Instagram has one of the world’s largest deployments of Python and we are CPU-bound while executing Python. With our unique set of constraints—many machines, many forked workers per machine, and frequent deploys—it is not straightforward to apply an off-the-shelf optimized Python runtime. We present our new optimizing JIT compiler, static type system, and runtime type optimization built on top of CPython. We show significant performance improvements over stock CPython and developer experience improvements as we replace C extension code with Static Python code.
Tue 7 JunDisplayed time zone: Amsterdam, Berlin, Bern, Rome, Stockholm, Vienna change
09:00 - 10:30
|Cinder: We didn't start the fire
Maxwell Bernstein Meta
|HPy: How To Design a C API For Optimizing Runtimes
Tim Felgentreff Oracle Labs, Potsdam