Getting the most out of your Apps: Knowing how to use the G4 and G5 compiler flags for best performance
A great majority of scientific apps come in the form of source code that must first be compiled prior to installation. For the most part this means running the standard routine of:
./configure
make
make install
However most people fail to realize that having the source code for an application is an amazing opportunity to optimize the application for your particular system. Very often these applications were initially intended for use on Linux or Solaris and were easily ported to the Mac because of OS X's UNIX underpinnings. Because of this the original author was not likely aware of various compiler optimizations available for Apple hardware when they created the Makefile. Through the inclusion of a few G4 or G5 compiler flags you could realize a huge gain in the runtime performance of your application. The following links provide a great overview of the G4 and G5 compiler optimizations for gcc and how to use them when compiling source code, or to keep in mind when you are writing a high performance application for OS X.
Optimizing on the Mac
G5 Performance primer
Tuning for the G5
Comments
compiler flags for Intel SSE?
Does GCC already have compiler flags for intel SSE optimization like it does for G5?
Don't quote me, but yes
I believe recent versions of gcc have some (early) attempts at auto-vectorisation. But don't quote me! (Good thing we have anonymous posting...!)
I still feel techy about Apple moving off the G5s: I've always been under the impression that bus architecture and related issues are more of an issue than raw CPU performance these days... (?)