Summer Progress Tracker – Akash

[polybeandip]

Weekly reports on how work is going over the summer

[polybeandip]

Work done between 6/3/24 and 6/7/24

• Read "Formal Abstractions for Packet Scheduling".
  • Sections 1 to 5.2 and 6.1 were carefully worked through
  • Sections 5.3 and beyond were skimmed
• Familiarized myself with handwriting Calyx code and the Calyx CI
  • wrote a matrix multiplier
  • setup runt test
• Familiarized myself with using the eDSL
  • wrote a matrix multiplier
  • side quest: implemented seq_mem_d2 and comb_mem_d2 (PR)
  • side quest: documented a bug with using invoke and reading from seq_mem (issue)
• Studied FIFO and PIFO implementations
• Began adding to the binary heap (branch)
  • tweaked the heap to accept and order values by rank

[anshumanmohan]

Really great stuff, and I hope you're having fun! I will add one more thing to your list: you found a bug in the docs. We were incorrectly using std_add in a non-combinational manner. I'll fix this now. Good luck with the fancier binary heap!

[polybeandip]

Oops, it totally slipped my mind to fix that part of the docs, apologies. Would it still be okay if I tried my hand at fixing it now? Nvm, just saw that you fixed it. Thank you!

[anshumanmohan]

Whoops sorry just seeing this now. I just opened calyxir/calyx#2122, take a look?

[polybeandip]

No worries, the PR looks great to me

[polybeandip]

Work done between 6/10/24 and 6/14/24

• More work on the binary heap (PR)
• Set up Havarti (Anshuman's related PR)
• Add error handling to width inference for cells in invoke (PR)
• Write up issue on nested if conditions
• Minor fix to name clashes with ComponentBuilder.incr (PR)

[anshumanmohan]

Way cool, and I'm really happy with the fuller-featured binary heap! Good luck on the scheduling transaction stuff :)

[polybeandip]

Work done between 6/17/24 and 7/12/24

More on the binary heap:

• Build stable_binheap and implement a general PIFO and FIFO via the heap (PR)
• Integrate stable_binheap into the randomized test suite (PR)
• Document both stable_binheap and binheap (PR)

On generalizing from PIFO to PIEO trees:

• Setup tracker (issue)
• Confuse people with my worries about metadata in Internal PIEO tree nodes (discussion thread)
• Alter pifo-tree-artifact to use PIEO trees and test the compilation process with two NWC policies (PR)

My PIFO of tasks (rank=importance):

• Write up a proof that PIEO tree compilation works: that is, our analogue of $\hat{f}$ does result in a simulation.
  Rather than copying proofs in Formal Abstractions with minor tweaks, I hope to piggyback off our results with PIFO trees.

  

  I think the key here is showing this diagram commutes*.
• After this PR is merged, spend some cycles on Queues: flag day
• Ask about waveform debugging (not needed for my work atm but it seems like a cool technique)

Apologies for the delay on this @anshumanmohan; the weekly updates are definitely coming back now!

*Here,

• $f$ is an embedding
• $\mathrm{proj}_p$ is map obtained by throwing away all packets and child references that don't satisfy predicate $p$
• $\hat{f}_{\mathrm{PIFO}}$ is our usual $\hat{f}$ from Formal Abstractions
• $\hat{f}_{\mathrm{PIEO}}$ is our analogue of $\hat{f}$ for PIEO trees

[polybeandip]

Work done between 7/15/24 and 7/19/24

• Finish formalizing PIEO tree compilation (PR); this completes this issue

[polybeandip]

Work done between 7/22/24 and 7/26/24

• Make Multiple tweaks via mini PRs:
  • Set up CI #45 (completes Set up CI #43)
  • Simplify helpers in eval.ml #47
  • Rearrange files #49
  • Factor policy functions into their own module #50
• Write up issue Agree on the workings of Round-Robin  #48
• Help with DSL: Better Error Messages #39 (still 99% Kabir's work though)
• Fix bugs in notes for Well-formedness and its preservation in a PIEO tree context #36
• Work on the blue path in Compiler paths: policy DSL to Calyx eDSL #38 on branch dsl+artifact

My PIEO of tasks, ordered by rank (=importance)

• Make issue for the blue path in Compiler paths: policy DSL to Calyx eDSL #38
  Sorry! this really should have happened before I started work on it
• Update Agree on the workings of Round-Robin  #48 with discussion from Monday's meeting (7/29)
• Benchmark queues
• Pick up some parts of flag day

[polybeandip]

Work done between 7/29/24 and 8/9/24 (last update)

• Update Round-Robin as per Agree on the workings of Round-Robin  #48 (PR)
• Port Over Simulator and Compiler from pifo-tree-artifact (PR)

So long, and thanks for all the fish!