Read pcap files with nanosec precision, set nano_sleep as default sle… #3
Conversation
plangarbalint
force-pushed
the
read_pcap_with_nanosec
branch
2 times, most recently
from
e5d3ebe to
52a6830
Compare
src/sleep.h
Outdated
| break; | ||
| } | ||
| get_current_time(now); | ||
| } |
There was a problem hiding this comment.
I'm a bit concerned about this one. Converting nanosleep to a busy loop might consume many resources. Am I miss something?
There was a problem hiding this comment.
I thought it is a good idea to use absolute sleep instead of relative sleep when an application needs to be suspended and activated in a periodic time, precise periodic activation cannot be achieved with a relative sleep() or nanosleep() function call. --> https://linux.die.net/man/3/clock_nanosleep
Do you think this solution is less efficient in ensuring precise periodic suspension and activation?
In case of nanosleep we do the following.
- calculate now
- calculate nap time (from pcap file)
- nanosleep(nap)
When CPU time is taken away from the process right after (2) before (3) then the application will not be activated exactly at now+nap. It will be activated at now+context_switch+nap
With busy loop it is easier to eliminate the effect of context switch since the time until the program should do nothing useful is an absolute time. What is your view on this?
plangarbalint
force-pushed
the
read_pcap_with_nanosec
branch
from
52a6830 to
80885b0
Compare
plangarbalint
force-pushed
the
read_pcap_with_nanosec
branch
from
80885b0 to
00e9f83
Compare
No description provided.