type traits, reflection, and whatnot

Gabriel Dos Reis gdr@integrable-solutions.net
Wed Jan 15 22:27:00 GMT 2003


Mark Mitchell <mark@codesourcery.com> writes:

| --On Tuesday, January 14, 2003 9:20 AM -0500 Jason Merrill
| <jason@redhat.com> wrote:
| 
| > On 14 Jan 2003 09:11:15 +0100, Gabriel Dos Reis
| > <gdr@integrable-solutions.net> wrote:
| >
| >> I was (and still is) optimistic.  Let's see where we end up at the end
| >> of the week-end.  I don't know whether other C++ folks have planed
| >> substantive work on the new parser.  Jason, Mark?
| 
| We're working on reducing the memory usage and compile times for templates,
| which does have some parser impact.  We're also working on fixing some
| correctness issues (not regressions) regarding dependent types, lookup,
| and related mangling issues.  Those too have some parser impact.

OK.  I've noted the name lookup issue (see my other patch) and how it
impacts on the overall front-end perfomance.  I would like to extend
the existing timevar framework to include time-report for at least:

   * class and object layout
   * template instantiation
   * overloading resolution

Reducing the "everything is a tree" paradigm to the struct minimum
should enable considerable correctness and performance gains also.

| But, nothing major -- I think the overall shape of the parser is as it will
| be for the forseeable future, at least from our point of view.

Thanks a lot!

-- Gaby



More information about the Libstdc++ mailing list