[Monotone-devel] Re: why is monotone so slow?

[graydon hoare]

On 14 Sep 2004 13:15:26 +0200, Peter Simons <address@hidden> wrote:

> I have a related question: Is there any chance of compiling
> Monotone with more aggressive optimizing flags than -O2? It
> appears that using -O3 (I suspect: in-lining) breaks the
> code when compiling with gcc 3.3.4: Almost all "make check"
> tests fail when I try that.

I can hazard a guess that it's cryptopp. did you compile with
just -O3 or -O3 -fno-strict-aliasing? the latter is absolutely
*required* for cryptopp; it does union member re-casting all
over the place and will break alias analysis.

anyways, if that doesn't work I can look at it further to see
if there's anything really obvious, but my plan is to look
instead at replacing cryptopp with botan. it looks to be more
maintainable and well-written code, but I can't tell if there
are lurking troubles with that plan.

> I seem to recall this this is a known problem, but is it
> also a problem that will be fixed "real soon now"?

not real soon now, unless someone wants to make a botan-testing
branch and give it a go (the upside is it will be quite isolated
work and likely will merge after the fact easily; the downside
is you'll be doing compatibility testing between two certificate
and key formats. yuck).

if it's left to me, this will be several months off. there's
a lot of more pressing stuff on my list.

-graydon