Using either the 64-bit binary or compiling from source as 64-bit, I’m having a problem with running classify.seqs. When “Calculating template probabilities”, I see my memory usage go from ~500M to > 4G and that task never completes (I’ve let it go over an hour). However, when I run the command with the same RDP training set and input fasta on our linux server, the memory at the probabilities step stays ~500M, completes in just over 1 minute, and classifies my sequences in about 20 seconds.
My Macbook has 8G of ram and 4 processors, so I can’t see it being a hardware limitation. Sampling the process indicates maybe something to do with __memcpy? I’m running 10.6.7.