Mailman 3 December 2013 - pypy-dev

AttributeError: 'socket' object has no attribute '_reuse'
by Robert Voigtländer 19 Dec '13

19 Dec '13

Thanks to Piotr I now got gevent installed. But now I get a new error when trying to run my python program. The Webserver (with SSE) runs but when I try to load a site I get the error below. I need to get it running .. pypy is so awesomely faster than cpython. Robert Traceback (most recent call last): File "build/bdist.linux-armv6l/egg/gevent/greenlet.py", line 328, in run result = self._run(*self.args, **self.kwargs) File "build/bdist.linux-armv6l/egg/gevent/pywsgi.py", line 653, in handle handler = self.handler_class(socket, address, self) File "build/bdist.linux-armv6l/egg/gevent/pywsgi.py", line 175, in __init__ self.rfile = socket.makefile('rb', -1) File "build/bdist.linux-armv6l/egg/gevent/socket.py", line 379, in makefile return _fileobject(type(self)(_sock=self), mode, bufsize) File "/usr/lib/pypy-upstream/lib-python/2.7/socket.py", line 300, in __init__ sock._reuse() AttributeError: 'socket' object has no attribute '_reuse' <Greenlet at 0x4737a10L: <bound method WSGIServer.handle of <WSGIServer at 0x465d630L fileno=6 address=0.0.0.0:9999>>(<socket at 0x4737a30L fileno=8 sock=192.168.1.199:, ('84.63.102.205', 52661))> failed with AttributeError

4 5

gevent
by Robert Voigtländer 19 Dec '13

19 Dec '13

Hi, I am relatively new to python and more or less completely new to pypy. So far I have pypy running on my raspberry pi. I don’t use virtualenv. Modules for pypy I simply installed via easy_installer from the pypy bin dir. I now try to get gevent installed. I found this remark: “stackless support - eventlet <http://eventlet.net/> just works and gevent <http://www.gevent.org/> requires pypycore<https://github.com/gevent-on-pypy/pypycore> and pypy-hacks <https://github.com/schmir/gevent/tree/pypy-hacks> branch of gevent (which mostly disables cython-based modules)” On the github site there is a small instruction on how to install gevent with pip in the virtualenv. $ virtualenv -p /path/to/bin/pypy venv $ source venv/bin/activate (venv)$ pip install git+git://github.com/schmir/gevent@pypy-hacks (venv)$ pip install cffi (venv)$ pip install git+git://github.com/gevent-on-pypy/pypycore (venv)$ export GEVENT_LOOP=pypycore.loop But I am not able to translate this into an installation with easy_install for my environment (without venv) Can someone give me a tip for getting started? Thanks Robert

2 1

FreeBSD 9.2 buildslave: first run
by Tobias Oberstein 16 Dec '13

16 Dec '13

Hi, the new FreeBSD buildslave has run the first time: http://buildbot.pypy.org/builders/pypy-c-jit-freebsd-9-x86-64/builds/6 PyPy was built successfully, but there are a couple of issues: https://bugs.pypy.org/issue1637 https://bugs.pypy.org/issue1638 https://bugs.pypy.org/issue1639 I'll address those. There is another issue that bugs me .. look at the built times: [Timer] Timings: [Timer] annotate --- 576.0 s [Timer] rtype_lltype --- 1323.7 s [Timer] pyjitpl_lltype --- 1098.6 s [Timer] backendopt_lltype --- 314.2 s [Timer] stackcheckinsertion_lltype --- 161.5 s [Timer] database_c --- 451.3 s [Timer] source_c --- 668.5 s [Timer] compile_c --- 2403.8 s [Timer] =========================================== [Timer] Total: --- 6997.6 s and compare with "Linux 64 on allegro64": [Timer] Timings: [Timer] annotate --- 641.5 s [Timer] rtype_lltype --- 1528.5 s [Timer] pyjitpl_lltype --- 938.7 s [Timer] backendopt_lltype --- 242.2 s [Timer] stackcheckinsertion_lltype --- 206.8 s [Timer] database_c --- 325.3 s [Timer] source_c --- 357.7 s [Timer] compile_c --- 246.4 s [Timer] =========================================== [Timer] Total: --- 4487.1 s ***** Why the heck does "compile_c" take 10x the time on FreeBSD? Note that FreeBSD builder uses Clang .. could that be a reason? Any hints on that are welcome! Cheers, Tobias

3 3

faster rbigint multiplication
by Mario Pernici 12 Dec '13

12 Dec '13

Hello, I would like to submit a patch to rbigint.py (50 lines of code in _x_mul_), speeding up multiplication; on my computer (x86_64) it seems to be at least as fast as CPython for numbers with few words, about 2x faster for numbers with more than 20 words. Where should I post it? Mario

2 1

Compiling stmgc-c4 fails on 32bit
by Jean-Philippe Caissy 09 Dec '13

09 Dec '13

Hi, I'm trying to compile stmgc-c4 on a fresh Ubuntu 12.04 32bit machine (lxc containers if that matters), and it fails. However, a fresh 12.04 64bit installation, it compiles without any problem. Both the jit and no-jit fail (-Ojit and -O2). The stack trace for the jit is really long, and can be found here : https://gist.github.com/jpcaissy/7862899 and the stack trace for the no-jit can be found here : https://gist.github.com/jpcaissy/7862913 Could that be a bug, or am I doing something wrong? Thanks, JP

2 1

Questions about pypy log
by andrea.jeradi＠studenti.univr.it 09 Dec '13

09 Dec '13

Hello, I'm Andrea Jeradi, a computer science student of university of Verona (Italy). I'm doing a project and I'm using pypy. On Internet I have not found explanation about the meaning of some entry of the log. [b6faf360f285] {jit-summary Tracing: 2 0.006712 Backend: 2 0.000884 TOTAL: 0.955633 ops: 705 recorded ops: 168 calls: 3 guards: 58 opt ops: 263 opt guards: 58 forcings: 0 abort: trace too long: 0 abort: compiling: 0 abort: vable escape: 0 abort: bad loop: 0 abort: force quasi-immut: 0 nvirtuals: 0 nvholes: 0 nvreused: 0 Total # of loops: 1 Total # of bridges: 1 Freed # of loops: 0 Freed # of bridges: 0 [b6faf3623ec5] jit-summary} Exist a site, a paper or something else that explain it? Particularly what means 'abort: vable escape', 'abort: bad loop', 'abort: force quasi-immut', 'nvirtuals', 'nvholes' and 'nvreused'? Thanks for the help, best regards Andrea Jeradi

5 5

Pluggable HTM
by Dimitri Vorona 08 Dec '13

08 Dec '13

Hi, the original STM proposal spoke of HTM as of a thing of a far future. Now, Haswells are out and provide built-in HTM support in form of TSX. In the near future I expect more and more systems to have it. Are there plan to make PyPy use HTM if it is available on the system? Regards, Dimitri.

3 2

Integrating CPython and PyPy, maybe
by Armin Rigo 03 Dec '13

03 Dec '13

Hi all, Starting from last night, the nightly builds of PyPy are able to run this kind of code: http://bpaste.net/show/155166/ (You have to adjust the libpython2.x.so path manually.) The idea is to link in the libpython.so, using cffi. Then you can use it much like you would in C. For example, you can import complete CPython packages and call them. This is similar to the earlier experiments done by fijal about calling matplotlib via CPython, except this time done with cffi. The code above supports __add__(), which lets you just write "x3 = x1 + x2" for any two PyObject instances x1 and x2. It's unclear if it's a good idea, but it's at least a cool hack. To be really useful, it needs much more work, of course, starting with a clearer design of what the goal really would be :-) A bientôt, Armin.

2 3

clarification cffi vs _ffi_ vs _ffi vs libffi vs _rawffi vs rffi
by Dima Tisnek 03 Dec '13

03 Dec '13

there seems to be some confusion (could be all mine) between N terms related to FFI: are there both _ffi_ and _ffi, or is one of those a typo? extending page mentions libffi and not cffi, is it out of date, or is there a reason to stay away from cffi, although recommended in general faq page? in my own testing, I noticed that equivalent code cffi code is faster than ctypes, is there an explanation for this? thanks, d. docs in question: http://doc.pypy.org/en/latest/faq.html http://doc.pypy.org/en/latest/extending.html http://doc.pypy.org/en/latest/rffi.html

4 6

PyPy3 release?
by Armin Rigo 03 Dec '13

03 Dec '13

Hi all, Any volunteer to handle the PyPy3 release corresponding to PyPy 2.2? A bientôt, Armin.

5 6