[BioC] memory trouble in using loess

[echang4 at life.uiuc.edu]

Dear BioConductor users,

      I am trying to analyze my Affy U133A chips (38 total) using the
gcrma protocol but substituting loess instead of quantile
normalization. I am doing this on a remote UNIX server using R 1.9.1
 and BioConductor 1.4. I have already processed this dataset using
>eset<- gcrma(dat)
There was no problem at all. But when I substituted the QN for loess, it
took almost 6 hours to complete the normalization!
      I ran into memory trouble after all the comparisons were done and
got the error message (Failure to allocate a vector of ~145,000KB.)
Is there any way of working around this problem? Because I used the
expresso function in the following order-gcrma correction, pmonly,
loess, and medianpolish. I suspect the memory failure is when R is
trying to summarize the probesets. Can I summarize the probesets
before doing loess normalization? That way the memory usage is less?
Is the operation order still "legal"?

Any suggestions will be great! Thank you for your time,
Edmund Chang