VEGAS Team Meeting
November 04, 2013
CALTECH, UC-B, & Columbia
Glenn, Dan, Mark
Marty, Mark, Joe, Ray, Jason, Randy, Anish, Amanda
- Old Business
- 11/11 Leave
- The VEGAS mode documentation, which is a Google Doc, available at https://docs.google.com/document/d/1C_it02j8yqu_VZcYnN6aVC-jrgP4ClAxbj8KP4cH4EQ/edit?usp=sharing.
- Needs more detail (block diagram & descriptions)
- Ranges on parameters, nominals values
- Values are different for different bof files
- Needs to have somewhat variable parameters set by observers (gains)
- GPU code documentation: This is a low-level documentation of the GPU spectrometer code, and is a CASPER wiki page, available at https://casper.berkeley.edu/wiki/VEGAS_GPU_Code_Documentation. This page is linked to from the main VEGAS page on the CASPER wiki.
- No comments at this time
- Check for information for Config Tool (link on VEGAS web site)
- The packet format page on the CASPER wiki has been updated to reflect the current LBW packet format: https://casper.berkeley.edu/wiki/VEGAS_Spead_Packet
- Just shows header for LBW modes - other questions need addressed
- Counter questions
- Time samples is not covered in linked document (data order in packet)
- look in .pdf document drafted by Simon
- Glenn will add LBW modes, and registers (some desc of purpose & use ing FPGA document already exists)
- Plan to move to GB all documents after next round of creations and edits
We want to do final, final tests of modes 1,2 and 3, to get these out of the way, and so we can send fixed versions to SHAO. Remember the issue was the reoccurrence of the 1/8 band problem, which Mark fixed earlier this week. But we also want to rebuild the BOF files with a rationalized build environment which Glenn and Mark are working on.
- Mark W rebuit them and added them to git
- Needs to be retested by GB (Amanda)
Maintenance days next week will be Tuesday and Thursday. Richard signed Amanda up for two hours on Tuesday, and Anish for eight hours on Thursday.
Slightly less happy path:
- [partial] Glenn and Mark sort out the build environment on or before Monday, and recompile the BOFs
- Not from rational build environment - probably requires a re-test
- Ray installs the BOF files on Monday.
- . Amanda uses two hours to do final engineering checkout of modes 1,2 and 3 (i.e. look at X-band passband using the full IF system, but with telescope parked). I think this will be the first real use of VEGAS for some time, so she will probably need Ray to check Managers are running, etc.
- Assuming engineering tests pass, Amanda requests two hours of real observing time from Toney to do final on-sky tests (looking at RRL).
- Anish and Glenn use Thursday for stand-alone Mode 20-29 testing (this would be ahead of schedule!)
- Glen , Mark and Ray have until Wednesday to build and install the new BOFs.
- Amanda does final engineering checkout of modes 1,2 and 3 on Thursday.
- Amanda requests real observing time after Thursday
- Glenn and Anish test Modes 20-29 the following week (starting 11th)
1. Rationalize build environment for BOFs.
Progress on this? When do we expect it to be completed? Still having issues with the environment
From John Ford, Friday 1st: The arp spamming fix is a Manager thing. [i.e. Ray Creager} It involves reading the config file and setting the arp table in all of the ROACHs.
Progress on this? Hard code addresses
When do we expect it to be completed? This week
4. Fix task pinning to cores.
From Joe Brandt, Wed 30th: The thread pinning code similar to what I added to guppi is now available in both the gbt and DIBAS branches of our /home/gbt2 repo. Example config file (vegas_threads.conf) is included in the source.
Progress on this? Code done; need to test to see what cores to pin too
When do we expect it to be implemented in the system? Values need to be set in config once pinning determinations have been completed
5. Install CUDA version 5.5 in VEGAS:
From me, Nov 1st: Task 5 (install CUDA 5.5) is assigned to Wolfgang. Does this need to be coordinated with HPC work (e.g. do you need to rebuild everything?) Joe, could you ask Wolfgang to proceed with that, when it is convenient for you?
Progress on this? When do we expect it to be implemented in the system? completed
6. Fixing data drop at the end of scans.
Choice to be made - data comes through system in blocks. Should we use DIBAS solution (one extra block) or round down? - Uncertain how other systems react downstream so Bob recomends confroming data. Rounding up (DIBAS) solution is prefered.
Has it been implemented yet? No
- Code and build repository at Green Bank See above
Next Teleconference: November 19, 2013
- NRAO will be off November 11.
- Meeting will be on Tuesday next week