All of this sounds AWESOME! Looking forward to the new version.
Best, Alex On Wed, Nov 25, 2009 at 01:07:15PM -0500, Michael Rubinstein wrote: > > Thanks for the update. > > I've spent the past few weeks making major improvements to lcalc. > I plan to release this updated version in a few weeks. > > 1) I got rid of the deprecated header files and the unused variables > so it compiles much cleaner. > > 2) I got lcalc to compile and run with Bailey's double double and quad double, > and also with mpfr (using a newer c++ wrapper for that), > so it now works in arbitrary precision. This took some effort to make sure > all the routines work in higher precision and to deal with some quirks > of each package. > > 3) I wrote my own cos and sin which is 4 times faster than my machine's (at > least > on my core2 laptop) and more than twice as fast Bailey's. It uses table a > lookup table of Taylor series. I spent two days hacking it as much as I could. > This makes some key routines 2-3 times faster. > > 4) I now handle values of s anywhere in the complex plane (rather than a half > plane). I had previously avoided slight complications due to trivial zeros > cancelling > out poles of the gamma factors in Lambda(s) (completed L-function). > > 5) I improved the derivatives option substantially using formulas for central > differences that give much better precision and allow for higher derivatives > numerically. > > 6) I am in the process of finishing to wire in Hiary's band limited > interpolation > routine for the Riemann zeta function when many values are needed. > At height 10^12 it runs 15 times faster than the straight forward summation > of the > Riemann Siegel formula, and at height 10^16 it runs a few hundred times > faster. > > 7) I cleaned up and reorganized the makefile and am adding a 'make test' > option that will check whether routines are running to the expected precision. > > 8) I fixed a couple of bugs that were affecting precision slightly. > > 9) There is a subtle bug that appears when looking for zeros of L-functions- > sometimes (very rarely) a zero is duplicated and another zero is skipped. > I'll be fixing that up in a little while. > > 10) Now that openmp is available on gcc with most machines I am going > to make use of that (as a compile option) in some of the key routines. > > 11) Many other small improvements are planned for later this year: > application of the fft and band limited interpolation to make > zero finding much faster (by a few orders of magnitude- more > improvement the further one goes). > > About the guy who asked about number fields- I might try to get this > to work in my code using pari's routines (not immediately), but having him > try it in sage with Rishi's wrapper would also be a good idea. > > Best, > Mike > > > > > > On Wed, 25 Nov 2009, William Stein wrote: > > > On Wed, Nov 25, 2009 at 8:24 AM, Dr. David Kirkby > > <david.kir...@onetel.net> wrote: > >> Pablo De Napoli wrote: > >>> It would be really nice if we can include the new weapper for the lcalc > >>> library. > >>> > >>> http://trac.sagemath.org/sage_trac/ticket/5396 > >>> > >>> [marked ad "with patch, neeeds review"] > >>> > >>> Pablo > >> > >> I have been less than impressed with lcalc itself. Of all the packages I > >> have > > > > For balance, I have to comment that I'm more than impressed by what > > lcalc itself actually *does*. It's by far the world's best program at > > computing zeros of L-functions (such as the Riemann Zeta function), > > far surpassing anything available in any other general purpose (or > > otherwise) math software, as far as I know. > > > > It would indeed be very nice to have much more functionality from > > Lcalc exposed in Sage, so I'm glad for 5396 getting finished, finally! > > Of course, I'm also glad David pointed out the various quality > > issues, and hope they can get addressed too. Mike -- maybe you can > > hire an undergrad CS major using our grant to knock them out? They > > don't require any real knowledge of math. > > > > -- William > > > >> met in Sage, it seems to have presented with more than its fair share of > >> problems. > >> > >> * At one point there was a non-portable option to suppress warnings from > >> the > >> assembler. I fixed that, so it would actually build on Solaris with gcc, > >> with > >> either the Sun or GNU assembler. > >> > >> http://trac.sagemath.org/sage_trac/ticket/6609 > >> > >> * I note it used the -no-deprecated (or similar option) to hide the fact > >> the > >> code is deprecated. > >> > >> * It will not build on Solaris with the Sun compiler > >> http://trac.sagemath.org/sage_trac/ticket/7065 > >> > >> * It will not build on HP-UX with gcc > >> http://trac.sagemath.org/sage_trac/ticket/7178 > >> > >> * If you run 'lint' on the sources, one find numerous variables that are > >> declared but never used. > >> > >> IMHO, lcalc needs a major overhall. > >> > >> I always find it worrying when the source is deprecated, efforts are made > >> to > >> suppress warnings, it is not portable, variables are unused ...It always > >> makes > >> me wonder what other issues might exist, which I do not know about. > >> > >> > >> Dave > >> > >> > >> -- > >> 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 > > > > > > > > -- > > William Stein > > Associate Professor of Mathematics > > University of Washington > > http://wstein.org > > > > -- > 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 -- Alex Ghitza -- Lecturer in Mathematics -- The University of Melbourne -- Australia -- http://www.ms.unimelb.edu.au/~aghitza/ -- 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
