On Monday, I spent most of the morning trying to get some work done, in order to not completely screw myself over upon return. Since my job that morning was to be a "floater", I couldn't do anything anyway (a floater is someone who just sits around unless the volunteer captains need them to do something). I missed my advisor's keynote at the Dynamic Languages Symposium (DLS) as a result, but I'm pretty sure I've heard the material many times before.
I was able to attend the afternoon session of DLS, which had some interesting talks. A paper on type reconstruction of dynamic languages looked like an idea that had briefly crossed my mind awhile ago, but it was fully explored, so it was interesting to hear about it. There was also a talk on running a VM on a VM (mind blowing, right?) to take advantage of the host VMs JIT, in order to eliminate the need to write a JIT on the guest VM. I find [nearly] all things JIT exciting, so it was a good talk, even if their performance benchmarks were extremely slow.
The Ruby Intermediate Language was presented as an easy-to-analyze intermediate form of Ruby. It was mildly interesting, but personally, I don't find Ruby's hard-to-parse intricacies very interesting, as I ended up ditching Ruby for Python a few years ago. RIL basically eliminates the hard-to-parse syntax from Ruby, converting them to their easier-to-parse equivalents, with the goal of making analysis tools easier to write.
There was a complaint at the end where someone questioned the necessity of RIL instead of just providing a standalone parser (which RIL may have, since none existed at the inception of RIL), which I tend to agree with — even though I think source transformations that simplify the AST are common, I don't think they're usually as game-changing as the ones presented in RIL. For example, the Java compiler collapses concatenation of string literals down to one literal, and it's completely transparent (and irreversible, as far as I can tell) to compiler plugins. However, this isn't the same as eliminating the ambiguities in Ruby that make life "pleasant" or "natural" for Ruby developers. It would be equivalent to a "JIL" that converted Java for-each loops into the ugly for-loop equivalents, for example, only Ruby offers many more conveniences than Java.
The last talk of the day was about object heaps on manycore (56) hardware. David Ungar from IBM presented experimental results using a Smalltalk VM. It's always interesting to hear about VMs and concurrency (and this is both!), since the topics are often at the threshold of my knowledge. However, this presentation was less "controversial" than the RIL paper, so I don't have much to say about it, aside from feel free to read all of the mentioned papers, if they sound interesting.
An interesting day overall. After all, who doesn't like dynamic languages?
No comments:
Post a Comment