Re: New GLib string routines

[Raph Levien <raph acm org>]

Jeff Garzik wrote:
> 
> I'm about to add two new macros to GLib, g_strdup_a, and g_strconcat_a.
> By using alloca, they benchmark significantly faster than g_strdup and
> g_strconcat.
> 
> Timing test results, strconcat test program, and patch are attached.

I'd like to see profile data that suggests users will actually be able
to see a difference. My tests show that a malloc/free roundtrip is about
a microsecond on my PII 233.

There are a lot of serious speed issues remaining in gnome-libs,
particularly on startup. I'd really rather see effort going into
identifying and resolving those than random optimizations of
questionable value, at least for right now.

Raph