httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "William A. Rowe, Jr." <wr...@lnd.com>
Subject RE: request_rec change inside ifdef APACHE_XLATE (?)
Date Tue, 16 May 2000 20:00:03 GMT
> From: Greg Stein [mailto:gstein@lyra.org]
> Sent: Tuesday, May 16, 2000 2:00 PM
> 
> On Tue, 16 May 2000, William A. Rowe, Jr. wrote:
> > > From: Greg Stein [mailto:gstein@lyra.org]
> > > Sent: Tuesday, May 16, 2000 3:37 AM
> > > 
> > > There is very little reason to use bitfields in this 
> > > structure. This is a
> > > runtime structure, allocated once per request. All the 
> > > bitfields will do
> > > is to bungle up your code (admittedly, a little bit) 
> later on when you
> > > get/set the values.
> > > 
> > > IMO, bitfields are useful only when you are trying to *really* be
> > > conservative about your memory usage. Even then, I might 
> tend to use
> > > bitmasks.
> > 
> > Agreed to your point in this case.  But taking the example 
> of a set of
> > flags that have many instances (ergo - need memory 
> conservation), can't
> > we assume any compiler worth it's salt is playing a machine 
> optimized
> > bit flag game that is most appropriate to that cpu?
> > 
> > Sometimes ya just gotta presume that the compiler knows its 
> cpu better
> > than we (cross platform app) would.
> 
> That is why I qualified my response as "admittedly, a little 
> bit." I was
> attempting to prevent a berating from Dean, pointing out that 
> all modern
> processors have efficient bit-testing opcodes. :-)
> 
> Of course, if you're trying to run Apache on a 6502, then avoiding bit
> fields will be helpful.
> 
> hehe...
> 

I know... they are a pain on the 1802 port I'm writing, of course the
8 bit native int's are more annoying :-)



Mime
View raw message