Can pypy handle streams and sockets quickly compared to C hand written?

Question

Does pypy help handle streams and sockets quickly compared to C hand written? Compared to regular python?

I would just try, but the python code in question was written for a small cluster of computers on which I am not an admin. I ask here because my google attempts provided a comparison with citon, bottomless swallowing, etc., and I don’t want to trick the administrator about it if this is unlikely to work.

I really don't need pypy to be as good in C; I want to use it because right now the overhead of the translator completely overshadows the calculation I'm trying to do. I just need pypy to get me in the area of ​​handwritten C.

Answer 1

Does pypy help handle streams and sockets quickly compared to C hand written?

Not. This is usually the same or worse.

PyPy retains the global interpreter lock (GIL) that CPython has. This means that native threads cannot run Python code in parallel. Python threads also have extra semantics that are expensive. Greater synchronization surrounds starting, shutting down, and tracking threads of Python objects. In comparison, C threads run faster, are cheaper to use, and can run completely parallel.

Efficient socket handling requires minimizing the time that was not spent waiting for the next socket event. Since the PyPy thread model is still GIL-linked, this means that threads that return from blocking socket calls cannot act until they acquire a GIL. Equivalent C code is usually faster and may return to wait for socket events sooner.

Compared to regular python?

Yes. But not so much.

For the above reasons, PyPy, with the exception of random spikes due to JIT and other overhead, requires less processor time for equivalent code. Therefore, thread and socket processing is faster and more flexible.

I would just try, but the python code in question was written for a small cluster of computers on which I am not an admin. I ask here because my google attempts provided a comparison with citon, bottomless swallowing, etc., and I don’t want to trick the administrator about it if this is unlikely to work.

PyPy will only improve performance if your code is CPU related. PyPy is the fastest Python implementation that I know about what happens initially. You can explore some of the other implementations or consider writing C extensions if true parallelism streaming is your priority.

I really don't need pypy to be as good in C; I want to use it because right now the overhead of the translator completely overshadows the calculation I'm trying to do. I just need pypy to get me in the area of ​​handwritten C.

Closing the performance gap with C is currently PyPy's biggest feature. I highly recommend you to try.

Answer 2

It should handle sockets and streams pretty quickly, I can’t assure you that it will be as fast as C, PyPy uses protective devices to make sure that the values ​​are acceptable to run in function C. Since PyPy uses JIT, it will offer the greatest performance gain in the case of large cycles. You can see the pypy speed tests here

Answer 3

The problem is not so much in routine calls in standard APIs as in pypy processing. The less pypy processing, the more you are going to get closer to C speed (since after execution has entered the actual API code, the difference in execution speed is no different). I would not expect miracles, though.

I found this comparison about how different programming languages ​​stand against each other, and although pypy is not enabled, python does not do all that well in terms of raw speed.