I have written my own ado file, which spends most of its time in a forvalues loop. The loop uses either reg or sureg (which you folks have parallelized), but seems to run at much the same speed as in Stata 11. I'm guessing that the routine must be spending much of its compute time on drawnorm or correlate commands. Are these high on the list for parallelization? Also, is there any prospect of commands which will allow us to parallelize a forvalues loop?
↧