Helmut Zeisel ([EMAIL PROTECTED]) writes: >Can you please give me some information about >your Mersenne code for a DEC Alpha running Linux? The source is at ftp://nigel.mae.cwru.edu/pub/mayer/Mlucas_2.6.f90.gz; the README file in the same directory tells how to use it. >As I understand, it is Fortran 90. >Do I need a Fortran 90 compiler >to use it on a DEC Alpha chip under Linux >or do you also have a binary distribution? Thanks to my SPEC98 contacts at DEC, er, Compaq, as of today I have a Linux executable, compiled using a pre-beta version of their soon-to-be- released F90 compiler for Alpha/Linux. This is in ftp://nigel.mae.cwru.edu/pub/mayer/bin/ALPHA_LINUX/Mlucas_2_6X.gem4494u_linux.exe.gz; once you have it, use gunzip or gzip -d to uncompress it. Note that it's a pretty big file (618KB gzipped, 2MB uncompressed) because it consists of the basic executable bundled together with all the run-time library (RTL) files it needs to run. Please make sure to run the self-tests described in the README file. At this point, I must insert a word from my sponsors: * * * * * * Compaq Computer Corporation has said that it will release a Fortran compiler for Alpha Linux during 1999. (See http://www.compaq.com/newsroom/pr/1999/pr060499a.html for the press release.) MLucas V2.6 has been compiled with a pre-beta version of this compiler. The executable is available for users of Alpha Linux systems, with the following conditions: 1) This text must be included with all distributions of the executable and prominently displayed at download locations. 2) Please understand that the compiler is pre-beta. While we do not know of any bugs that would affect the correct execution of MLucas, surprises are certainly possible. 3) If you do encounter any surprises, please send a complete description of the circumstances to the author of MLucas, Ernst Mayer, [EMAIL PROTECTED] , including information about your Linux environment (e.g. which software distribution) and information about your hardware (exact model number as printed on the back of the case would be very helpful). 4) Usage must be summarized to the author. Whether you have any surprises or not, please drop a line to [EMAIL PROTECTED] to let us know how things work out. * * * * * * I have only Unix on my Alphas, so have been unable to do timings myself, but the folks at DEC, er, Compaq assure me it's about the same speed as the Unix version compiled with the native compiler. >Do you have any numbers how fast it runs >compared to George Woltran's MPrime on >an Intel chip? Under DEC Unix, it's 1.5-2 times slower than George's Prime95 at the same clock speed. On SGI (for which executables are also available), it's about equal at the same clock speed. >On what ranges of exponents does it work? The same as Prime95, i.e. up to 20 million or so. It also uses similar FFT lengths (i.e. of form 2^n, 3*2^n, 5*2^n, 7*2^n) for the various exponent ranges, although Prime95 can go about 1% higher for a given FFT length due to the extra accuracy of the Intel 80-bit register type. Happy hunting, Ernst ________________________________________________________________ Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
