Exploratory action

AlaMVic

Distilling VMs – A Generative Approach to Virtual Machine Construction
Distilling VMs – A Generative Approach to Virtual Machine Construction

Language Virtual Machines (VMs) are pervasive in every laptop, server and smart-phone. Industry-level VMs (e.g., Google, Oracle, Microsoft)  use highly-engineered optimisation techniques, often hand-crafted by experts, difficult to reproduce, replicate and change. Such optimisation techniques target mostly speed improvements, and are incompatible with constraints such as space and energy efficiency important in the fields of IoT or robotics.

In AlaMVic we propose to approach VM construction using a holistic generative approach, in contrast with existing approaches that focus on speed and single VM components such as the JIT compiler. We explore how to transform handcrafted optimisations into generation heuristics, how they are applied and combined in fields such as IoT and robotics, and new methods and metrics to evaluate VMs in such fields.

Inria teams involved
RMOD

Members

Stephane Ducasse

Scientific leader