Speed up Python (according to pybench and 2to3-on-itself) by 1-2% by caching
whether any thread has tracing turned on, which saves one load instruction in
the fast_next_opcode path in PyEval_EvalFrameEx().  See issue 4477.
1 file changed