Results 1 to 2 of 2
Back Ground -------------- I have a program that processes 200 ms of data in 2 seconds. So I am 10 times slower than realtime. If I was to run multiple ...
Enjoy an ad free experience by logging in. Not a member yet? Register.
- 09-28-2011 #1
- Join Date
- Aug 2011
When will performance be affected?
I have a program that processes 200 ms of data in 2 seconds. So I am 10 times slower than realtime. If I was to run multiple copy of the same program and all was perfect in the world after 10 copies I would be able to run in realtime.
But that is not what happens, the best I can get to is 50% of realtime.
I set the CPU affinity for each instance of the program and that did not seem to help either.
My system has 2 6-core Hyper-threaded CPUs with 24GB of RAM. In /proc/cpuinfo, it shows up as 24 CPUs. I am running RedHat 5, kernel 2.6.18-195.el5PAE with SMP.
I have run Oprofile and Vtune and am not to sure what to make of the output.
I have not tried SCHED_RR or SCHED_FIFO. I hopping to not have to do that.
Any insight would appreciated.
What kind of performance should one except in this situation?
How can I tell what is going on?
Any additional information on interpreting Oprofile or Vtune?
Is this just a stupid test and why?
- 09-28-2011 #2
Hyperthreading provides a hardware route to splitting a processor into two virtual cores - it doesn't provide real CPU cores. If you're maxing out your physical cores, then with hyperthreading I'd expect load to show as you're seeing it. You could try turning it off and do some more testing (but don't forget to turn it back on when you're done).Linux user #126863 - see http://linuxcounter.net/