Mailman 3 September 2024 - pypy-dev

Help me understand why PyPy is so slow in my benchmark
by Jeremy Brown Sept. 23, 2024

Sept. 23, 2024

Hello, I'm working on a project that involves benchmarking CPython and PyPy across a few benchmarks I've devised, but PyPy is so much slower than CPython I feel I must have made some sort of mistake. One major headscratcher is that CPython runtimes hold stable while PyPy appears to get worse as the benchmark progresses. I've included an excerpt of the benchmark below that shows the issue, if I can attach files that won't be lost I can attach a full copy; I don't believe the string operations are the issue because the micro-tuning tips suggest the JIT reduces concatenation overhead in scenarios like these (if I'm reading the page right). [1] def recurse(num): if num >= 1000: return "M" + recurse(num - 1000) elif num >= 900: return "CM" + recurse(num - 900) elif num >= 500: return "D" + recurse(num - 500) elif num >= 400: return "CD" + recurse(num - 400) elif num >= 100: return "C" + recurse(num - 100) elif num >= 90: return "XC" + recurse(num - 90) elif num >= 50: return "L" + recurse(num - 50) elif num >= 40: return "XL" + recurse(num - 40) elif num >= 10: return "X" + recurse(num - 10) elif num >= 9: return "IX" + recurse(num - 9) elif num >= 5: return "V" + recurse(num - 5) elif num >= 4: return "IV" + recurse(num - 4) elif num >= 1: return "I" + recurse(num - 1) else: return "" -Jeremy [1]: https://pypy.org/performance.html#micro-tuning-tips

4 7

Operating system package...
by steve.b＠osfda.org Sept. 5, 2024

Sept. 5, 2024

I have a service and companion utilities that runs on both debian and ubuntu (to start...) This is linux-specific, requiring root permissions of the server service since I am making adjustments to file permissions for the purposes of an extensible archiving system for end-users (the server validates any requests made over a unix file socket based on a peer's uid and gid, and manages files for archival for the user -yes, turns out you can get that info from a file socket on linux; you can't with a network socket...) I made a python3 script for the server service and a python3 install script that sets up everything right for the server service. Because it is doing some pretty privileged service setup, root permissions are needed for the installation script. Right now the programs check the effective uid of the operator, and if it's not root it exits with a message that root is required (so it could be done with sudo or su...) Here's an important point: I need to install a few linux packages (my python install script does that by shelling to apt-get...); I get the appropriate package names for the distro in question (my install script consults a file that comes with it...) Right now if the distro is not debian or ubuntu, it tells the person installing that the distro is unsupported and exits. I tried doing a search on this, but I am getting a hell of a lot of noise in the results. My question is: can pypy be used to distribute a linux-oriented installer for specific distros such as I describe here? Can you point me at the documentation to do that, or a good example pypy package that does something like this?? [Because it is closely coupled to linux API calls and has to be root, I do not have it use venv...] I am now also working on distro-specific install packages; but it would be nice to open access to any and all pypy users...

1 0

How to test code that uses weakrefs
by Dima Tisnek Sept. 4, 2024

Sept. 4, 2024

We've got a library that supports py 3.8 .. 3.12 I though I'd give it a go with newest pypy release, and a couple of unit tests fail. I've traced it down to the expectation that objects will disappear from weakref.WeakKeyDictionary. In some cases, this is a user-facing thing, we issue a warning for "unused" resources, by virtue of them missing from the weak container when we expect the user code to keep the reference around. Is there a good idiom or some way to support this behaviour with pypy? Thanks, Dima Tisnek

3 2