At work, the 64bit Power4/Power5 IBM AIX systems come with 32-bit Perl. After reading the README, I wasn't interested in building everything in 64-bit, so when I build a Perl to build DBI with, I build it with similar settings to the vendor's (unsupported, "contributed") 32-bit Perl. If one of my users needs more than 4GB address space in a DBI script, I'm going to suggest a change of algorithm might be in order!
> Does your app spend a significant amount of its time splitting huge > numbers into 32-bit "chunks" so it can cope with them? This would likely be indicated by using Math::Big[Something] packages. Or, do you have a lot of FLOATING POINT math, such that you are or should be using PDL, the Perl Data Language? Do you need the longer floating point registers, assuming your platform's 64-bit mode has longer such? On my home machine -- nothing to do with work -- I have 64bit Perl since that's all that's available with Debian Alpha. I'm going to try some PDL on it. LIMITS -- As near as Google can tell, the limit on 32bit is 2GB for filesystem (signed numbers! Fie!), 4GB for process memory (unsigned, yeah!). http://www.google.com/search?hl=en&lr=&q=Perl+32-bit+2-GB+OR+4-GB+x86 (Perl6 may have a 2GB limit on the bytecode file in 32bit, but program ought to be LESS than half your address space or you're in trouble enough already!) Bill Ricker Member, Boston.PM NOT speaking for the Firm. _______________________________________________ Boston-pm mailing list [email protected] http://mail.pm.org/mailman/listinfo/boston-pm
