages. If not, I might look at porting some of them once I finish my thesis. Cheers, Tim. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send email to sage-devel@googlegroups.com To unsubscribe from this group, send email to sage-devel-unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/sage-devel URLs: http://www.sagemath.org -~----------~----~----~----~------~----~------~--~---
