Wednesday, March 12, 2014

Adwait's report on ASPLOS-2014

This was my 2nd ASPLOS and I was impressed with the quality of the papers and delivery of the talks. I attended GPGPU and Approximate Computing Workshops in addition to the ASPLOS conference.

Workshop/Tutorials
The workshops were conducted very nicely, and I liked both GPGPU and Approximate Computing Workshops

- Both workshops had papers that would interest to broad audience. Even though I don't know much about all the research areas in-depth, I got the high-level ideas of almost all the workshop papers. I would like to thank the organizers of both workshops -- Great Work!

- As most of the works were in their nascent stages, it was nice to talk to authors (and also get feedback on my GPGPU workshop paper), and get their thoughts on potential future works. 

- In GPGPU Workshop, talks/papers from Martin Burtscher (K20 Power) and Nalia (instrumentation of GPGPU kernels) were most interesting to me. Martin's talk was hilarious :-)

- In approximate computing workshop, talk/paper from Benjamin Ransford (Approximate Semantics for Wirelessly Networked Applications) and Mehrzad Samadi (CPU-GPU Collaboration for Output Quality Monitoring) were most interesting. 

Main Conference
- The lightning session helped me significantly in identifying the papers that are of interest to me. 

- Poster Session was also nice -- as I could talk to authors directly and in-depth.

- WACI session was interesting -- Liked the idea of Dirty Caches for Useless (FaceBook) updates. Talk was hilarious. Although not all talks were as WACI as I would have expected. 

- The conference program was very strong. I really appreciate PC Chair and Committee's effort in carving out a strong program. I liked two papers the most : (1) "Disengaged Scheduling for Fair, Protected Access to Fast Computational Accelerators" from University of Rochester, (2) REF: Resource Elasticity Fairness with Sharing Incentives for Multiprocessors from Duke. 
In my opinion, both papers are very neat and can spur lot of future work.

- According to me, talks from Di Wang (Underprovisioning Backup Power Infrastructure for Datacenters) and Seyed Majid Zahed (REF: Resource Elasticity Fairness with Sharing Incentives for Multiprocessors) were most interesting and clear. Conclusion of Di's paper was: "We don't need Diesel generator for backup power". I think this is a very strong conclusion and can have long-term impact. 
Although in Seyed's paper, there was lot of maths, but his talk was very clear and simple, and I could understand almost all of the math he presented in the talk. Looking forward to reading his paper in detail!. 

- Debate: It was very informative, but there was lot of confusion about "YES" or "NO". I guess many people voted in wrong direction :-)

General
- Organization was great. Thanks to Rajeev, and all students of Utah :-). Presence of Indian food in conferences (especially held in US) is probably rare (first time for me). Thanks for it. 

-  The idea of giving fleece jackets was superb. This is the first conference souvenir, which is probably I am going to use for long time. Food during Sunday's reception was a good idea too.

-  ASPLOS was co-located "not-deliberately"with many funny non-technical events :-). It was both good and bad :-)





No comments:

Post a Comment