Re: [Openexr-devel] half

[Paul Schneider]

> I'd like to keep the generation of the half<->float lookup tables a
> compile-time thing.  For some reason, this step is really slow on my 
> OS X
> box.  It might be due to the fact that it's a 400MHz G4, or it might 
> be an
> OS X thing; I haven't had time to investigate.  Anyway, I don't want to
> have to pay the cost of this at runtime when it can easily be done at
> compile time instead.  It may not matter for some long-running
> applications, but simple, fast-running utility programs could be 
> adversely
> affected.

I agree.  It's pretty slow on my Mac (733 Mhz), and it's pretty darn 
noticeable on my Linux machine as well (1.8 Ghz).  This step is 
automatic with the unix make system; I'm sure it's possible to automate 
it in VC++ as well.

I think Jukka also mentioned replacing the system-style includes 
(#include <file>) with user-style includes (#include "file").  I'm 
actually for this - we consider EXR a "standard" library at ILM, but I 
think most 3rd party developers will install it as part of their own 
code, not as a system-level thing.  Unix style compilers (cc, gcc) 
don't differentiate between the two at all, as far as I know, but other 
development systems definitely do, and while it's possible to tell the 
system where to look, it's one more thing to be aware of when trying to 
get the source to build on different systems.  Comments?

- Paul