On Tue, Oct 01, 2002 at 03:48:00PM -0400, Lord Isildur wrote:
> doesnt vms do something like this for some of their equivalent of shared 
> libs? 
> isildur
> 
> On Tue, 1 Oct 2002, Wojciech Puchar wrote:
>
> > > >
> > > > Shared libraries and the dynamic linker *already* share in this way...
> > >
> > > OTOH mapping libc at a fixed and reserved address would, on some
> > > architectures at least, allow the code pages to use global TLB
> > > entries.  This would mean that cache lines could be shared between
> > > different processes.
> > 
> > and lot of time saved of ld_elf.so work.

I think MVS has something similar in it's LPA. It makes things faster,
and also takes up crazy amounts of space if you ever get a text crash
dump of a program -- the entire LPA shows up in the printout. 

Oh, and watch MVS people squirm when they have to update the LPA at
run time without rebooting, uh, without a re-IPL. Don't mess up! Also,
don't delete datasets in the LPA, and applying maint. gets hairy as
well. 

From reading about people who get/have to deal with the LPA, it sounds
like a interesting optimization. It also sounds like a real PITA, and a
big-ole' length of rope at that.
-- 
Kevin P. Neal                                http://www.pobox.com/~kpn/

"It sounded pretty good, but it's hard to tell how it will work out
in practice." -- Dennis Ritchie, ~1977, "Summary of a DEC 32-bit machine"