On Oct 8, 11:10 pm, Bill Hart <goodwillh...@googlemail.com> wrote: > Hi all, > > We have just fixed the last known major bugs for the up and coming > MPIR 1.3.0. If anyone would like to try it out, it is available from > svn. Once we have more broadly tested it on our build farms in a day > or two, we'll upload a tarball and go through the usual release cycle > of alphas, etc. > > svn cohttp://modular.math.jmu.edu/svn/mpir/mpir/trunkmpir-1.3 > > Brian has done testing on Windows and so far it is passing. Jeff, it > is up to you if you would like to either test from svn or wait until > we have a tarball. > > The new benchmarks look good. They show we are still well behind GMP > on root extraction, especially cube roots, but that we've caught up a > lot since our last release and that we are a long way ahead on new > primality tests we added to the benchmark. > > The main things we really need now are: > > * Optimise the function mpn_nhgcd2 function in the > mpn/generic/nhgcd2.c file. This will speed up both gcd and xgcd. > * Fast algorithm for extracting cube roots.
Our current root code , although asymtotically much faster than mpir-1.2 is not at all optimised. 1) we can use partial products , all the code is there , its just a few hours to mix it together 2) at the final step we do a division whereas we only need an inversion , we have a number of inversions , just a matter of making sure they are compatible , and mixing them together with some thresholds 3) There is a special algorithm just for cube root , we could try that. 4) Our code is actually a native inverse root ( ie division free) , perhaps a native root (with division) would be faster , I think asymtotically they are all the same run time. > * Fast exact division using David Harvey's Hensel division code (we've > already used his approximate quotient code to get fast tdiv_q code) > * Asymptotically fast division. > > If anyone is keen for a challenge for the next release, mpir 1.4, > those would be the main areas where we could really use a boost. > > Bill. --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---
