Actually it was released before we released - we knew it was coming, hence the MPIR 1.1 release. I had assumed that jernej's question was in relation to the GMP 4.3.0 release.
Obviously some of us have spent some time off list analysing their release. There's a lot of specific improvements in GMP that show up in the benchmarks (and some that don't): * An improved FFT * Many assembly improvements for Core 2, Atom, AMD 64, Pentium 4 * Toom 4 and unbalanced variants * Divide and conquer division and the so-called "misunderstanding algorithm" * Fast GCD and XGCD * Faster 1 and 2 limb division * redc basecase * exact division by certain constants * other things We've targeted slightly different things in MPIR so far: * Assembly improvements for Core 2, i7, K8, K10, Pentium 4 * Better MIPS support * Better CPU detection code, especially for x86_64 * Toom 4 and Toom 7 (balanced only) * Fast GCD * Some improvements to 1 and 2 limb division * Better build support for Apple and Sun compilers * MSVC support * redc basecase * exact division by certain constants * other things There's been some off list discussion about working on further assembly optimisation in MPIR. So far only our K8 code is close to optimal (though still not quite there). Jason Moxham is the guy to speak to about that. Specifically we don't use our new addmul_2 code for K8 yet. In the next few releases we'll have better assembly support for other x86_64 cpus. Jason and I have been discussing improving the FFT. We think putting some version of Paul Zimmermann's et al's FFT in MPIR as a stop gap until the new MPIR FFT is ready is probably sensible. We've been discussing possible improvements of that. I'm giving some thought as to whether a polynomial divide and conquer division algorithm I invented last year can be turned into an integer version. I don't know about the other MPIR developers, but I'm not personally planning on a new release in the next two weeks. But our intention has generally been to make smaller releases at regular intervals. Bill. On 15 Apr, 17:18, Jeff Gilchrist <jeff.gilchr...@gmail.com> wrote: > On Tue, Apr 14, 2009 at 8:22 PM, Bill Hart <goodwillh...@googlemail.com> > wrote: > > We are about to release MPIR 1.1 which will improve our benchmarks, so > > a comparison would be premature. You can check our MPIR 1.1 code out > > from svn if you want to time it. Just do make bench in the bench > > directory to run the benchmark utility. > > Looks like GMP 4.3.0 was also just released:http://gmplib.org/gmp4.3.html > > Jeff. --~--~---------~--~----~------------~-------~--~----~ 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 -~----------~----~----~----~------~----~------~--~---
