• Current status

• Giovanni generates RAM image files from his high-level formats

• Coldstart ensures that these RAM images are loaded into the boards at start of each run

• Optimization using compression, and skip of unneeded writes based on checksum comparison (VME i/o is much cheaper than network i/o)

• Choice of memory images for each board determined by hardware database "metakey" parameter

• Currently no elegant way to select among different sets of constants (e.g. 4/5, 4*, which layer out for 4/4, different beamspot, etc.)

• Big memory images archived in CVS

• Full implementation for all boards used in commissioning run; 95% implementation for TF; missing for XTF

• Desiderata

• Need way to specify (trigger DB, R_C GUI, etc.) different sets of constants

• Want to eliminate storage of big RAM files, since they are essentially object code, for which we have the source code

• Want to generate RAM images on crate CPU

• Need to have full implementation for all boards; also need some minor clean-up of existing coldstart routines (e.g. don't hard-code memory addresses)

• Think about other degrees of freedom, such as offset beam spot, "broad" vs "narrow" beamspot, hot channel suppression

• Read as much as possible about channel-by-channel silicon issues from silicon DB, not our own; Petar assures me that we can predict hot channels from his pedestal measurements; we should check

• Continue to use checksum, as a means of ensuring that on-crate algorithm gives correct RAM image, can be reproduced years later on different computer, inside trigsim, etc.

• Make sure there is an easy way to get trigsim to know what we did online; this is equivalent to the problem of making sure we can reproduce trigger results offline

• To do items

• Reproduce CR state (never as easy as it sounds)

• Make sure TF is configured

• Get my hands on Giovanni, Guram, Vladimir programs that expand high-level formats into RAM files

• Make these able to run on crate CPU, do coldstart from high-level data instead of low-level data

• Work out database issues so that we can manage different sets of constants

• Etc

• I would like to find a grad student who is (a) living full-time at Fermilab and (b) interested in a Run 2 B physics thesis topic, who would be willing to work with me in January and/or February; this person may then continue to be involved in the SVT group, if it works out

• This talk is very sketchy; I'll try to fill in more details for next week, since I'm talking then about svtsim as well, and there are many interconnecting issues