Re: [VANILLA-LIST:2200] Ouch

[Bob Tanner <tanner@real-time.com>]

Quoting Dave Ahn (ahn@vec.wfubmc.edu):
> On Thu, Mar 25, 1999 at 08:24:39PM +0100, Kurt Siegl wrote:
> > 
> > sprintf is a huge performance leak as well. In COW we replaced all time
> > critical sprintf calls with hand coded string utilities which helped a lot.
> 
> I don't think it is as important in the server.  Aside from the fact that
> computers are getting ever faster (any Sparc IPC's or 386-16 PC's out there?),
> the only significant usage of sprintf in the server is the messaging.  If
> we ever get around to redoing the entire messaging system, we could make it
> more efficient, but I don't see any real reason to change existing sprintf
> code to string utils if all it saves is a few microcpuseconds each call.
> 
> I am in favor of sacrificing some performance for better code and ANSI/POSIX
> compliance.  Portability and GPP is more important than hand optimized code
> that lets us run the server on an obsolete box.
> 

Well, I thought the hand coded string utils would be just as good of a
solution as snprintf().

-- 
Bob Tanner <tanner@real-time.com>       | Phone : (612)943-8700
http://www.real-time.com                | Fax   : (612)943-8500
Key fingerprint =  6C E9 51 4F D5 3E 4C 66 62 A9 10 E5 35 85 39 D9