L1 & L15 Verification comments

• D Owen 5/22/96

Compare Program

1) Required substantial effort

• need for "bit for bit" simulation requires effort beyond that needed for trigger simulation studies
• setup and maintenance of on-line monitoring non-trivial
• monitoring requires someone to spend a measurable fraction of their time DAILY. (more effort in setup could help minimize daily effort)

2) Few problems detected

• only of order 2 hardware problems were first detected by the on-line monitoring. (one "logic bug" would have been seen in by the L1.5 Compare if the latest modifications in hardware had been implemented in the simulator on time)

3) Nevertheless, Compare is essential

• Provides earliest detection of trigger hardware problems

4) Improvements (or what we did wrong)

• L1.5 hardware and simulator design differences lead to more difficult simulator code changes when hardware upgraded.
• implement hardware changes in simulator code EARLY
• better communication
  • Monitor person needs to be notified of new releases of the Trigger List and the Simulator and On-Line code.
• more automated implementations
  • new trigger releases required too much unnecessary work
  • Compare output could be improved
    • distinguish "known features" from real discrepancies
    • minimize number of "known features"

Other On-Line Trigger Monitoring

1) GM Trigger Examine

• caught "logic error" in a L1.5 algorithm
• sensitive to trigger list mistakes as well as hardware problems

2) Trigger List checks:

• TCC, COOR, & Trig Parse each applied useful checks to trigger lists.

3) DAQ and GM Shifters:

• monitored trigger rates and cross sections

4) Suggestions

• Perhaps L1 & L1.5 data needed on DST's should have been put there by RECO rather than L2. Then some L1.5 changes would not have required new releases of the on-line code