PATCH: Use a hash table for 26X linker speedup

[Alan Modra]

On Fri, May 27, 2005 at 09:41:02AM -0700, H. J. Lu wrote:
> > > > I will work on lang_output_section_find_1 next. I am planning to use
> > > > a hash table. I hope to enable 64k section on all ELF targets.
> 
> Here is the updated patch. I have been using this for a while. Is there
> any objection?

I think I made this comment before:  It seems wasteful to have two hash
tables working with virtually the same set of data.  We have a section
hash table holding output section names, and you're adding another hash
table to hold output_section_statement names.  I realize the two aren't
identical, but all orphans will be in both tables (and it is our orphan
section handling that makes the output_section_statement list so long).
I'd prefer if you don't commit this patch, at least until I've had time
to look at changing orphan section handling.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre