Board level simulation

• Specifications

• Simulators are instantiable

   myhf = hfsim_new()
   myams = amssim_new(wedgenum)
  

• makes it easy to write a small test that simulates one or two boards

• provides a natural place to collect instance-specific info, instead of e.g. arrays of global variables

• Input/output formats correspond to spy buffer / "cable" data (but are split into distinct events)

• Alex is working on code to interpret/decode bits on cables (software equivalent of Stefano's web pages)

• Don't copy data around; pass pointers instead, with agreements about who owns the data

• Allow both "compact" and "memory image" representations of configuration data

• No question we need to keep memory use to a minimum on crate CPUs

• Often a simple function can replace a huge lookup table

• Giovanni is working out the plan for this

• Want to make sure boards and simulators can be initialized from same ASCII file formats

• We have some thoughts on memory allocation, which we won't bore you with

• Want to use efficient algorithms, e.g. precompute lists of roads that contain a given superstrip

• Eventually want a non-hard-coded way to indicate which boards exist and are connected to which other boards

• To be decided

• Do we configure board simulators using many method calls (analogous to board code)?

   amssim_loadPattern(myams, amboardnum, pattdata)
   hfsim_setStripQCut(myhf, stream, chip, counts)

• ... or does e.g. each wedge have a monster data structure pointing to all relevant configuration data?

   myams[wedgenum] = amssim_new(wedgenum, &configdata[wedgenum])
   myhb[wedgenum] = hbsim_new(&configdata[wedgenum])

• Do we explicitly shuffle data between simulators?

   while (1) { 
   . find(svx_raw_data); 
   . hfsim_inputRawData(myhf, svx_raw_data); 
   . hfsim_procEvent(myhf); 
   . hitnout = hfsim_getHits(myhf, hitout, maxhitout); 
   . amssim_inputHits(myhf, hitout, hitnout); 
   . amssim_procEvent(myhf); 
   . amsnout = amssim_getRoads(myams, amsout, maxamsout); 
   . hbsim_inputRoads(myhb, amsout, amsnout); 
   . hbsim_inputHits(myhb, hitout, hitnout); 
   . hbsim_procEvent(myhb); 
   . hbnout = hbsim_getOutput(myhb, hbout, maxhbout); 
   . doSomethingWith(hbout, hbnout); 
   } 

• ... or do we tell each board once when it is created where to look for its input?

   myhf = hf_new(&svx_raw_data); 
   myams = amssim_new(&myhf->output); 
   myhb = hbsim_new(&myhf->output, &myams->output); 
   while (1) { 
   . find(svx_raw_data); 
   . hfsim_procEvent(myhf); 
   . amssim_procEvent(myams); 
   . hbsim_procEvent(myhb); 
   . doSomethingWith(myhb->output); 
   } 

• Do we want to provide more than just board output? HF, for example, could make cluster-length and cluster-charge data available for histogramming

• Status

• Less progress than expected since January

• One or more bit-level simulators exist for HF, AMS/B, HB, TF, but all will need some work on interfacing, once decisions are made

• Now have 36 HF simulators finding clusters in SIXD written by Run2 GEANT simulation (but SvxSim puts stereo data before axial, uses wrong chip ID convention ... oops!); sending them to 12 AMS/B and 12 HB simulators in a crude, ad-hoc AC++ module

• Need to include XFTSim module's output

• TF simulator now looks much more like the others

• We said that by now we would have spelled out exactly how we wanted to interface to the outside world; we haven't done it yet

• We also said that by now we would have a prototype actually finding and fitting GEANT-simulated tracks; we're not there yet