Speaker
Description
Resource management is a very difficult challenge on embedded devices that run gaming workloads. Different games can have very different workload patterns, and use resources in different ways that are often difficult to predict. For example, one game may be CPU bound and have a single main thread, whereas another game is heavily memory bound and have multiple threads on the rendering path. Of course, embedded devices also tend to be resource constrained.
In order to accommodate these eclectic and unpredictable workload patterns, operating systems that run on these embedded devices must provide a rich set of building blocks for defining resource-management policies. In this talk, we'll describe a solution being built out at Meta right now called Orchestrator, which allows developers to create well-encapsulated, predictable policies using a reactor pattern.