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
Is your feature request related to a problem? Please describe.
Requests have been received to include software bus performance as part of the test matrix.
Describe the solution you'd like
Start by implementing a functional test case (as part of cfe_testcase) that emulates using the software bus for bulk message transfer, by sending a fixed large number of messages (e.g. 1 million). This process can be timed, and this in turn will give some indication of relative performance of the implementation.
Additional context
This obviously could be much more sophisticated. The proposed test only covers the simplest case, in single thread/single core context, and the results are only comparable to successive runs on the exact same hardware with the same compiler and same compiler flags. Nonetheless, monitoring the result before and after a change will provide at least one simple metric of whether that change affects performance or not.
A follow-on change should perform a similar test with multiple threads, and thus give some metrics to how well SB scales across cores.
Requester Info
Joseph Hickey, Vantage Systems, Inc.
The text was updated successfully, but these errors were encountered:
Add a functional test that moves a large number of messages across the
bus as fast as possible (i.e. with no delay). The process is timed, thus
offering some metric as to the single thread/single core performance of
the software bus implementation underneath.
Note that due to vast differences between hardware, results should only
be compared between successive runs on the exact same hardware, such as
before and after a change was made.
Is your feature request related to a problem? Please describe.
Requests have been received to include software bus performance as part of the test matrix.
Describe the solution you'd like
Start by implementing a functional test case (as part of cfe_testcase) that emulates using the software bus for bulk message transfer, by sending a fixed large number of messages (e.g. 1 million). This process can be timed, and this in turn will give some indication of relative performance of the implementation.
Additional context
This obviously could be much more sophisticated. The proposed test only covers the simplest case, in single thread/single core context, and the results are only comparable to successive runs on the exact same hardware with the same compiler and same compiler flags. Nonetheless, monitoring the result before and after a change will provide at least one simple metric of whether that change affects performance or not.
A follow-on change should perform a similar test with multiple threads, and thus give some metrics to how well SB scales across cores.
Requester Info
Joseph Hickey, Vantage Systems, Inc.
The text was updated successfully, but these errors were encountered: