AFNI on Mac + Intel
The new macosx_10.4_Intel.tgz binaries package contains AFNI (et al.) compiled on an Intel-based iMac. With the Duo processor, 3dDeconvolve runs nearly twice as fast using the '-jobs 2' option.
For installation details, see http://afni.nimh.nih.gov/pub/dist/HOWTO//howto/ht00_inst/html/index.shtml
Also see this message board post for the announcement.
Interestingly, the Accelerate framework didn't speed up 3dDeconvolve_f at all (unlike on the G4 and G5 CPUs), and so the binaries on the system are not compiled with vectorization. For some timing results, see the bottom of the page http://afni.nimh.nih.gov/afni/doc/misc/afni_speed/index_html/ -- where the new longer-running speed test results can be found.




