> I'm curious to know if performance of Maple 15 are following the same as for Maple 14.
Very similar. SDMP was refactored for Maple 15. It's size was cut in half, it was made re-entrant to allow parallel polynomial algorithms, and we now support Zp for multiprecision p. We added Kronecker substitution for univariate multiplication (calling GMP, sadly version 4.2), and implemented powering algorithms in C. There's some internal stuff for recycling memory which makes the parallel multiplication use the same amount of memory as sequential. We needed that for parallel division, which sadly did not ship because the Windows scheduler threw monkey wrenches at us. We fixed the issue but it was too late for Maple 15. There are lots of other goodies in the pipeline now. We have been coding for about 8 months, and Maple 15 is just the first 2 months of that. As for Trip, those results are very impressive. We have improved the performance and scalability of sdmp, but we can't match those results. We're getting a big machine soon so there is a chance to try. The heap algorithms were mainly chosen for their memory use. For multiplication f*g the heap is O(min(#f,#g) + result) and for division f/g = q it is O(#g + #q), this is both sequential and in parallel. We emphasized superlinear speedup on a modest number of cores (16 or less) to offer stiff competition on the desktop and because higher level algorithms should be parallelized too. We have quite a lot on our plate so we concede the speed crown to Trip for now. I can't say I mind, because the scalability they achieved at 64 cores is an extremely difficult thing to do. We're still looking good next to Magma and Singular (which Sage uses). But we're not stingy. If anyone is implementing these algorithms or other sparse algorithms, we're happy to discuss details in newsgroups or by email. -- 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 URL: http://www.sagemath.org
