Mailman 3 May 2011 - pypy-dev

Re: [pypy-dev] Tealets Re: Stackless work
by Andrew Francis 23 May '11

23 May '11

Hi Da_Blitz: --- On Mon, 5/23/11, Da_Blitz <pypy(a)pocketnix.org> wrote: > i have seen some tealets work but no documentation from > what i have overheard it seems like stackless should be posible if > someone devoes quite a bit of effort to it, I just googled and I see work has been done. Too cool! > got anymore info on this join pattern, i am hoping/assuming its not a fork >and join to offload to other cores as i am more concerned > about having huge amounts of threads/tasklets Join Patterns come out of JoCAML and Polyphonic C#. They are a concurrency mechanism. Hopefully, Join Patterns will allow a Stackless programmer to do stuff like # block until all nine reindeer channels are ready to receive, OR # any three of Santa's too many to count army of elves OR # a timeout occurs reindeer = [CUPID, DASHER, PRANCER, BLITZEN, DANCER, COMET, VIXEN, \ DONNER,RUDOLPH] results = stackless.select(joinPattern(reindeer, joinPattern(elves, 3), timeout) if results == reindeer: for result in reindeer: # read data from each reindeer ..... right now join patterns and stackless.select() are not official parts of C based Stackless Python or stackless.py Cheers, Andrew

1 0

Re: [pypy-dev] [pypy-commit] pypy default: Move the loop in contains() into a seperate function, so it can be inlined if it doesn't hit the fallback case.
by Maciej Fijalkowski 23 May '11

23 May '11

We should do such stuff automatically - it's getting annoying. On Sat, May 21, 2011 at 5:08 AM, alex_gaynor <noreply(a)buildbot.pypy.org> wrote: > Author: Alex Gaynor <alex.gaynor(a)gmail.com> > Branch: > Changeset: r44339:5809a5ace634 > Date: 2011-05-20 22:19 -0500 > http://bitbucket.org/pypy/pypy/changeset/5809a5ace634/ > > Log: Move the loop in contains() into a seperate function, so it can be > inlined if it doesn't hit the fallback case. >

1 0

help with translation (error in JIT code)
by wlavrijsen＠lbl.gov 23 May '11

23 May '11

Hi, trying to translate reflex-support branch, which was recently move up to 1.5, I'm stuck at the error below. Any ideas? Thanks! Best regards, Wim [translation:ERROR] Error: [translation:ERROR] Traceback (most recent call last): [translation:ERROR] File "translate.py", line 306, in main [translation:ERROR] drv.proceed(goals) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/translator/driver.py", line 809, in proceed [translation:ERROR] return self._execute(goals, task_skip = self._maybe_skip()) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/translator/tool/taskengine.py", line 116, in _execute [translation:ERROR] res = self._do(goal, taskcallable, *args, **kwds) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/translator/driver.py", line 286, in _do [translation:ERROR] res = func() [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/translator/driver.py", line 397, in task_pyjitpl_lltype [translation:ERROR] backend_name=self.config.translation.jit_backend, inline=True) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/metainterp/warmspot.py", line 46, in apply_jit [translation:ERROR] **kwds) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/metainterp/warmspot.py", line 196, in __init__ [translation:ERROR] self.codewriter.make_jitcodes(verbose=verbose) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/codewriter.py", line 72, in make_jitcodes [translation:ERROR] self.transform_graph_to_jitcode(graph, jitcode, verbose) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/codewriter.py", line 41, in transform_graph_to_jitcode [translation:ERROR] transform_graph(graph, self.cpu, self.callcontrol, portal_jd) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 25, in transform_graph [translation:ERROR] t.transform(graph) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 39, in transform [translation:ERROR] self.optimize_block(block) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 61, in optimize_block [translation:ERROR] oplist = self.rewrite_operation(op) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 196, in rewrite_operation [translation:ERROR] return rewrite(self, op) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 275, in rewrite_op_direct_call [translation:ERROR] return getattr(self, 'handle_%s_call' % kind)(op) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/jtransform.py", line 323, in handle_residual_call [translation:ERROR] calldescr = self.callcontrol.getcalldescr(op) [translation:ERROR] File "/home/wlav/pypydev/pypy/pypy/jit/codewriter/call.py", line 208, in getcalldescr [translation:ERROR] assert NON_VOID_ARGS == [T for T in ARGS if T is not lltype.Void] [translation:ERROR] AssertionError -- WLavrijsen(a)lbl.gov -- +1 (510) 486 6411 -- www.lavrijsen.net

1 1

Re: [pypy-dev] JVM backend - status and future work possibilities?
by Antonio Cuni 22 May '11

22 May '11

On 16/05/11 17:10, Bo Chen wrote: > Anto, Hello Bo, (I've CCed pypy-dev again, please click "reply all" to make sure the list is included) > Is there any advice on reading the source code of the pypy? that's a hard question, it depends what is your goal. My experience is that if you have a clear goal and start by writing a failing test, then it's easier to follow what's going on in that particular case. Personally, I love to put pdb.set_trace() here and there to watch the execution of the code in the pieces I'm interested in. > And here is my question, the JVM backend is to invoke java classes in pypy or > translate the python binary to java binary, or both? The JVM backend translates RPython source code into JVM bytecode (actually, it generates a textual form of the JVM bytecode, which is then assembled by jasmin). Currently, you cannot create/call Java entities from RPython code, but as I explained in the earlier emails, this is a prerequisite feature before doing the rest of the work. > There is a prototype use JPype, is this mean the JVM backend turn the java > binary code into the python's. no. JPype allows to call Java code from CPython, but there is no translation of java into python involved. Note that JPype would be used only for tests, and not for the actual translation. This is because the RPython code can be either run on top of cpython (for testing) or translated to jvm code. JPype (or equivalent) is needed for the first case For the second case, JPype does not play any role. What it's needed is to teach the RPython toolchain how to deal with Java classes. Remind that the JVM backend already emits JVM bytecode, so it's "enough" to teach the RPython toolchain to transform e.g. "rjvm.System.out.println" into the corresponding invokestatic bytecode. > And only the JIT complier will transfer the python to java. no JIT is involved at this point. I think it's better to keep the JIT out of this particular discussion, to avoid confusion. > And the Point(1) would be not only just use the JPype, but also need > implement totally? see my explanation above, hoping that it answer your question. ciao, Anto

2 1

Stackless work
by Da_Blitz 21 May '11

21 May '11

Hi keeping this one short and simple I am interested in using JIT enabled stackless with pypy (obviously a feature that doesn't exist). i have been playing with the coroutines interface (which i am finding nicer to work with than tasklets) but don't have the skill to pull of the stackless work it has been suggested that pypy devs are willing to work on paid for features. so if anyone is interested in doing work on stackless and being paid for it please feel free to get in touch of me so we can discuss it (either on irc or via email) there isnt a deadline for compleation of this feature, just me wanting to throw a firecracker at someone to at the very least get it started Da_Blitz

1 0

PyPy compile memory requirements
by David Naylor 20 May '11

20 May '11

Hi, I'm sure you all know that translating pypy requires a large amount of RAM. It has been my experience that GCC also uses a large amount of RAM when compiling, about 4G. Is it possible to get python/pypy to stop once it has written the source files and then to run that build process separately, thus avoiding having both pypy and gcc in memory at the same time? This is on FreeBSD/amd64: # gcc -v Using built-in specs. Target: amd64-undermydesk-freebsd Configured with: FreeBSD/amd64 system compiler Thread model: posix gcc version 4.2.2 20070831 prerelease [FreeBSD] I've observed similar behavior using other versions of gcc. Clang does not exhibit this problem. Regards, David

4 4

PyPy platforms/architectures
by David Naylor 18 May '11

18 May '11

Hi, I am busy writing a port for pypy on FreeBSD. I have the basics done but I have a slight issue with the package list. It appears pypy creates some files based on the architecture name [1] however the names it uses for those files differ to what FreeBSD uses [2]. I've mapped the obvious name difference (i386->x86_32, amd64->x86_64) however I am not sure what pypy uses for the other architects. Also it appears FreeBSD uses "powerpc" for ppc (even though [2] says otherwise). Please can you clarify the architecture names. Regards, David [1] Under lib_pypy/ctypes_config_cache [2] http://www.freebsd.org/platforms/

2 1

_pypy_wait.py fixes
by Da_Blitz 17 May '11

17 May '11

Hi when calling wait3 or wait4 in the os module a ctypes error about return types is printed. a quick patch is included bellow to silence the message --- Da_Blitz diff -r eaba7adf5188 lib_pypy/_pypy_wait.py --- a/lib_pypy/_pypy_wait.py Thu Apr 28 11:46:10 2011 +0200 +++ b/lib_pypy/_pypy_wait.py Tue May 17 22:43:29 2011 +1000 @@ -8,8 +8,10 @@ c_wait3 = libc.wait3 c_wait3.argtypes = [POINTER(c_int), c_int, POINTER(_struct_rusage)] +c_wait3.restype = c_int c_wait4 = libc.wait4 +c_wait4.restype = c_int c_wait4.argtypes = [c_int, POINTER(c_int), c_int, POINTER(_struct_rusage)]

1 0

new bugs.pypy.org / ssl warning
by holger krekel 16 May '11

16 May '11

Hi all, the pypy issue tracker has moved from the good old codespeak host to a new one. If you go to one of the old URLs you should be transparently redirected to https://bugs.pypy.org You should be able to use your original login, however please note: - if you used your codespeak credentials and don'T know your actual roundup password you may need to trigger a password reset via "Lost your Login?" - you'll get an SSL warning because we use a self-signed certificate (the IP is shared with some other domains). FYI the cert is from "merlinux" and has this SHA1 Fingerprint: 1D:EF:49:8D:77:BE:45:B6:16:1E:79:BC:3D:2C:81:67:CE:3B:86:D8 - The tracker email address is now tracker(a)bugs.pypy.org in case you are using filters. - there maybe some more problems lurking, please report here or to me privately. many thanks to Ronny Pfannschmidt who majorly helped with the migration of the pypy tracker. It is now running on roundup-1.4 instead of an ancient version. This is good preparation for when we potentially/ eventually move to python.org / PSF infrastructure. For now, the tracker remains hosted courtesy of my little company http://merlinux.eu. best, holger

1 0

sqlite warnings
by Joe 16 May '11

16 May '11

I ran a script I had that used SQLite with storm, and saw a bunch of RuntimeWarnings about args and return times. I took some time to fix them up, and ran the sqlite tests to fix any remaining ones. Joe

2 1