Committing TS Firmware Hit Reconstruction Stagger for the Purpose of Triggering Studies #1473
Conversation
|
@rodwyer100 let me rebase this for you, you have all these commits from the past, I'll push in a second |
|
I think the NumericalRecHitProducer may be introducing errors that prevent the build test from working. I am trying to remove it for another commit; its the other stuff I really only need commited on the short term. |
so I tried this, but it leads to way to many conflicts, so I dont wanna touch too much of the code. This is what I did this will then give you options to pick if you want the new or old |
I think we should introduce a DQM analyzer that plots these variables, and then we can compare things like what you mention above more easily (I guess this can be done in another PR) |
rodwyer100
force-pushed
the
TSfirm
branch
4 times, most recently
from
51aa024
to
a4fa598
Compare
|
Okay so I am trying to pass the build test with warnings interpreted as errors. It seems that there are many warnings unrelated to what I am doing in code thats old. Lmk if you want me to try to fix it, but seeing as its in the HLS primitives used in Trigger I doubt Id be able to. |
|
@rodwyer100 See the changes here: #1470 We can choose to ignore the warnings from HLS but there are several other parts of TS code that can be patched by us. You can also quicken development by testing the warnings-as-errors compilation locally with before building to update your configuration to treat warnings as errors. |
|
Alright everythings been rebased, checked that its building, and fixed w.r.t the changes yall made on the last push |
In my last pull request, I added the validated processor which stagged analysis digi objects to be inserted into our most current version of cluster reconstruction. In this pull request, I am adding the validate processor which inserts the most currect validated version of hit reconstruction into the analysis chain so that Elizabeth may play with it. It is a little different from the S30XL implementation primarily in that its input FIFO pragmas mean that it would need to run around 5 clock cycles as written (though LK said we would likely integrate over 3) and would need 5 copies to run continuously; the S30XL runs continuously and if we were to do that here the pragma structure would need to be altered a little bit.
This is all unnecessary detail, just note that this is the first validated piece of hit making firmware designed for the full LDMX environment that emulates LK's original manner of reconstruction. This is done in TrigScintFirmwareHitProducer, which takes in QIEDigis and outputs TrigScintHit objects. The processor will ultimately be adopted to emulate the NumericalHitProcessor of Andrew and Niramay as that seems the easiest to reproduce in something that can run in firmware. I am including it here because it is the "official" processor I will use to emulate in firmware. If it needs some refashioning or being made pretty, please lmk. I can also include it in a latter pull request alongside the firmware pileUp processor if necessary.
Check List
Here are proofs that the firmware emulates the software hits (from QIE adcs and tdcs) faithfully. Note that our ap_int resolution here means that if you fiddle with the gains (i.e. make them bigger) you will note get the right hit PE amplitudes:
SAMPLE OUTPUT:
Analysis barID: 2, PE Number: 59.4437
Analysis barID: 3, PE Number: 97.9569
Analysis barID: 22, PE Number: 79.7444
Analysis barID: 23, PE Number: 89.955
Analysis barID: 35, PE Number: 158.878
Analysis barID: 36, PE Number: 61.3231
DID I GET HERE 3
DID I GET HERE 4
DID I GET HERE 5
Firmware barID: 2, PE Number: 59
Firmware barID: 3, PE Number: 98
Firmware barID: 22, PE Number: 79
Firmware barID: 23, PE Number: 90
Firmware barID: 35, PE Number: 159
Firmware barID: 36, PE Number: 61
Firmware barID: 49, PE Number: 8
DID I GET HERE 6
Note that the reason there is an extra hit printed in the Firmware in bar 49 in this case is because I didn't print analysis hits with amplitude below 30. I can fix that, but I have seen enough of the output to know that the firmware is replicating the software appropriattely.