(Done) March 12th Meetup: Decompilation and clang plugins

[frenchy64]

👍 to RSVP

https://madclj.com/

Location

Wednesday, March 12th, 2025
6:30 PM to 9:00 PM CT

StartingBlock Madison
821 E Washington Ave 2nd floor · Madison, WI

Schedule

Sayan will give a presentation on a clang plugin she is designing.

Ambrose will demonstrate how to to decompile Clojure to inspect the equivalent Java code. We will use this to show how let{fn} is implemented in Clojure, then compare this to jank's actual LLVM IR output.

6:30pm arrive, food & drink provided
7pm start meetup
9pm end meetup

[brookeswanson]

I shall be there!

[frenchy64]

Any ideas for topics? Would people be up for taking turns to plug into the projector and show a thing or two they've been working on?

[sivakusayan]

Maybe we could do a papers we love type thing?

If that doesn't work, I could do a rehash of a presentation I gave at Singlewire at the end of January, although James Elliot would have seen it already 🙂 it was a ~50 minute presentation that fleshes out some mathematical ideas behind the clang plugin I want to write, although ~15 minutes of that was computer architecture background people here are probably more familiar with.

Otherwise, I've been trying to figure out how lock-free programming works recently, and assuming I can learn the (very basics) of it in a month and try to do some implementation, I could prepare a presentation for what I learned so far in May. Although, I'm moving to Milwaukee soon, so not sure if I'll still be here in May.

[ericnormand]

I'll be there with the girls!

[sivakusayan]

Did we decide on a schedule for tomorrow? Or were we thinking of skipping this month?

[frenchy64]

Still on, but haven't decided what to do exactly yet.

[frenchy64]

@sivakusayan could you do a 30 minute version of your clang talk?

[sivakusayan]

Sounds good :)

[frenchy64]

@madclj/meetup tomorrow's meetup will feature two talks:

Sayan will give a presentation on a clang plugin she is designing.

Ambrose will demonstrate how to to decompile Clojure to inspect the equivalent Java code. We will use this to show how let{fn} is implemented in Clojure, then compare this to jank's actual LLVM IR output.

[frenchy64]

Pizza is ordered. I should be in the lobby at 6:30pm to let people in.

[sivakusayan]

I'm downstairs!

[frenchy64]

Some relevant thoughts to Sayan's talk wrt jank:

• jank objects are structs whose first members are a uint8_t tag. there is no subclassing, instead objects are cast from the object struct---which just has a tag---to other structs that also start with a tag with a big switch statement to simulate polymorphism.
• this llvm optimization pass can automatically convert struct { T field} into T (apparently that's subtly different). It's used to optimize closures that close over 1 variable like (let [v 1] (fn [] v)). The extra argument passed to initialize the closure is just 1 instead of struct { v=1 }.