I asked because I just implemented some profiling in XEmacs, and I
was trying to figure out a way to get more accurate timings. If
getrusage() had so much overhead, it might not be useful, but it
would still help immensely if there were a way of getting SIGPROF's
to occur more than 100 times per second.
If I change HZ to 1000 and recompile my kernel, will this work?
How about a way to change this without recompiling the kernel?
E.g.
-- make HZ be a variable instead of a preprocessor constant.
-- add a system call (callable by root only, of course) that
changes the frequency of the system clock.
I have no idea how difficult this would be, i.e. if it would wreak
havoc on some device drivers.
ben