Peter Tribble wrote: > On Feb 20, 2008 6:32 AM, Matt Ingenthron <Matt.Ingenthron at sun.com> wrote: >> I keep meaning to send out a pointer on this one. I don't know if it's >> been noticed, but there was a slight thread of discussion on 32/64 after >> I wrote that blog on using svccfg to switch from 32 to 64-bit: >> http://blogs.sun.com/mingenthron/entry/opensolaris_web_stack_setting_services#comments >> >> Based on those comments, I still don't think isaexec is appropriate >> here, but I'm very much out of my area. There could be a good argument >> here I'm not privy to. >> >> He makes a good point on the 64-bit being faster on by default on x64 >> based systems. >> >> Before this arc case goes forward adopting "enable_64bit", I thought >> this was worth bring up on the list to discuss. Is there any case here >> where isaexec makes sense? >> >> I believe we'll need to deliver both, and there are probably still a >> number of reasons (for now) that one may choose to run 32-bit even on a >> system with a 64-bit ISA, which means there has to be a switch somewhere >> to pick what one wants to run. > > If you have an x64 system then running in 64-bit is clearly the > right choice, so isaexec would work fine and would probably > be the right choice there. > > It might be useful to do a comparison on sparc. But my general > experience there is that 64-bit isn't generally better and is only > usually worth it if you need the extra addressability. >
Yes, but I think it should be done the same way on sparc as on x64, to have 64bit as default only on amd64 platforms will be very confusing. I would also prefer that this property had the same name and default as other components in Webstack, although only Apache have that choice. Regards, Jan S
