Potential project list¶
Simple tasks for newcomers¶
- Tkinter module missing support for threads: https://bitbucket.org/pypy/pypy/issue/1929/tkinter-broken-for-threaded-python-on-both
- Optimize random: https://bitbucket.org/pypy/pypy/issue/1901/try-using-a-different-implementation-of
- Implement AF_XXX packet types of sockets: https://bitbucket.org/pypy/pypy/issue/1942/support-for-af_xxx-sockets
Below is a list of projects that are interesting for potential contributors who are seriously interested in the PyPy project. They mostly share common patterns - they’re mid-to-large in size, they’re usually well defined as a standalone projects and they’re not being actively worked on. For small projects that you might want to work on look above or either look at the issue tracker, pop up on #pypy on irc.freenode.net or write to the mailing list. This is simply for the reason that small possible projects tend to change very rapidly.
This list is mostly for having an overview on potential projects. This list is by definition not exhaustive and we’re pleased if people come up with their own improvement ideas. In any case, if you feel like working on some of those projects, or anything else in PyPy, pop up on IRC or write to us on the mailing list.
Explicit typing in RPython¶
RPython is mostly based around type inference, but there are many cases where
specifying types explicitly is useful. We would like to be able to optionally
specify the exact types of the arguments to any function. We already have
solutions in that space,
@rpython.rlib.signature.signature, but they are inconvenient and limited.
For instance, they do not easily allow to express the type “dict with ints as
keys and lists of instances of Foo as values”.
Additionally, we would like to be able to specify the types of instance attributes. Unlike the function case, this is likely to require some refactoring of the annotator.
Make bytearray type fast¶
PyPy’s bytearray type is very inefficient. It would be an interesting task to look into possible optimizations on this. (XXX current status unknown; ask on #pypy for updates on this.)
Implement copy-on-write list slicing¶
The idea is to have a special implementation of list objects which is used
myslice = mylist[a:b]: the new list is not constructed
immediately, but only when (and if)
mylist are mutated.
Our cpyext C-API compatiblity layer can now run upstream NumPy unmodified. Release PyPy2.7-v5.4 still fails about 60 of the ~6000 test in the NumPy test suite. We could use help analyzing the failures and fixing them either as patches to upstream NumPy, or as fixes to PyPy.
We also are looking for help in how to hijack NumPy dtype conversion and ufunc calls to allow the JIT to make them fast, using our internal _numpypy module.
Improving the jitviewer¶
Analyzing performance of applications is always tricky. We have various tools, for example a jitviewer that help us analyze performance.
The old tool was partly rewritten and combined with vmprof. The service is hosted at vmprof.com.
The following shows an old image of the jitviewer. The code generated by the PyPy JIT in a hierarchical way:
- at the bottom level, it shows the Python source code of the compiled loops
- for each source code line, it shows the corresponding Python bytecode
- for each opcode, it shows the corresponding jit operations, which are the ones actually sent to the backend for compiling (such as
i15 = i10 < 2000in the example)
The jitviewer is a web application based on django and angularjs: if you have great web developing skills and want to help PyPy, this is an ideal task to get started, because it does not require any deep knowledge of the internals. Head over to vmprof-python, vmprof-server and vmprof-integration to find open issues and documentation.
Optimized Unicode Representation¶
CPython 3.3 will use an optimized unicode representation (see PEP 0393) which switches between different ways to represent a unicode string, depending on whether the string fits into ASCII, has only two-byte characters or needs four-byte characters.
The actual details would be rather different in PyPy, but we would like to have the same optimization implemented.
Or maybe not. We can also play around with the idea of using a single representation: as a byte string in utf-8. (This idea needs some extra logic for efficient indexing, like a cache.)
(XXX this is unlikely to be feasible.)
- Incremental or distributed translation.
- Allow separate compilation of extension modules.
PyPy has pluggable garbage collection policy. This means that various garbage collectors can be written for specialized purposes, or even various experiments can be done for the general purpose. Examples:
- A garbage collector that compact memory better for mobile devices
- A concurrent garbage collector (a lot of work)
- A collector that keeps object flags in separate memory pages, to avoid un-sharing all pages between several fork()ed processes
STM (Software Transactional Memory)¶
This is work in progress. Besides the main development path, whose goal is to make a (relatively fast) version of pypy which includes STM, there are independent topics that can already be experimented with on the existing, JIT-less pypy-stm version:
- What kind of conflicts do we get in real use cases? And, sometimes,
which data structures would be more appropriate? For example, a dict
implemented as a hash table will suffer “stm collisions” in all threads
whenever one thread writes anything to it; but there could be other
implementations. Maybe alternate strategies can be implemented at the
level of the Python interpreter (see list/dict strategies,
- More generally, there is the idea that we would need some kind of “debugger”-like tool to “debug” things that are not bugs, but stm conflicts. How would this tool look like to the end Python programmers? Like a profiler? Or like a debugger with breakpoints on aborted transactions? It would probably be all app-level, with a few hooks e.g. for transaction conflicts.
- Find good ways to have libraries using internally threads and atomics,
but not exposing threads to the user. Right now there is a rough draft
lib_pypy/transaction.py, but much better is possible. For example we could probably have an iterator-like concept that allows each loop iteration to run in parallel.
Introduce new benchmarks¶
We’re usually happy to introduce new benchmarks. Please consult us before, but in general something that’s real-world python code and is not already represented is welcome. We need at least a standalone script that can run without parameters. Example ideas (benchmarks need to be got from them!):
Make more python modules pypy-friendly¶
A lot of work has gone into PyPy’s implementation of CPython’s C-API, cpyext, over the last years to let it reach a practical level of compatibility, so that C extensions for CPython work on PyPy without major rewrites. However, there are still many edges and corner cases where it misbehaves.
For any popular extension that does not already advertise full PyPy compatibility, it would thus be useful to take a close look at it in order to make it fully compatible with PyPy. The general process is something like:
- Run the extension’s tests on PyPy and look at the test failures.
- Some of the failures may be solved by identifying cases where the extension relies on undocumented or internal details of CPython, and rewriting the relevant code to follow documented best practices. Open issues and send pull requests as appropriate given the extension’s development process.
- Other failures may highlight incompatibilities between cpyext and CPython. Please report them to us and try to fix them.
- Run benchmarks, either provided by the extension developers or created by you. Any case where PyPy is significantly slower than CPython is to be considered a bug and solved as above.
Alternatively, an approach we used to recommend was to rewrite C extensions using more pypy-friendly technologies, e.g. cffi. Here is a partial list of good work that needs to be finished:
Status: the repo is an older version of matplotlib adapted to pypy and cpyext
TODO: A suggested first step would be to merge the differences into matplotlib/HEAD. The major problem is the use of a generic view into a numpy ndarray. The int* fields would need to be converted into int[MAX_DIMS] c-arrays and filled in.
Status: A GSOC 2013 project to adapt the Phoenix sip build system to cffi
TODO: Merge the latest version of the wrappers and finish the sip conversion
Status: see blog post <http://morepypy.blogspot.com/2014/03/pygamecffi-pygame-on-pypy.html>
TODO: see the end of the blog post