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
Since #97 the device produces a timestamp that is stored in the optional output file.
However the host does not use this in the power calculations. It still use the wall time as measured on the host whenever data arrives from the device.
While the device timestamp is more accurate, it has a potential issue: it rolls over every 1024 us. If the link with the host
drops for longer than that, the calculated total time would be inaccurate. This could cause a desync that gets worse over time.
It this high risk enough that we care? Or would it be ok to use the device timestamps in the host power calculations?
In the latter case we should manually keep track of the total time in a variable wider than the effective 10 bits of the device timestamp.
The text was updated successfully, but these errors were encountered:
Since #97 the device produces a timestamp that is stored in the optional output file.
However the host does not use this in the power calculations. It still use the wall time as measured on the host whenever data arrives from the device.
While the device timestamp is more accurate, it has a potential issue: it rolls over every 1024 us. If the link with the host
drops for longer than that, the calculated total time would be inaccurate. This could cause a desync that gets worse over time.
It this high risk enough that we care? Or would it be ok to use the device timestamps in the host power calculations?
In the latter case we should manually keep track of the total time in a variable wider than the effective 10 bits of the device timestamp.
The text was updated successfully, but these errors were encountered: