Simon King wrote: > On 22 Aug., 23:53, "Dr. David Kirkby" <david.kir...@onetel.net> wrote: > [...] >>> sage: R.<x,y,z> = PolynomialRing(GF(13)) >>> sage: time _ = expand((x+y+z+1)**100) >>> CPU times: user 0.07 s, sys: 0.00 s, total: 0.08 s >>> Wall time: 0.08 s >>> In[1]:= Timing[Expand[(x+y+z+1)^100, Modulus -> 13]][[1]] >>> Out[1]= 4.20826 > > So, over QQ, MMA is slightly faster, but over finite fields Sage > clearly wins? That is already something worth pointing out. > >> IMHO, it would be good to have one exactly as Karl has. By all means >> have others. > > I know almost nothing about MMA, but from the discussion here I get > the impression that you believe that MMA is playing on its strengths > (= prefering benchmarks that MMA make look good) and tries to tune it > so that well-known commonly used benchmarks work well.
No. I was not implying that at all. You must have mis-understood me. I am saying: 1) Mathematica does have a benchmark of its own. This times 15 different tasks. Data Fitting Digits of Pi Discrete Fourier Transform Eignenvalues of a matrix Elementary functions Gamma function Large integer multiplication Matrix Arithmetic Matrix Transpose Numerical Integration Polynomial expansion. Random number sort Singular Value Decomposition Solving a linear system. Times for each task are shown for about 15 computers. It shows you your own computer ranked against the other machines. No reference is made to the speed in computing any of these things compared to any other software - only compared to other hardware. Data is referenced to a 2.4 GHz Pentium 4 running XP, so that has a benchmark score of 1.0. 2) Karl has produced his own Mathematica benchmark. (I believe he had one long before Wolfram Research included one in Mathematica). His also does 15 different tasks, but they are different to the one Wolfram Research wrote. At this point in time, I do not think Wolfram Research are going to care too much what their own benchmark does. They only really offer it as a method of comparing Mathematica on different platforms. I suspect if Sage used the same benchmarks, then Wolfram Research would take some interest and ensure the benchmarked items were ones where they know other products can't do, or things where they know they are quicker than anyone else. > So, why shouldn't Sage do benchmarks in a similar spirit? Having only > the *same* benchmarks as MMA is weakness, IMHO. So, we should > definitely think of adding genuine "SAGE" benchmarks to the test > bench. Several good examples have been mentioned already. I agree. I suspect there needs to be a suite of benchmarks, but keep one the same as Karls', and possibly (legal issues??) the same as Wolfram Research's would be good IMHO. > Sorry for the self-promotion, but I doubt that any other software is > currently able to compute the complete ring structure + Poincaré > series + a-invariants of H^*(G;GF(2)) for all groups G of order 64 in > less than a day. Sage can do in less than an hour. > > Best regards, > Simon I'm all for such a comparison. --~--~---------~--~----~------------~-------~--~----~ To post to this group, send an email to sage-devel@googlegroups.com To unsubscribe from this group, send an 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 -~----------~----~----~----~------~----~------~--~---
