Monday, July 6, 2015

Post-PLDI 2015

Attending PLDI 2015 was a great experience. I had a chance to see many great talks at both PLDI and FCRC, meet with fellow students and faculty in both PL and architecture, and take a little time to explore Portland.

I was especially happy to meet my future colleague Peter-Michael Osera, who will be joining me as a new faculty member at Grinnell College this fall. I enjoyed seeing Michael Stonebraker's Turing Award lecture. Some stand-out talks from the PLDI program include Dan Barowy's work on FlashRelate, Stelios Sidiroglou-Douskos' work on repairing vulnerable programs with automatic code transfer, and Nuno P. Lopes' work on provably correct optimizations with Alive.

I think the dual tracked conference format worked well (it was the first time I had attended a dual-tracked PLDI), especially with the one minute madness videos playing in the lobby. PLDI 2015 was a great experience, and I'm already looking forward to PLDI in Santa Barbara, CA next year.

Post-PLDI 2015

PLDI 2015 was a great experience. I had lots of chances to talk to people about interesting work and heard some very good talks, just like I hoped. I also got great feedback after my talk, which I really appreciated.

I also really enjoyed hearing the presentations at the SIGPLAN Townhall meeting. I was shocked at how much planning and work Steve Blackburn put in as PC chair, and he did a good job summarizing the lessons learned from the whole experience. Jan Vitek described the status of things like artifact evaluation, blind reviewing, and open access across the different SIGPLAN conferences. He also explained the results of various polls of SIGPLAN members on their views of these issues. Although there is a lot of disagreement, it's great that the SIGPLAN community actively discusses these issues. I think it's important for graduate students like me to pay attention, since some day our cohort will take a larger role in managing these things.

Friday, July 3, 2015

Back from PLDI

This is somewhat belated, but I recently (sort of) returned from a trip to Portland to attend PLDI, where I gave a talk entitled "Peer-to-peer Affine Commitment Using Bitcoin", or, more whimsically, "Massively Multiplayer Online Linear Logic". I really enjoyed the opportunity to present this work and discuss it with people; people seemed to find it very intriguing, if perhaps a little silly.

I really enjoyed the chance to talk to people about their work, and to catch up with and make new friends from other parts of the PL community.

Some of the talks I found most interesting were:
  •  "Automatically Improving Accuracy for Floating Point Expressions"; a very well delievered talk about automatically implementing numerical methods transformations on floating point computations
  •  "Verifying Read-Copy-Update in a Logic for Weak Memory"
  •  "Composing Concurrency Control"
  •  "A Formal C Memory Model Supporting Integer-Pointer Casts"; this work is near and dear to me because I am also working on designing an alternative for a problematic part of the C/C++ spec. The particular problem even has a similar root cause: an unfortunate reliance on syntactic data dependence to define semantic behavior.
  •  "Profile-Guided Meta-Programming"
  •  "Synthesizing Data Structure Transformations from Input-Output Examples" and "Type-and-Example-Directed Program Synthesis"; I'm a huge sucker for example based program synthesis even though I have yet to be convinced that it is actually useful