On 01/20/21 12:50 PM, Michael Schuster wrote:
what you're describing is memory overcommit, which Solaris never did (at
least not while I was at Sun).
Good. BTW HP or IBM Unix was the first one I saw do this and they even
had special signals, to tell you. Sorry you app had to die as we lied
about how much memory you could have. It was v annoying.
If memory serves, pages weren't cleaned at reservation (ie when "malloc"
was called), but only when actually accessed ("zfod" - zero fill on
demand). Memory*did* have to be accounted for ("reserved" as the name
says), so you'd know at malloc-time if you were out of memory, and not when
someone tried to access it later.
Thanks for the clarifications.
_______________________________________________
openindiana-discuss mailing list
[email protected]
https://openindiana.org/mailman/listinfo/openindiana-discuss
