Re: [Openexr-devel] A few fixes before the 1.1.0 release

[Drew Hess]

It's easy to have autoconf detect whether the compiler supports "restrict" 
or not and choose the appropriate flag when building OpenEXR.

The problem occurs when the developer is using Imath in her own project.  
We can't put the flag in the pkg-config file because the developer can use 
a different compiler to compile the app than the compiler used to compile 
OpenEXR itself.  So that puts on the burden on the developer to use the 
right flag in order to get Imath to compile, which I don't wanna do.

I guess I'll set up an ImathPlatform.h file with #ifdef's for the various 
compilers we support.


-dwh-



On Mon, 2 Feb 2004, Kevin Wheatley wrote:

> Drew Hess wrote:
> > I agree that the OpenEXR file format should be as high-performance as
> > possible, but OpenEXR only uses Imath for portability and a few
> > attributes, so eliminating the restrict keywords from ImathMatrix.cpp
> > won't have any noticeable effect on OpenEXR performance.  Like I said, I'm
> > not even convinced it makes a noticeable difference on anything except on
> > IRIX.
> 
> Intel's C++ compilers will perform additional vectorisation if you use
> the restrict keyword also. It's better than messing around with
> Compiler flags to force the pointer aliasing to be discounted. It also
> helps to ensure memory alignement to get the best out of this.
> 
> '-restrict' on the Linux command line will enable restrict support.
> '-Qrestrict' on the Windows command line will enable restrict support.
> 
> Kevin
> 
>