Re: Urgent: Manipulating large libraries
On Wed, 20 Aug 1997 mdorman@debian.org wrote:
> [...]
> Well, a Jaz drive isn't all *that* speedy. Where/why are you using
> ranlib at all? I am 100% certain GNU ar doesn't need it. Could it be
> that ranlib on the rs/6000 is effectively a NOP?
Ah .... I've been porting to Linux a multi-arch package. Some platforms
don't need ranlib, I wasn't aware that Linux was one of them. Am I now
right in thinking that, if I use the -s modifier to ar, I never need
ranlib?
> > I'm rather worried. Swapping was not excessive - I have 24MB RAM,
> > and at most 2MB of swap was in use. The disk was going most of the
> > time, so I think ranlib on my PC was reading the huge lib in each
> > time, whereas on the 43P (which has 192 MB), it was held in memory.
>
> Yes, but during that operation, how much was buffered/cached according
> to, say, top. I'd suspect not a lot.
Please ignore my earlier worries - I tracked down the problem to 19
ranlibs on the same library when only 1 (or in fact 0) were required!
> I can summarize my thoughts:
>
> 1) Don't bother to use ranlib, with gnu ar you shouldn't need it.
Provided -s?
> 2) The AIX ranlib may not have actually been doing anything.
Not sure ... possibly.
> 3) If the AIX ranlib *was* doing something, it may have been caching
> (which your machine probably doesn't have a whole lot of memory
> for, with only 24MB)
Quite.
> 4) Unless you're talking a dramatic difference in processor speed,
> always get more memory.
Right, I'm now down to a generality (ranlib worries out of the way). The
choice is:
366MHz 21164, 1MB cache, 128MB DRAM;
433MHz 21164, 2MB cache (Microway claim 30% performance boost), 64MB
DRAM.
Any thoughts? I'll probably have to decide tomorrow ...
> The only way to make sure you're doing the right thing, though, would
> be to do some concerted testing.
Sure. Unfortunately ...
Many thanks,
Nikhil.
--
Nikhil Nair
Trinity College, Cambridge, England
Tel.: +44 1223 368353
Email: nn201@cus.cam.ac.uk
nnair@debian.org
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-alpha-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: