It's actuallly kind of fun doing my master's thesis over again. But maybe wait 2 years and then the 20 year research do-over rule will kick in and a future student of mine get their master's doing that too? Nah...

It took me like 12 years, but I finally documented what the compiler options do. github.com/titzer/virgil/…

When you get to needing array slices in bash, maybe you should reevaluate your life choices. `args=${@:$i:$SHARDING}`

Just one more optimizer bug to squash, I swear. Just one more.

Deep down, I worry that if my paper has a tiny mistake it looks like this to reviewers.

My colleague Albert Ziegler wrote this nice blog post summarizing our MAPS '22 paper on the productivity impact of #GitHubCopilot . 1/

I think the guitar thief of Gatwick might have permanent back problems after my legendary 2018 flying tackle.

Congratulations Steve Blackburn 🌻, looking forward to seeing what this lab produces!

Visualizing graph-based IRs like C2's is absolutely critical to work on these compilers. Nice work!

Retpoline was never a panacea for BTI and this paper now proves that conclusively. regmedia.co.uk/2022/07/12/han…

The numbers for the OOPSLA artifact size are in: 10GB zip file, of which 2.8MB is Wizard (including its entire git history!), 6MB is generated data, 57MB is Virgil (and its git history), 25MB other Wasm engines, and then enormous, enormous JavaScript engine source and binaries.

The entirety of the past 24 hours has been me wrestling with how to package up 24GB of 3 WebVMs for the OOPSLA artifact. Just so they can run a couple dozen Wasm programs in the few tens of kilobytes range.

Me: man, assembly isn't that hard. People shouldn't be afraid of machine code. Also me: man, I think we should invent something so no one ever has to write machine code again.

Excited to be (remotely) speaking on Friday at plclub @ Penn on my work on Filibuster and microservice resilience testing!

If you're at Penn, come visit and say hi!

seas.upenn.edu/~plclub/club/

Every time I find a bug in my optimizer I feel like there are cracks in the foundations everywhere and this whole house is a-comin' down any second.

Pleased to get the notification today that my paper on how to interpret #wasm fast was accepted to OOPSLA 2022! Here's the arXiv version I put up a few months back but didn't advertise because blind review. arxiv.org/abs/2205.01183

Hi, I'm Ben and my decades-long PL project is 'unpublishable' and projects using it suffer from 'an unfortunate language choice'.

Wizard generates a fast #wasm assembly interpreter at startup time. That takes 300 or so microseconds, which always bothered me. So I hacked it to puke the generated code directly into the executable to skip that step. github.com/titzer/wizard-…