This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: parallelized 'ld'?
- From: Alan Modra <amodra at bigpond dot net dot au>
- To: DJ Delorie <dj at redhat dot com>
- Cc: cynbe at muq dot org, binutils at sources dot redhat dot com, asmundak at cisco dot com
- Date: Wed, 27 Aug 2003 12:38:20 +0930
- Subject: Re: parallelized 'ld'?
- References: <200307150234.h6F2YsNW028337@tully.CS.Berkeley.EDU><87d6f63jes.fsf@muq.org> <200308261804.h7QI4VW26104@greed.delorie.com>
On Tue, Aug 26, 2003 at 02:04:31PM -0400, DJ Delorie wrote:
>
> > Automatically expanding hashtables when they reach ludicrous overload
> > levels isn't that hard -- I've been doing it without a second thought
> > in everything I've written for the last twenty years. No?
>
> Here's a quick patch to implement that, at least in a naive way. See
I'd suggest starting with a smaller table size. Many object files only
have a small number of symbols. For example, I took a look at a glibc
build and found the average symbol table size for .o files was only 27
symbols. Many of these are not globals so won't find their way into
the linker hash table.
--
Alan Modra
IBM OzLabs - Linux Technology Centre