Last modified: 2014-08-19 21:37:35 UTC
HHVM uses a persistent SQLite database as a bytecode cache. This cache will be dynamically maintained by the HHVM server process, but it can also be pre-populated to bootstrap a newly deployed codebase [0]. Pre-compiling the cache during scap and distributing that cache to the HHVM worker nodes should provide a significant performance increase. [0]: http://hhvm.com/blog/4061/go-faster