Re: [R] 64 bit R slower than 32 bit R on Sun Sparc Solaris?
On Wed, 8 Sep 2004, Jason Liao wrote: > Thank you very much, Profs. Ripley and Peng! It corrected a big > misconception in my mind. > > By the way, does the Sun Forte 7 compiler produce faster R than the GCC > 3.4.1 compiler (which we use)? Not in my experience, although libsunperf helps a lot on some problems (but ATLAS seems competitive). I mainly use Forte 7 to test the validity of source code: gcc is far too generous, especially as is compiles `GNU Fortran' not ISO Fortran. > > Jason > > --- Prof Brian Ripley <[EMAIL PROTECTED]> wrote: > > > On Tue, 7 Sep 2004, Roger D. Peng wrote: > > > > > Are you using an optimized BLAS for both builds? That's one > > > possibility. Also, 64-bit builds use up more memory initially > > since the > > > pointers are bigger. I've tried both 64-bit and 32-bit builds on > > > Sparc/Solaris and haven't seen any slowdown. > > > > It uses more memory at all times and so gc() takes longer. There > > *is* a > > slowdown, for example 90 vs 80 secs for a run of R-devel's stats-Ex.R > > (for > > either Sun's Forte 7 or gcc 3.4.1 compilers). But `25-30% slower' is > > unexpected and needs investigation. > > > > The only difference between 32-bit and 64-bit versions of R on > > Solaris > > will be the size of the pointers and (probably) less efficient PIC > > code. > > There is no reason to expect a performance boost with 64-bit > > applications: > > they have to do more work and are only worthwhile if you need the > > address > > space (in memory or also on disc as 64-bit applications use large > > files > > natively). > > > > > > > Jason Liao wrote: > > > > > > > Hello, everyone! I guess no one is still using R on Sun Sparc > > these > > > > days. But our department has a (pretty new) two-CPU Sun server. > > We > > > > recently compiled R as a 64 bit application and expected a > > performance > > > > boost. But it runs 25-30% slower than the 32 bit version of R. > > Anyone > > > > knows why this is so? Thanks! > > > > -- > > Brian D. Ripley, [EMAIL PROTECTED] > > Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ > > University of Oxford, Tel: +44 1865 272861 (self) > > 1 South Parks Road, +44 1865 272866 (PA) > > Oxford OX1 3TG, UKFax: +44 1865 272595 > > > > > > > = > Jason Liao, http://www.geocities.com/jg_liao > Dept. of Biostatistics, http://www2.umdnj.edu/bmtrxweb > University of Medicine and Dentistry of New Jersey > phone 732-235-5429, School of Public Health office > phone 732-235-8611, Cancer Institute of New Jersey office > moble phone 908-720-4205 > > __ > [EMAIL PROTECTED] mailing list > https://stat.ethz.ch/mailman/listinfo/r-help > PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html > > -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] 64 bit R slower than 32 bit R on Sun Sparc Solaris?
Thank you very much, Profs. Ripley and Peng! It corrected a big misconception in my mind. By the way, does the Sun Forte 7 compiler produce faster R than the GCC 3.4.1 compiler (which we use)? Jason --- Prof Brian Ripley <[EMAIL PROTECTED]> wrote: > On Tue, 7 Sep 2004, Roger D. Peng wrote: > > > Are you using an optimized BLAS for both builds? That's one > > possibility. Also, 64-bit builds use up more memory initially > since the > > pointers are bigger. I've tried both 64-bit and 32-bit builds on > > Sparc/Solaris and haven't seen any slowdown. > > It uses more memory at all times and so gc() takes longer. There > *is* a > slowdown, for example 90 vs 80 secs for a run of R-devel's stats-Ex.R > (for > either Sun's Forte 7 or gcc 3.4.1 compilers). But `25-30% slower' is > unexpected and needs investigation. > > The only difference between 32-bit and 64-bit versions of R on > Solaris > will be the size of the pointers and (probably) less efficient PIC > code. > There is no reason to expect a performance boost with 64-bit > applications: > they have to do more work and are only worthwhile if you need the > address > space (in memory or also on disc as 64-bit applications use large > files > natively). > > > > Jason Liao wrote: > > > > > Hello, everyone! I guess no one is still using R on Sun Sparc > these > > > days. But our department has a (pretty new) two-CPU Sun server. > We > > > recently compiled R as a 64 bit application and expected a > performance > > > boost. But it runs 25-30% slower than the 32 bit version of R. > Anyone > > > knows why this is so? Thanks! > > -- > Brian D. Ripley, [EMAIL PROTECTED] > Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ > University of Oxford, Tel: +44 1865 272861 (self) > 1 South Parks Road, +44 1865 272866 (PA) > Oxford OX1 3TG, UKFax: +44 1865 272595 > > = Jason Liao, http://www.geocities.com/jg_liao Dept. of Biostatistics, http://www2.umdnj.edu/bmtrxweb University of Medicine and Dentistry of New Jersey phone 732-235-5429, School of Public Health office phone 732-235-8611, Cancer Institute of New Jersey office moble phone 908-720-4205 __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] 64 bit R slower than 32 bit R on Sun Sparc Solaris?
On Tue, 7 Sep 2004, Roger D. Peng wrote: > Are you using an optimized BLAS for both builds? That's one > possibility. Also, 64-bit builds use up more memory initially since the > pointers are bigger. I've tried both 64-bit and 32-bit builds on > Sparc/Solaris and haven't seen any slowdown. It uses more memory at all times and so gc() takes longer. There *is* a slowdown, for example 90 vs 80 secs for a run of R-devel's stats-Ex.R (for either Sun's Forte 7 or gcc 3.4.1 compilers). But `25-30% slower' is unexpected and needs investigation. The only difference between 32-bit and 64-bit versions of R on Solaris will be the size of the pointers and (probably) less efficient PIC code. There is no reason to expect a performance boost with 64-bit applications: they have to do more work and are only worthwhile if you need the address space (in memory or also on disc as 64-bit applications use large files natively). > Jason Liao wrote: > > > Hello, everyone! I guess no one is still using R on Sun Sparc these > > days. But our department has a (pretty new) two-CPU Sun server. We > > recently compiled R as a 64 bit application and expected a performance > > boost. But it runs 25-30% slower than the 32 bit version of R. Anyone > > knows why this is so? Thanks! -- Brian D. Ripley, [EMAIL PROTECTED] Professor of Applied Statistics, http://www.stats.ox.ac.uk/~ripley/ University of Oxford, Tel: +44 1865 272861 (self) 1 South Parks Road, +44 1865 272866 (PA) Oxford OX1 3TG, UKFax: +44 1865 272595 __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html
Re: [R] 64 bit R slower than 32 bit R on Sun Sparc Solaris?
Are you using an optimized BLAS for both builds? That's one possibility. Also, 64-bit builds use up more memory initially since the pointers are bigger. I've tried both 64-bit and 32-bit builds on Sparc/Solaris and haven't seen any slowdown. -roger Jason Liao wrote: Hello, everyone! I guess no one is still using R on Sun Sparc these days. But our department has a (pretty new) two-CPU Sun server. We recently compiled R as a 64 bit application and expected a performance boost. But it runs 25-30% slower than the 32 bit version of R. Anyone knows why this is so? Thanks! Jason = Jason Liao, http://www.geocities.com/jg_liao Dept. of Biostatistics, http://www2.umdnj.edu/bmtrxweb University of Medicine and Dentistry of New Jersey phone 732-235-5429, School of Public Health office phone 732-235-8611, Cancer Institute of New Jersey office moble phone 908-720-4205 __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html __ [EMAIL PROTECTED] mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide! http://www.R-project.org/posting-guide.html