Increase no. of injection threads to 2

Test

mentioned in issue #1 (closed)

The strange thing about this is that we still have some ranks that are slow (some do indeed go faster). From looking at the output logs it seems that when you have a rank that injects a lot of requests quickly, it waits the longest. This is the only correlator I can see. Note running with -f, so no delays.

Will merge !2 (closed) and see if an artificial delay is helpful using -f.

added 1 commit

• 0406525b - Add option to delay a fixed amount of time between injections

Compare with previous version

added 1 commit

• cc5f2362 - Report deadtime when using a fixed delay, note how bad nanosleep is, must be a…

Compare with previous version

Seems nanosleep() has a fixed floor at around 50us on a COSMA node. That's not helpful.

added 1 commit

• 9cf31d24 - Add in a nanosleep() to the delay loop timer and some fudge...

Compare with previous version

added 5 commits

• 9cf31d24...7c8a108a - 4 commits from branch master
• 0eb80acb - Merge remote-tracking branch 'origin/master' into multi-injectors

Compare with previous version

added 2 commits

• 1e8f7d0e - 1 commit from branch master
• 521d15df - Merge remote-tracking branch 'origin/master' into multi-injectors

Compare with previous version

added 3 commits

• 521d15df...a19e0095 - 2 commits from branch master
• e89ae239 - Merge remote-tracking branch 'origin/master' into multi-injectors

Compare with previous version

closed

mentioned in merge request !3