HELP: nanosleep causes huge amount of local timer interrupts

neyshule · 10-08-2011, 10:49 PM

Hi all:
I have a user mode application which uses nanosleep in a while(1) loop, unfortunately it brings a huge amount of local time interrupts, which greatly damage the application performance, I don't know why nanosleep can cause this large increase on local timer interrupts?? How should I do to eliminate this unwanted interrupts. Thanks for your help.

corp769 · 10-08-2011, 11:16 PM

Hello,

Could we see the code, if you don't mind?

Cheers,

Josh

neyshule · 10-09-2011, 12:09 AM

The code is long， here is the main framework：
while（1）{
check some flag；
if true，nanosleep；
use memcpy to copy from one memory area to anthoer；
}

If I don't use nanosleep, everything works fine without huge LTI, but the CPU usage is 100%.
The weird thing is this code works well in the old environment, after I reformat the disk and reinstall the linux the behavior changes.(same machine and kernel), what could be the difference, some configuration???

dwhitney67 · 10-09-2011, 06:11 AM

Quote:

Originally Posted by neyshule

The code is long

Then why not develop a small application that attempts to duplicate the issue? Perhaps all that is wrong is your usage of nanosleep?

ta0kira · 10-09-2011, 10:30 PM

Why not use a pthread condition instead of a flag and nanosleep? If you still need the flag, you should be able to broadcast the condition when the flag gets updated. nanosleep isn't really a good way to synchronize threads and processes. If necessary, you can share a condition between different processes, also.
Kevin Barry