CF Bolz-Tereick
CF Bolz-Tereick
I don't know, it does feel a little bit weird that pyperf gives a result that is a factor of 500x slower than timeit? --loops and --warmups shouldn't really come...
FWIW, I have been running the pytest-html test suite with coverage locally as a kind of real world benchmark for coverage performance (see linked issue). seems we can speed up...
FWIW, that's the definition for ``unicode`` strings in pypy, where you need to look up in the unicode db to find out what the whitespace chars are. I doubt that's...
It's possible to pass the underlying buffer of a PyPy string to a C function without copying it, so that wouldn't be a problem.
this depends on #17 (it's not terribly hard but a bit of a mess, we "just" need to implement signal handling)
Improving the list specialization size might improve things: https://gist.github.com/cfbolz/3d354b82245ad5b020f5 (eg the ConsEnvArbitrary allocations)
What kind of guarantees does the rng make? We can probably also just use rlib.rrandom.
the first problem I suspect is a complicated bug/limitation in rpython's handling of function pointers. I tried to make primites rely less on those in the branch prims-fewer-func-ptrs, which makes...
successfully in the sense that I now get a ``SchemeException``, meaning it behaves the same as cpython at least.
@davidhewitt thanks for the ping. the problem has been fairly annoying to debug for me, I haven't managed to get it to show up on my dev machine. Does it...