Back to proposals-2013

Title: Fast Enough VMs in Fast Enough Time
Proposer: laurence_tratt
Type: Tutorial/Experience Report
Duration: 90 minutes
Description: Programming language designers face an unpleasant dilemma when it comes to   implementing their languages: too little implementation, and it will be   laughed at as too slow; too much, and it will divert energy away from design.   Lacking the manpower to make a plausibly fast implementation, many   interesting language design ideas have faded unfairly into obscurity. We all   live with the consequences.

  In this talk I look at a new mode of creating “fast enough virtual machines   in fast enough time” Virtual Machines (VMs), using the meta-tracing JIT   language RPython. Unlike previous approaches, RPython creates VMs that   automatically come with a JIT customised for the language being interpreted.   RPython has been used to implement a new VM for Python that gives an average   speed-up of 5x over the stock VM, and thus demonstrably scales to “real”   languages. Such VMs will change the language landscape of the future.

  I will explain the performance trade-offs of existing approaches to language   implementation, how meta-tracing works, and what the performance   consequences (good and bad) for programmers of this new breed of VMs are.

Asti - eh, I'm not crazy about this
Roger: hard to tell if this is a 'brave new idea' or yet another cul-de-sac.