Casper.Dik at Sun.COM wrote: > >In any even, it is silly to talk about the overhead without > >any measurements. > > Certainly we know from our own testing that making "/bin/ls" > an isaexec wrapped program was costly, so we did ship /usr/bin/sparcv9/ls > but did not make /bin/ls an isaexec wrapped binary. > > So the cost can be measured (but shells are bigger and slower to > start and the cost of isaexec done in another location might > well be negligble)
It is negligble. I've been trying to say that over and over again. Using /usr/lib/isaexec has a price, however this overhead is completely canceled out when you hit one single builtin command or a construct "( ... )". Other shells fork() in this case while ksh93 doesn't - and that is more compensation than anything consumed by isaexec. Note that we tested this on our side over and over again. It's working, the overhead is negligble (both isaexec and 32bit vs. 64bit code) and IMO the usage of a 64bit shell as default on a 64bit platform is a HUGE step forward for Solaris. ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)