(http://www.scl.ameslab.gov/Projects/HINT/index.html.
Hint runs different using data types on consecutively larger sized problems,
allowing the user to determine the effect of codes that stay in L1 cache, grow
to L2 cache, grow to size of ram memory, swapbound etc.
I ran the Float benchmark and monitored the CPU activity using xosview. The task
changed cpus about every 2 seconds. And this was the only user process running
on the machine. The only competition for cpu would come from xosview and the
system tasks. (Dual 450MHz xeon on asus XG-DLS mainboard, 512 mB memory, 2.2.5
kernel, RedHat 5.2).
I had assumed, maybe incorrectly, that the benchmark code would execute faster
if the content of L2 cache was preserved, at least for the code if not for the
data?
I will study the problem further. My hope was to understand how/if the size of
modern L2 cache memory was being factored into the scheduling decision. The
problem seems similar to that posed by distributed computing?
don
-- ------------------------------------------------------------------- | Don Fisher dfisher@as.arizona.edu | | MMT Observatory VOICE: (520)621-7647 | | University of Arizona FAX: (520)670-5740 | | Tucson, AZ 85721 | -------------------------------------------------------------------- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/