You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When debugging the "heavy loss" tests, the execution time is different on 64 bits (a little under 23sec) and on 32 bits (a little over 23 sec). Comparing the qlog traces shows that packets are sent at slightly different times. This cause the order of transmissions and acknowledgements to differ, thus slight differences in congestion control and also in relative order of retransmissions and first transmissions. In the case of the "heavy load" test, this causes a cascading effect.
The text was updated successfully, but these errors were encountered:
This is mostly annoying for testing. It is probably due to the mix of floating point and integer computation in the pacing logic. Replacing that by fixed point would alleviate the problem.
When debugging the "heavy loss" tests, the execution time is different on 64 bits (a little under 23sec) and on 32 bits (a little over 23 sec). Comparing the qlog traces shows that packets are sent at slightly different times. This cause the order of transmissions and acknowledgements to differ, thus slight differences in congestion control and also in relative order of retransmissions and first transmissions. In the case of the "heavy load" test, this causes a cascading effect.
The text was updated successfully, but these errors were encountered: