This is the mail archive of the
mailing list for the libstdc++ project.
Re: [Patch]: Corrected patch for stl_tree.h to improve performanceand memory usage
- From: Benjamin Kosnik <bkoz at redhat dot com>
- To: Loren James Rittle <rittle at latour dot rsch dot comm dot mot dot com>
- Cc: libstdc++ at gcc dot gnu dot org, gp dot bolton at computer dot org, green at redhat dot com
- Date: Thu, 6 Mar 2003 11:53:25 -0600
- Subject: Re: [Patch]: Corrected patch for stl_tree.h to improve performanceand memory usage
- References: <3E667887.email@example.com><200303060055.h260tP53002136@latour.rsch.comm.mot.com>
>Again, sweet work. I think you have nailed all of the performance
>issues raised. If you ever answered the first time I asked in regards
>to your std::list improvements back around Feb 12, I'm sorry I missed
Ditto! This stuff looks good.
>First, do you[/Nortel --- where you posted from last month] have an
>assignment for GCC on file with the FSF? Based on the informal
>10-line rule, your work appears to be beyond a simple bug/performance
>fix. Opinions on this point by other library maintainers?
I don't know off hand. I could go either way with it.
It would probably be best to get an assignment though.
Gawain, Phil Edwards will know the current status of gcc/libstdc++
assignments. I know that contributions from Nortel (or their
subsidiaries) have been accepted in the past, and that in general,
Nortel seems quite friendly to free software. Perhaps they have a
blanket company assignment already in place for all of gcc? I've cc'd
Anthony Green on this email, who might know more details.
>Next, since this change hits a file included in application code and
>the libray ABI, we would need to consider when to apply it. If we
>strongly believe that the new implementation satisfies the standard in
>every way (and I believe that you have studied and reported on the
>issues well enough to satisfy me) and only has rare cases where
>performance could ever worsen in practice (I'm sold), then I'd vote
>before the 3.3 release.
You have my full agreement.
Loren, would you be willing to merge this and the std::list bits posted
previously, when the assignment bits get answered?