Ok,
some of you received single processors G5 PM (I am waiting my dual but the date slipped!). It would be useful, instead of benchmarking these machine with xbench, to take C/C++ or Fortrand code you wrote and recompile it in two versions: the first with gcc that comes with OS X ADT, the second one with IBM (beta) compilers specific for G5/OS X (
http://www14.software.ibm.com/webapp...amp;rs=vacpp3, http://www14.software.ibm.com/webapp...=y&rs=xlf3).
Processor and disk intensive benchmarking will test processor/disk/bandwith capabilities and the two compilers (you will get a big speed improvement with IBM compilers because they are specific for the G5).
The ideal code would be a scientific calculation (numerical computation) that access a database to store or retrieve data.
As a second exercice (and for dual G5 owners lust) try to check if IBM compilers (like Intel specific) adhere to OpenMP (check
www.openmp.com, you will find .PDF/.PPT docs with code).
OpenMP, thru precompiler directives, automatically parallelize your code. Check all the features of IBM comilers and let me (us) know. I'm too excited!
Sam.
Top
