Renaissance Grrrl

I announced this on twitter a while back, but tomorrow I'm flying out to Nara Japan. I'll be out there all week for ICFP and all that jazz. It's been about a decade since last time I was in the Kansai region, and I can't wait. As I've done in the past, if you want to meet up for lunch or dinner, just comment below (or shoot me a tweet, email, etc).

It's official, I'm heading to Google at the end of July to work with Mark Larson on the Chrome OS security team (of all things!). Seems an unlikely match, but Mark was pretty keen on my background (including the gender stuff!) and wasn't put off by my fusion of linguistics, type-theory, constructive logic/maths, et al. I guess the security team is more concerned with semantic models and constructive correctness than the other teams I talked with. Makes sense, I suppose, but it's sad that these are still thought of as "security" concerns rather than "everything in programming" concerns.

I'm totally looking forward to it, but I am still thinking of it as a bit of an experiment. I'm an academic at heart, so I could see myself heading back to academia in a few years. (Wanting to be a professor is, afterall, what motivated me to start gradschool in the first place.) Then again, I've never really thought of industry as being interested in the sorts of theory I work on. In any case, I'll still be around at the various conferences I've been frequenting; I won't just disappear into industry.

I know a bunch of y'all are in the bay area. I'd love to hear any pointers you have on apartment hunting or on what neighborhoods (nearish Mt View) are nice— i.e., places an artistic disabled radical queer woman might like. Feel free to comment below or get in touch by email, twitter, etc.

In the spirit of Brent's post, I figure I'll make a public announcement that I'm in Vancouver all week attending HOPE, ICFP, and the Haskell Symposium. I love reconnecting with old friends, as well as meeting new folks. Even if I've met you at past ICFPs, feel free to re-introduce yourself as ...things have changed over the past few years. I know I've been pretty quiet of late on Haskell cafe as well as here, but word on the street is folks still recognize my name around places. So if you want to meet up, leave a comment with when/where to find you, or just look for the tall gal with the blue streak in her hair.

Unlike Brent, and unlike in years past, I might should note that I am looking to "advance my career". As of this fall, I am officially on the market for research and professorship positions. So if you're interested in having a linguistically-minded constructive mathematician help work on your problems, come say hi. For those not attending ICFP, you can check out my professional site; I need to fix my script for generating the publications page, but you can find a brief background/research statement there along with the other usuals like CV, recent projects, and classes taught.

This summer I've been working on optimizing compilation for a linear algebra DSL. This is an extension of Jeremy Siek's work on Built-to-Order BLAS functions. Often times it's more efficient to have a specialized function which fuses two or more BLAS functions. The idea behind BTO is that we'd like to specify these functions at a high level (i.e., with liner algebra expressions) and then automatically perform the optimizing transformations which have made BLAS such a central component of linear algebra computations.

The current/prior version of BTO already handles loop fusion, memory bandwidth constraints, and more. However, it is not currently aware any high-level algebraic laws such as the fact that matrix multiplication is associative, addition is associative and commutative, transposition reverse-distributes over multiplication, etc. My goal is to make it aware of these sort of things.

Along the way, one thing to do is solve the chain multiplication problem: given an expression like ∏[x1,x2...xN] figure out the most efficient associativity for implementing it via binary multiplication. The standard solution is to use a CKY-like dynamic programming algorithm to construct a tree covering the sequence [x1,x2...xN]. This is easy to implement, but it takes O(n^3) time and O(n^2) space.

I found a delicious alternative algorithm which solves the problem in O(n*log n) time and O(n) space! The key to this algorithm is to view the problem as determining a triangulation of convex polygons. That is, we can view [x0,x1,x2...xN] as the edges of a convex polygon, where x0 is the result of computing ∏[x1,x2...xN]. This amazing algorithm is described in the tech report by Hu and Shing (1981a), which includes a reference implementation in Pascal. Unfortunately the TR contains a number of typos and typesetting issues, but it's still pretty legible. A cleaner version of Part I is available here. And pay-walled presumably-cleaner versions of Part I and Part II are available from SIAM.

Hu and Shing (1981b) also have an algorithm which is simpler to implement and returns a heuristic answer in O(n) time, with the error ratio bounded by 15%. So if compile times are more important than running times, then you can use this version as well. A pay-walled version of the article is available from Elsevier.