This page summarizes progress of the 3rd  HLT review requested by Nu in July of 2010.

A.1) The charge

A.2) Previous documents:

• LOI , February of 2009
• 1st review report, April 2009
• Revised HLT proposal , October 2009
• 2nd review report, February 2010
• HLT response to 2nd review, August, 2010

A.3) Responses of HLT to this review:

• response 1  (August 30, 2010)
• response 2 ( September 2, 2010) 
• J/Psi & v2 from run 2010, AuAu200 ( Sept 3, 2010)

A.4) Reports from this review:

• preliminary report (September 9, 2010)
• full report (November 22, 2010)

 A.5) Further reports from HLT

• Status of HLT prior to run 11 (Oct 1 2010)

Lose notes from the review process itself, view t as my private notes.

B) Recommendations of 2nd review (order not prioritized) 

Many of the other points raised in this report can be better evaluated based on the experiences gained operating the prototype HLT in the FY10 run. The committee thus recommends to prepare a timely report on the results of the test run. This report should at least address the following points:

– performance measurements, impact on regular DAQ and network

– efficiency of the HLT during regular data taking

– effort (both time and manpower) to achieve sufficiently good calibrations

– resources needed for algorithm development and associated offline analysis

– re-evaluation of resources needed for construction and operation of the SL3 tracking part

Additionally:

1.  The main physics impact of the HLT will thus be the improvement of the fast-stream selection capability

2.  enhancement factor ...on the planned storage and computing capabilities of STAR... will probably be low

3. focus on one or two algorithms to demonstrate the sensitivity on calibrations and the computing speed requirements.

4.  The project can be roughly split into two phases:

   • the first one focussing on TPC tracking (provided by the SL3 and few GL3 machines),

   • the second one focussing on integrating further (tracking) detectors and more computing power for more demanding physics algorithms (the full GL3 computer part)

5.  Calibration will be the limiting factor for which fraction of a run the HLT can be used

6. It remains completely unexplored what level of TPC calibration is sufficient to calibrate an HFT system together with the TPC

7.  A total of 3.25 FTE, including one to be still hired ...  is on the low side

8.  The cost estimates of the proposal are on the qualitative level and will need to be extended

9.  discussion of the integration of the HLT with the DAQ system should be extended

10.  the committee recommends to study an alternative system with separate readout and tracking PCs

11. recommends to study the adoption of OpenCL instead of CUDA.

C) Concerns, issues raised recently:

1. off-line test bed for HLT algo development, testing
2. interface to run control
3. archiving setup, thresholds in DB
4. purpose of HLT: tagging or veto?
5. implication for scaler system if HLT vetoes

1. TPC calib stability: drift velocity, beam offset, space charge
2. pileup rejections for VF:  BTOW, ETOW, BTOF
3. offline trigger emulator
4. CVS : common or different code w/ offline trig emul
5. test bed to run full HLT code on M-C events
6. Preference to use OpenCL and drop CUDA

 Manuel's page with summary of 4 of us