iiinova — where your digital self keeps evolving

getting some cool visuals going… github.com/nyu-parcour/...

020

and, check out some of the other cool stuff we're up at NYU here! nyu-parcour.github.io

nyu-parcour.github.io

ParCour

000

Scheduler augmentation is easily deactivated when not in use, requires no special compiler support, and can be incorporated into lightweight "library-only" schedulers with essentially no performance impact or maintenance burden.

100

Critically, none of these analyses incur major overhead: just a few percent in typical applications and use-cases, and fully parallelized, with almost no impact on scalability.

100

In another case study, we measure the impact of a subtle space optimization. With just a few lines of code, we can empirically confirm that the optimization works as intended, by measuring peak memory consumption as an abstract property on the computation graph.

100

For example, using just a few lines of code we can identify a granularity control issue inside of a library we did not author ourselves. We describe a case study in detail in the paper.

100

Now we can easily peek inside a computation, even deeply across library boundaries. This is great for finding performance issues buried in places you did not expect.

100

Customization is easy: just define a notion of vertex-local data and functions to update this data across forks and joins. These custom vertices are handed off to an “augmented” scheduler, which incrementally maintains the computation graph under the hood during execution

100

This paper is all about making it easy to dynamically observe the computation graph of a parallel execution (for profiling purposes, dynamic analyses, etc)

100

ACM Symposium on Parallelism in Algorithms and Architectures – Royal Holloway, University of London, in London, UK, July 6-10, 2026

Joint work between myself and two of my students at NYU: Darshan Dinesh Kumar and Seong-Heon Jung come check out our presentation at SPAA this Thursday: spaa.acm.org/wp-content/u... (live streams will be available, I believe?)

spaa.acm.org

100

Sam Westrick @shwestrick.bsky.social · 24/06/2026

delighted to announce that our SPAA paper received an outstanding paper award! Scheduler Augmentation: A Lightweight, Low-Cost Profiling Technique for Fork-Join Parallel Programs cs.nyu.edu/~shw8119/26/...

cs.nyu.edu

130

Reposted by Sam Westrick

We've extended the deadline for the ML Family Workshop 2026! The new deadline is ** July 1, AoE ** The submission requirements are fairly minimal: only 2 pages. Please consider submitting, and encourage your friends to do so too!

103

Sam Westrick @shwestrick.bsky.social · 24/06/2026

103

Sam Westrick @shwestrick.bsky.social · 16/06/2026

the ML Family Workshop (@ ICFP'26) submission deadline is approaching -- **Wednesday June 24** Please consider submitting! We especially welcome works-in-progress. Submission requirements are minimal -- just 2 pages. icfp26.sigplan.org/home/mlfamil...

icfp26.sigplan.org

ML Family Workshop 2026 - ICFP 2026

ML is a large family of programming languages that includes Standard ML, OCaml, F#, CakeML, SML#, Manticore, MetaOCaml, JoCaml, Alice ML, Dependent ML, Flow Caml, Reason ML, Flix, MaPLe, and many othe...

1110

Sam Westrick @shwestrick.bsky.social · 05/06/2026

it helped me catch a nasty whitespace change that latexdiff totally missed

010

Sam Westrick @shwestrick.bsky.social · 05/12/2025

my student Seong-Heon talking about one of our new projects at NYU! (We’re at NJPLS today, come say hi)

1102

Sam Westrick @shwestrick.bsky.social · 04/12/2025

Enjoying this presentation about performance engineering in the Go GC implementation: youtu.be/gPJkM95KpKo Really fantastic talk!

youtu.be

GopherCon 2025 - Advancing Go Garbage Collection with Green Tea - Michael Knyszek

YouTube video by Gopher Academy

030

check out the paper for more details! cs.nyu.edu/~shw8119/25/...

cs.nyu.edu

000

Additionally, OAC outperforms existing quantum circuit optimizers, often producing a better quality final circuit in orders of magnitude less time.

100

In our experiments, we confirm that OAC performs strictly better than the naive chunked strategy, sometimes significantly so, removing thousands more gates from large circuit instances. This confirms that "melding" is important for circuit quality.

100

We call our algorithm OAC, for "optimize and compact". We prove that this algorithm only calls the oracle productively: the number of additional calls to the oracle (due to melding, etc) is bounded by the number of optimizations found.

100

Melding can in turn expose even more opportunities for optimizations of individual chunks. So, after melding, we can continue optimizing recursively until convergence on a circuit which is (in some sense) "locally optimal" relative to the size-constrained oracle.

100

To address this, we develop a "melding" algorithm which identifies and performs additional optimizations across the boundaries.

100

But this immediately raises a question. Surely, this approach must miss optimizations that are possible across the boundaries between pieces, right? (As you might expect, the answer is yes.)

100

thinking of these tools as black box optimizers or "oracles", a natural strategy would be to chunk up the circuit into small pieces and apply the oracle to each piece. This easily guarantees that the exponential searches are bounded and do not explode (in time/space)

100

the key challenge here is optimizing a large quantum circuit (with, e.g., hundreds of thousands of gates) many existing tools are essentially superoptimizers, with exponential search spaces, and therefore can only handle "small" circuits in a reasonable amount of time/space

100

Sam Westrick @shwestrick.bsky.social · 23/10/2025

happy to announce that, earlier this Fall, our QCE'25 paper "Local Optimization of Quantum Circuits" received a Best Paper award! cs.nyu.edu/~shw8119/25/...

cs.nyu.edu

150

These days, you can buy such a machine for ~$3K. That’s a lot of compute per buck. Medium-scale parallelism is quite affordable (“Medium scale” meaning large-ish single-node.)

000

Sam Westrick @shwestrick.bsky.social · 23/10/2025

I believe the idea (in ~2015) was: if you’re considering purchasing a car, consider spending your money on this 72-core 1TB multicore machine instead. They’re a similar price!

100

Sam Westrick @shwestrick.bsky.social · 23/10/2025

for even more context: Laxman has been exploring the limits of what is possible on single-node parallel machines. This talk is about a recent SPAA paper. They’re getting great scalability up to 8TB on a single node

100

Sam Westrick @shwestrick.bsky.social · 22/10/2025

for context, this slide is ~10 years old

100

Sam Westrick @shwestrick.bsky.social · 22/10/2025

the famous Nissan Leaf slide returns! Laxman Dhulipala emphasizes the benefits of single-node, shared memory parallelism

140

Sam Westrick @shwestrick.bsky.social · 18/10/2025

one day, my apartment will look like this

1212

Reposted by Sam Westrick

Anil Madhavapeddy @anil.recoil.org · 17/10/2025

Extremely cool work on a mechanically verified garbage collector for OCaml using F* by Sheera Samsu, @kcsrk.info and colleagues at the OCaml Workshop #icfpsplash25

0257

Martin Elsman wraps up the day with Compositional Deep Argument Flattening, a method for summarizing flattening transformations to enable optimizations across compilation unit boundaries

000

Yanni Lefki (together with Arthur Charguéraud) develop “Binding Boolean Expressions”, a core calculus for generalized pattern matching, guards, and case statements

021

Samuel Vivien presents his recent work (together with Didier Rémy) on Implicit Modules, an important step towards the long-standing Modular Implicits proposal for OCaml: a way to get the compiler to insert the function you need, automatically

081

When putting together the program, I noticed that there hadn’t been much CakeML at the ML Family Workshop, at least in the past decade. Yong Kiam’s talk was excellent! 👏 I am endlessly impressed with CakeML — such an exciting project

000

Yong Kiam overviews the design of CakeML and all of the successes of the project over the years!

100