I'm using R-1.6.1 on a dual Xeon system running Windows XP.
When I have two compute intensive programs running in the
background, and I start a calculation in an interactive R
session, that calculation gets just a few percent of the
available CPU time. This is true even if I use the task
manager to set the priority of the background jobs to "low"
and the R session to "high". Scheduling of the background
jobs seems to behave as expected; if I start more than two,
then jobs with higher priority get more CPU time than jobs
with lower priority. So it seems to be R that is behaving
strangely.
R seems to get its full allotment of CPU time when I invoke
single internal functions that take a significant amount of
time to run. It is starved when I run more complex R code
that is accumulating results of several thousand small
linear regressions.
Is this a known issue? I'm not sure what could cause this;
perhaps R is yielding the processor when the GUI checks for
interactive events, and is getting starved for CPU time
because it is yielding too frequently? If these GUI checks
happen between execution of R statements, then perhaps there
should be a way to throttle the frequency of the checks, or
an option to disable them?
-- Dave Hinds
