present: Roland, Enrico, Milo, Dave, David, Karol, Frank
Firmware organisational matters (targeted to co-processor option rather than to FELIX+CPU option):
- brief "historical" overview of interaction with the FELIX developers' team thus far:
- initially, the PD-SP FELIX effort was merely a "user" (but from the CERN/Nikhef side the fairly direct connection was felt to be a great advantage)
- the existing FELIX implementation did not provide the necessary throughput, so substantial contributions were made (especially by Enrico and Roland) to adapt it to PD-SP needs:
- notably, "jumbo blocks" + "super chunks", resulting in fewer (but larger) DMA transfers; also multi-descriptor DMA
- these improvements are also of interest to the ATLAS Phase-I upgrade and will be back-ported (with the exception of the multi-descriptor mode, which appeared less useful in the end)
- hope for future updates is to stay as close to "main" ATLAS development as reasonably possible
- from the UK side, the hope is that for the communication with the FMC it will suffice to add a "thin layer" on top of existing firmware functionality
- agree that it will be necessary to be in direct contact with the (ATLAS) FELIX developers' team, and are willing to contribute to it
- even if there is only a small overlap between UK-ATLAS and UK-DUNE person-power
- FF will talk to Nikhef FELIX developers to see if they are willing to host a small workshop
More technical aspects on FELIX+co-processor option:
- data rates & protocol for transmission between FELIX card and FMC decided? No, but low-level items (clocking etc.) have been fleshed out already
- aim to be able to use a cheaper alternative than the full "demonstrator" card. David has 4 options for a board featuring an FMC+ connector along with (2?) QSFP+ transceivers
Other: FELIX+CPU plans
- Thijs will make a start with trigger candidate finding
- Roland is sorting out the issue with backpressure noticed by Phil
- this comes in bursts, expectation is that this is due to the re-ordering stage
- also an increase of the PD-SP time window to 1 s (from the present 7.5 ms), for noise studies, is nearly done
AOB:
- there seems to be a mild preference for using the FNAL Indico server / Zoom for meetings, rather than the CERN Indico server / Vidyo
- at any rate, use of the Consortium mailing list is (for now) encouraged
- next meeting: expect to hold this in two or three weeks' time (but depending on progress)
There are minutes attached to this event.
Show them.