David Korn wrote: > > I am confused about the whole issue of isaexec() and 64 bit ksh93 > and whether /bin/sh should use this. > > Here are the questions I have: > 1. Why is a 64 bit address space needed? The current 32 bit > version handles 64 bit files and also uses long double for > arithmetic so the only gain would be a larger virtual space. > Are there applications that need this?
Yes, for example some of the NIH bioshell stuff employes a hacked ksh93 for postprocessing and those data can only be described as gigantic. We've seen 8GB heap usage on our (quite small) machines and postprocessing larger datasets require even more heap. Another very important point are loadable builtins which may require a 64bit address space. This is why we're delivering both a 32bit and 64bit version of ksh93 even on the 64bit-only Solaris/SPARC platform - if a plugin library is not available as 64bit the 32bit ksh93 version is needed. [snip] > 2. Assuming that a 64-bit address space is desired and that > measurement shows little performance degradation, why > wouldn't this be /bin/sh rather than having both the 32 and > 64 bit version? See above. > 3. If the isaexec() functionality is desired, why wouldn't > this be built into the internal PATH search algorithmr? Erm, I have a patch for that to do such a selection within isaexec. IMO it is not usefull to do this within the shell - it would only compliciate the matter. ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)
