PyPy for SCONE

PyPy is a Just In Time compiler for Python. We maintain a container image that includes PyPy running inside of an enclave.

The speed of pypy inside of an enclave is in many cases faster than running natively (i.e., outside an enclave) using the standard CPython interpreter. To compare the performance we are using the standard python / pypy speed center and add pypy for SCONE.

Below you can see a graph that gives an overview of the performance of different Python variants. This depicts normalized performance using pypy as the baseline, i.e., smaller values show better performance. For example, a ratio of 65 means that the program is indeed 65 times slower than pypy running in native mode (i.e., these are not percent values). We can see in this graph that pypy can achieve for some benchmarks dramatic speedups. pypy running inside of an enclave is in almost all benchmarks faster than native CPython:

You can compare the performance in more details using the speedcenter website maintained by UFCG.