Improve timers in standalone bridge mode?? (Low priority)

[valassi]

This is a followup to #461.

For cpp and cuda we have at least three sets of throughputs that we can measure

• through madevent, via fbridgesequence: this is what the experiments will see, and we should keep this
• through the standalone check/gcheck, in "bridge" mode: this should mimic the previous one, but it does not (see Understand performance difference in cudacpp between SA and madevent #461)
• through the standalone check/gcheck, in no-bidge mode: this is essentially our fastest benchmark, and we should keep that too

The question is, should the timers in SA bridge mode be improved so that they reproduce madevent? This seems unnecessary, because now we do have madevent. The bridge mode was only a way to test the functionality of the bridge before its integration in madevent. It has served its purpose, it needs no more work.

Note that for complex processes like ggttggg, all three are identical in cpp, and are within +-10% in cuda. So this is really low priority.

Anyway, I am filing this for bookkeeping...

[roiser]

I agree, let's not spend time on the SA bridge mode.

[valassi]

Low priority, but I would keep this open for now