Let me start by thanking William Stein for making this machine available. I would like to run a parallel polynomial multiplication benchmark on sage.math, however right now 16 cores are in use so performance degrades after 8 threads (the program does not handle heavy multitasking well). Here are the results I am getting:
> read "sdmp.mpl": > f := sdmp(expand((1+x+y+z+t)^30),plex(x,y,z,t),pack=4): > g := sdmp(expand((1+x+y+z+t)^30+1),plex(x,y,z,t),pack=4): > for i from 1 to 24 do sdmp:-num_cpus(i): p := sdmp:-Multiply(f,g): end do: time: 57.260 sec alg: 57.150 sec io: 0.110 sec real: 57.190 sec 0.999x time: 54.430 sec alg: 54.300 sec io: 0.130 sec real: 27.236 sec 1.994x time: 56.320 sec alg: 56.200 sec io: 0.120 sec real: 18.788 sec 2.991x time: 54.760 sec alg: 54.660 sec io: 0.100 sec real: 13.721 sec 3.984x time: 52.480 sec alg: 52.360 sec io: 0.120 sec real: 10.539 sec 4.968x time: 51.920 sec alg: 51.810 sec io: 0.110 sec real: 8.713 sec 5.946x time: 52.040 sec alg: 51.930 sec io: 0.110 sec real: 7.533 sec 6.894x time: 54.660 sec alg: 54.550 sec io: 0.110 sec real: 7.117 sec 7.665x time: 57.200 sec alg: 57.080 sec io: 0.120 sec real: 6.573 sec 8.684x time: 57.600 sec alg: 57.480 sec io: 0.120 sec real: 6.024 sec 9.542x time: 63.730 sec alg: 63.610 sec io: 0.120 sec real: 6.907 sec 9.209x time: 65.130 sec alg: 65.010 sec io: 0.120 sec real: 6.720 sec 9.674x time: 74.280 sec alg: 74.160 sec io: 0.120 sec real: 7.944 sec 9.336x time: 106.240 sec alg: 106.110 sec io: 0.130 sec real: 10.251 sec 10.351x time: 97.920 sec alg: 97.790 sec io: 0.130 sec real: 10.064 sec 9.717x Interrupted > Would it be possible to arrange a time when the machine is empty to run this test ? I would like to get a graph for a paper. In return, I am happy to share the binary with the sage community so you can do experiments with parallel computer algebra. You can find the binary in /home/mmonagan/sdmp start Maple and read in sdmp.mpl. Sorry no source code (it has bugs anyway) but the paper should be very helpful to anyone trying to implement something. Roman Pearce CECM/SFU --~--~---------~--~----~------------~-------~--~----~ 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 -~----------~----~----~----~------~----~------~--~---
