fniephaus_ has quit [Quit: OaO]
kipras is now known as kipras`away
kipras`away is now known as kipras
fniephaus has joined #pypy
vkirilichev has joined #pypy
marr has quit [Ping timeout: 260 seconds]
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
yuyichao_ has quit [Remote host closed the connection]
yuyichao_ has joined #pypy
lritter_ has quit [Ping timeout: 255 seconds]
vkirilichev has quit [Ping timeout: 240 seconds]
tbodt has joined #pypy
jcea has quit [Quit: jcea]
Garen has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
ArneBab has joined #pypy
vkirilichev has joined #pypy
ArneBab_ has quit [Ping timeout: 240 seconds]
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
vkirilichev has quit [Ping timeout: 260 seconds]
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
pilne has quit [Quit: Quitting!]
tbodt has joined #pypy
tbodt has quit [Client Quit]
oberstet has quit [Ping timeout: 246 seconds]
tbodt has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
vkirilichev has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
vkirilichev has quit [Ping timeout: 240 seconds]
tbodt has quit [Client Quit]
dcrosta has joined #pypy
tbodt has joined #pypy
tilgovi has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
vkirilichev has joined #pypy
tbodt has joined #pypy
tbodt has quit [Client Quit]
vkirilichev has quit [Ping timeout: 260 seconds]
jacob22_ has quit [Quit: Konversation terminated!]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
jamesaxl has quit [Read error: Connection reset by peer]
jamesaxl has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
vkirilichev has joined #pypy
jacob22_ has joined #pypy
tbodt has joined #pypy
vkirilichev has quit [Ping timeout: 258 seconds]
vkirilichev has joined #pypy
vkirilichev has quit [Client Quit]
amaury has joined #pypy
vkirilichev has joined #pypy
realitix has joined #pypy
adeohluwa has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
Guest54708 has quit [Ping timeout: 240 seconds]
Remi_M has joined #pypy
<kenaan> mattip cpyext-unhashable 21f235773e6d /pypy/: merge default into branch
<kenaan> mattip cpyext-unhashable c2f5f5816876 /pypy/: test, fix for cpytext hash of empty string subclass not returning -1
marr has joined #pypy
black_ant has joined #pypy
marvin has joined #pypy
marvin is now known as Guest11041
black_ant has quit [Ping timeout: 245 seconds]
mihaid has joined #pypy
black_ant has joined #pypy
oberstet has joined #pypy
amaury has quit [Ping timeout: 240 seconds]
cstratak has joined #pypy
vkirilichev has quit [Remote host closed the connection]
vkirilichev has joined #pypy
<mihaid> hello, I have a question regarding enabling profile-guided optimizations. In order to get this to work on any target (including non-pypy targets), after the profile generation you would have to run a training script / workload. However, different targets would require different scripts with a different set of arguments. I see two possible solutions
<mihaid> for this. Either, there would be a generic script, that by default would suit a PyPy target, which the user might change to suit any other target. Or there can be more profopt arguments added to the translation to include arguments regarding the training script. I'm not sure which these ideas is preferable, though?
vkirilichev has quit [Ping timeout: 240 seconds]
<LarstiQ> are you doing this for pypy or for rpython?
<LarstiQ> how much work is it to submit PRs for both?
antocuni has joined #pypy
amaury has joined #pypy
amaury has quit [Quit: Konversation terminated!]
amaury has joined #pypy
<mihaid> right now I have a PR for pypy that is fully working with PGO and shared, but it does not work for other binaries such as the one in described by targetrpystonedalone.py (mainly because of the issue above)
oberstet has quit [Ping timeout: 240 seconds]
<LarstiQ> mihaid: either you have something that works for pypy only, or you have something that works for all rpython users
<LarstiQ> doing the middle of targetrpystonedalone.py I don't think is useful, except for benchmarking perhaps
jcea has joined #pypy
<mihaid> LarstiQ: well, targetrpystonedalone.py was what cfbolz proposed on my PR to make it more generic. Do you have any advice about another target that is more suitable for a generic PGO, please?
<cfbolz> mihaid: I think the main problem for me is not the argument to the training run, but the fact that the executable name is hard-coded
<cfbolz> mihaid: basically it's fine if every language has to put into the target the default training run it wants to use
<cfbolz> (like what is already in the PyPy target)
amaury has quit [Ping timeout: 255 seconds]
<mihaid> oh, so just to make this clear. The target should be generic, but it is acceptable to expect that other target will have to modify the makefile for themselves when doing profile guided optimizations?
<cfbolz> mihaid: not the makefile
<cfbolz> Just the option
<cfbolz> Sorry, with 'target' I mean the language specific target*.py files
adeohluwa has quit [Quit: Connection closed for inactivity]
<mihaid> cfbolz: okay, i think I got it. One more thing, though. When I last update the PR, I added a profoptpath, as a path to the training script. However, I realized now that its arguments, regardless of the script, are the same as for regression testing, because they are hard coded. Should I also add a profoptargs option to the translation?
<cfbolz> sorry, I don't get the question?
<cfbolz> mihaid: basically what I am saying is that it's fine to ask people to change their target*.py file in the way that targetpypystandalone.py was changed here: https://bitbucket.org/pypy/pypy/pull-requests/540/pgo-enabling/diff#chg-pypy/goal/targetpypystandalone.py
arigato has joined #pypy
<mihaid> so, in my PR, right now you can change the training test for the pgo, but it will run with the same arguments (--pgo -x test_asyncore test_gdb test_multiprocessing test_subprocess ) as the regression test. A different training script however, would need different arguments. Should I pass them in a new variable?
<mihaid> I got the target*.py issue. This was about the training tests.
<antocuni> cfbolz, arigato: so, my mysterious obscure bug has been fixed by this commit: https://bitbucket.org/pypy/pypy/commits/0e91701e92f07e19f9c47e46ea8e4adce363f56f
<antocuni> although I'm still unsure what happens exactly
<antocuni> because it seems to manifest only if there is at least one graph which uses llop.gc_store_indexed AND it's seen by the JIT
<antocuni> (if you add @jit.dont_look_inside, the bug disappears)
<cfbolz> mihaid: ah, I see. so couldn't we rename profoptpath to contain all the arguments necessary?
<arigato> normally this kind of bug shows up as a typing error in rpython
<arigato> (and in the tests you run first too)
<cfbolz> mihaid: so it would be "profoptargs" or so?
<arigato> ah, wait, maybe a longlong-for-storage gets unified with a float?
<antocuni> yes, this is my theory
<mihaid> cfbolz: yeah, I think that would also work. I will try it now. Thanks a lot for the help!
<cfbolz> mihaid: you're welcome, thanks for doing the work!
<arigato> nonsense, then
<antocuni> I *think* that the argument of cpu.write_float_to_mem gets annotated as a float instead of a longlong
<arigato> ah, something like that
<antocuni> but what is puzzling me is why this doesn't happen if the llop is not seen by the codewriter
<arigato> that's because there are two paths that call write_float_to_mem
<arigato> if the new llop is seen, it makes the argument annotated as a float
<arigato> then the crash occurs even in the other path
<arigato> because write_float_to_mem doesn't really check its argument's type, just write it to memory
<arigato> so you get nonsense in the other path
<antocuni> I don't understand how an llop seen or not by the codewriter influences what graphs are annotated or not
<arigato> I mean:
<arigato> write_float_to_mem is called by two functions that we'll call OLD and NEW
<arigato> you have old code in OLD, which calls write_float_to_mem(12371848189798L), where the value is a longlong-encoded float
<arigato> because of the mere presence of NEW, then write_float_to_mem's arg is annotated as a float,
<arigato> so the code in OLD will first convert the value to a float,
<arigato> and call write_float_to_mem(1237E10)
<arigato> which is nonsense
<antocuni> yes, but the function NEW (which is, more concretely: llmodel.AbstractLLCPU.bh_gc_store_indexed_f) should be annotated even if the codewriter never sees the llop
<arigato> are you sure?
<antocuni> this is what I'd expect, and it's the part I don't understand :)
<arigato> parts of the jit are not annotated at all if some jitcodes don't show up
<arigato> notably, pyjitpl's and blackhole's implementations of the corresponding jitcodes
<antocuni> uhm, probably this explains it
<antocuni> who annotates e.g. executor.do_gc_store_indexed?
<arigato> who knows? maybe only the blackhole's version of the jitcode
<antocuni> we reached the point in which our JIT is self-conscious and decided what to annotate by itself :)
amaury has joined #pypy
<arigato> obviously :-)
<arigato> note that you had a failing test on 32-bit
<antocuni> yes, that's how I found the bug
<arigato> I *think* that there are hacks for non-translated runs where mixture of longlong-for-storage and floats are detected
<arigato> so it's basically your fault for not running tests :-)
<antocuni> well no, I *did* run the tests (on buildbot because I didn't have a 32 bit chroot locally)
<antocuni> my fault was to dismiss the failure as unrelated
<antocuni> because the bug showed up also in places in which the llop was not called at all
<antocuni> anyway, I suppose I could add some _annenforceargs_ or similar to write_float_at_mem, to catch this kind of error earlier next time
<arigato> ah righ
<arigato> yes, possibly
<arigato> normally, a function that receives a float unexpectedly will crash because of what it tries to do with it
<arigato> so yes, if write_float_to_mem is an exception, better make it crash
<antocuni> yes
<antocuni> also, I did not expect it is possible to union r_longlong and float
<antocuni> but probably it is too late to forbid it
<cfbolz> we should try to forbid it anyway
<cfbolz> it's clearly nonsense
tilgovi has quit [Ping timeout: 272 seconds]
oberstet has joined #pypy
inad922 has joined #pypy
<arigato> it's forbidden before translation
<arigato> in the case of longlong-for-storage
<arigato> after translation, I think I tried but gave up
<cfbolz> ah :-(
inad922 has quit [Ping timeout: 268 seconds]
<kenaan> antocuni faster-rstruct-2 65b6be2c1b4f /rpython/jit/backend/llsupport/llmodel.py: enforce the annotation of write_float_at_mem: the problem fixed by 0e91701e92f0 was that we mixed a r_f...
DragonSA has joined #pypy
DragonSA has joined #pypy
DragonSA has quit [Changing host]
amaury has quit [Ping timeout: 246 seconds]
antocuni has quit [Ping timeout: 268 seconds]
jacob22_ has quit [Quit: Konversation terminated!]
mihaid has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
arigato has quit [Read error: Connection reset by peer]
jkbbwr has quit [Remote host closed the connection]
jkbbwr has joined #pypy
jamadden has joined #pypy
amaury has joined #pypy
arigato has joined #pypy
vkirilichev has joined #pypy
vkirilichev has quit [Remote host closed the connection]
vkirilichev has joined #pypy
mattip has joined #pypy
amaury has quit [Ping timeout: 260 seconds]
antocuni has joined #pypy
mihaid has joined #pypy
<mihaid> hello, I have another question regarding profile-guided optimizations. Allowing the profopt for other targets will also change the goal directory, and this is a problem because the resulting pypy-c binary needs to be in the goal directory in order to be trained. Should I add the goal directory as a parameter, or should I assume that the translation
<mihaid> is run from the goal directory and get an absolute path to it?
<arigato> should we issue a 5.7.2 release? because of b2569cf1bd55, which I suspect is causing other issues too that we see
<cfbolz> mihaid: you should be able to find the correct output filename as an absolute path, somehow
<cfbolz> mihaid: but anyway, I think the makefile works like this: it builds everything in /tmp/usession*/testing_1/ and then later the binary is copied
<cfbolz> the Makefile can work completely in the testing_1 directory
<antocuni> the only drawback is that if you run pypy-c from the testing_1 directory, it might not be able to import the stdlib correctly, I think
<cfbolz> it will use the compiled in path
<arigato> which should always be the right one in this precise case and likely no other
<cfbolz> yes
<mihaid> I tried running the pypy-c binary in the /tmp/.../testing and I got an import error, unfortunately
<arigato> it prints big warnings, but it should work
<arigato> ah
<arigato> let's fix that then
<mihaid> how can this be fixed?
Tiberium has joined #pypy
<arigato> read carefully the CPython doc about "sys.prefix"
<arigato> so far we don't set it at all if we don't find the library paths
<arigato> maybe we should
<arigato> or, the alternative is to fix sysconfig.py
<arigato> to not crash if sys.prefix is not set
<arigato> ...no, this sysconfig.py is straight from CPython and it reads sys.prefix there too
<arigato> so the problem is really that nobody is expecting sys.prefix to be undefined, also in CPython
<kenaan> ltratt default 008c6f3a2f8f /rpython/translator/platform/: Build main binaries on OpenBSD with wxallowed. On RPython, JIT compilers that need to read and write to the same ...
<kenaan> ltratt default cfe544728497 /: Merge default
<cfbolz> arigato: why is prefix only set if you run it from goal?
<arigato> I can't tell
vkirilichev has quit [Remote host closed the connection]
dcrosta has quit [Quit: Textual IRC Client: www.textualapp.com]
vkirilichev has joined #pypy
<cfbolz> :-(
vkirilic_ has joined #pypy
vkirilichev has quit [Read error: Connection reset by peer]
<kenaan> cfbolz default 1d471fdc1963 /rpython/annotator/: fix issue #1877: replace bare Exceptions with AnnotatorError in the annotator
vkirilic_ has quit [Ping timeout: 246 seconds]
tbodt has joined #pypy
<mihaid> so I don't know how to fix the import site issue, but the argument that I added would only need to be used when profopt-ing something other than PyPy. Basically, right now if you need to translate with profopt for PyPy, you just specify --profopt.
<cfbolz> mihaid: it's a bit weird that the Makefile hard-codes the goal directory, though. in theory it would be cool if all the building happened in testing_1/
<mihaid> yes, but I see no other way out. Maybe I can fix this in another issue / pull request?
<cfbolz> agree that it's annoying to have the pgo work blocked by the prefix issue
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<kenaan> Alecsandru Patrascu sockopt_zero 74c6874f803f /pypy/module/_socket/: added the possibility to have zero sized buffer length
<kenaan> Alecsandru Patrascu sockopt_zero 4c2ff9e6a959 /pypy/module/_socket/: proper fix after review
<kenaan> Alecsandru Patrascu sockopt_zero f848cd62800a /pypy/module/_socket/interp_socket.py: shorten code path
<kenaan> cfbolz default 965ffd1ed1de /pypy/module/_socket/: Merged in palecsandru/pypy2_sockopt_zero/sockopt_zero (pull request #548) Passing a buffersize of 0 to socket.get...
vkirilichev has joined #pypy
<kenaan> rlamy py3.5 619bc1ea9e6d /pypy/module/cpyext/test/test_listobject.py: fix test
tbodt has joined #pypy
amaury has joined #pypy
jacob22_ has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
vkirilic_ has joined #pypy
<mihaid> I think I have found a quick fix. The pypy-c binary is looking for lib-python and lib_pypy to use pypy_find_stdlib in the current directory. If I copy them in /tmp/.../testing then I have no more import site errors.
<mihaid> is this ok?
Remi_M has quit [Quit: See you!]
vkirilichev has quit [Ping timeout: 260 seconds]
<kenaan> rlamy py3.5 e8a4ea639258 /pypy/module/cpyext/test/issue2482.c: Adapt test to py3
cstratak has quit [Quit: Leaving]
cstratak has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<cfbolz> mihaid: bit heavy-handed. arigato, any other clue?
<arigato> I'd say, fix sys.prefix
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
* arigato looks more carefully at clean-exported-state, the merging of that branch falls inside the (large) window for a rare unrolling crash
<mihaid> okay, I will look into it
mihaid has quit [Quit: http://www.kiwiirc.com/ - A hand crafted IRC client]
<cfbolz> arigato: good luck
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<arigato> cfbolz: thanks, but if I'm right I'll take the easy way out
<arigato> I'm translating now a pypy with this merge reverted
<arigato> if it fixes the problem, I'll really revert it
<arigato> and then try to contact Spenser Bauman about it
tbodt has joined #pypy
<cfbolz> arigato: I think spenser has *more* weird bugs in unrolling
<cfbolz> but somebody would have to find the motivation to do a deep dive
<arigato> if PR #490 seems to add more bugs than it removes, I'll remove it
<cfbolz> arigato: that's probably not an accurate summary. it rejiggles the world. but yes, I see your point too
<cfbolz> sbauman: ^^ in case you are around
<sbauman> That branch fixed some number of crashed for Pycket. It could have introduced others though.
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
realitix has quit [Ping timeout: 268 seconds]
<arigato> sbauman: it adds (rarely) in the 2nd loop iteration a variable that comes from nowhere, that is apparently only used inside guard fail_args
<arigato> admittedly we could also work around that problem by detecting the situation and raising InvalidLoop then
<sbauman> arigato: That is odd. It should only delete information. It could be losing information used to eliminate those fail_args I suppose.
<arigato> let's wait another hour before I blame your PR
oberstet has quit [Ping timeout: 255 seconds]
black_ant has quit [Quit: ZNC - http://znc.in]
marr has quit [Ping timeout: 260 seconds]
tbodt has joined #pypy
tbodt has quit [Client Quit]
yuyichao_ has quit [Ping timeout: 255 seconds]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
oberstet has joined #pypy
<arigato> there are other changes to unroll.py in that window
yuyichao_ has joined #pypy
* mattip trying to understand hashing
<cfbolz> hi matti :-)
<mattip> hi
<kenaan> mattip cpyext-unhashable 21ce0ac7d0de /pypy/: add passing tests
<mattip> there are multiple ways to get to the compute_hash() function, usually we go through DescrOperation.hash in descroperation.py
<mattip> but cpyext wraps all the lookups, so it goes directly to the descr_hash function without going through DescrOperation
<cfbolz> mattip: DescrOperation.hash is the same as space.hash, I think?
<cfbolz> yes, seems so
<cfbolz> so I guess basically all hashing outside of cpyext goes through that
<mattip> right, but there is special logic in descroperation.py to handle hash(obj) == -1, which is an error condition
<cfbolz> this stuff again :-(
oberstet2 has joined #pypy
<mattip> so for w_str('') I needed to add it directly to the descr_hash
<mattip> which seems OK, but why doesn;t w_unicode('') need it as well?
<mattip> ... and then some PyObject objects seem unhashable while others are, I don't see the logic
<mattip> so far in CPython
<cfbolz> mattip: it's probably just historical mess, not necessarily any logic
<cfbolz> we used to not special-case -1, but then we added special casing
<cfbolz> maybe it was just forgotten in some places
oberstet has quit [Ping timeout: 260 seconds]
<mattip> yeah, ok, so anyhow c2f5f5816876 fixes another numpy failure, where we used to return -1 from hash(numpy._string(''))
<mattip> arigato: should we do 5.7.2 now or 5.8 in a few weeks? Any pressing conferences coming up?
<arigato> sbauman: yes, it's the merge of PR #490 that causes the problem
<mattip> cfbolz: I'm not sure about the part of 552622605578 where I set w_dict["__hash__"] = None
<mattip> if there is no tp_hash in the original PyType_Object
<mattip> it makes some tests pass with -A, but then again, it may break other stuff
<cfbolz> mattip: why is that needed? it feels like just not having a __hash__ should be sufficient?
<mattip> cfbolz: since then the lookup will go to the parent's hash via mro, not simply fail
<cfbolz> mattip: and the parent has a hash (eg object)?
<mattip> right
<cfbolz> I see
<cfbolz> mattip: yes, I think the = None is right
<cfbolz> that's how lists do it too
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<mattip> ok, I will try a translation of this mess on cython/numpy/pandas and see what broke :)
<mattip> thanks
tbodt has joined #pypy
<arigato> sbauman: the problem is https://bitbucket.org/pypy/pypy/issues/2534/jit-crash and it has a nice reproducer example for pypy
<mattip> it's all so emperical
<arigato> mattip: I suppose that if we back out PR #490 and don't see a different crash, we should release 5.7.2 now, because we got several reports about that
<mattip> ok
<mattip> so we need to cherry pick the commits that are bug fixes into the release branch
<mattip> it might make sense to just go ahead with 5.8 and get the cpyext improvements (latest numpy requires them) and the PyBuffer improvements
<arigato> either that if you feel like it, otherwise we make 5.8 now, yes
<antocuni> I'd like to merge faster-rstruct-2 before the next release, if it's possible :)
<arigato> depends on you :-) but I don't expect the next release to occur tomorrow
<antocuni> perfect
<arigato> mattip: are you still ok to be release manager? you can opt out if you don't want it any more
<antocuni> the "now" in "we make 5.8 now" worried me :)
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<kenaan> rlamy py3.5 48119cb4c0db /pypy/objspace/std/: Add validation of __getnewargs__()/__getnewargs_ex__() return values
<mattip> arigato: if someone else _wants_ to take it, that's fine
<mattip> especially someone newish
<mattip> antocuni: can we say around two weeks?
<antocuni> sure, that's perfect for me. I really hope to have finished it before that date :)
<arigato> cool :-)
vkirilic_ has quit [Remote host closed the connection]
kanaka has joined #pypy
kanaka has quit [Changing host]
kanaka has joined #pypy
vkirilichev has joined #pypy
vkirilichev has quit [Ping timeout: 240 seconds]
tbodt has joined #pypy
tbodt has quit [Remote host closed the connection]
tbodt has joined #pypy
<kenaan> arigo default 1a06512dea4b /rpython/: Backout PR #490 (ab8a0b092d2d + follow-up bd7b1902c4df). See https://bitbucket.org/pypy/pypy/issues/2534. It give...
james has joined #pypy
james is now known as Guest80440
Guest80440 is now known as jamescampbell
lritter_ has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Client Quit]
jamescampbell has quit [Remote host closed the connection]
pilne has joined #pypy
arigato has quit [Quit: Leaving]
<kenaan> antocuni faster-rstruct-2 356dacc477eb /rpython/translator/backendopt/: backout 85d3ab6fe80b, which was broken, and start to rewrite the logic in a more generic way. Also, wri...
<kenaan> antocuni faster-rstruct-2 d5be1a60c5a3 /rpython/translator/backendopt/: add support & test for gc_load_indexed of a list of chars
<kenaan> antocuni faster-rstruct-2 1e87cd8c5f6c /rpython/translator/backendopt/: implement the analyzer for gc_store_indexed
<kenaan> antocuni faster-rstruct-2 64c71c919e55 /rpython/translator/backendopt/test/test_writeanalyze.py: add a passing test for gc_store_indexed on a list of chars
pf_moore has joined #pypy
antocuni has quit [Ping timeout: 260 seconds]
tbodt has joined #pypy
tbodt has quit [Client Quit]
tbodt has joined #pypy
fryguybob has quit [Ping timeout: 240 seconds]
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
fryguybob has joined #pypy
tbodt has joined #pypy
jamescampbell has joined #pypy
amaury has quit [Ping timeout: 272 seconds]
lritter_ has quit [Remote host closed the connection]
inhahe_ has quit [Read error: Connection reset by peer]
inhahe_ has joined #pypy
lritter has joined #pypy
tbodt has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tbodt has joined #pypy
tbodt has quit [Client Quit]
<kenaan> mattip cpyext-unhashable ff2ce5a214bd /pypy/module/cpyext/: backed out 552622605578, breaks too many things
Tiberium has quit [Remote host closed the connection]
jamesaxl has quit [Read error: Connection reset by peer]
aboudreault has quit [Excess Flood]
jamesaxl has joined #pypy
aboudreault has joined #pypy
amaury has joined #pypy
jamescampbell has quit [Quit: Leaving...]
cstratak has quit [Quit: Leaving]
tbodt has joined #pypy
commandoline has joined #pypy
marr has joined #pypy
lritter has quit [Read error: Connection reset by peer]
lritter has joined #pypy
aboudreault has quit [Excess Flood]
jamescampbell has joined #pypy
arigato has joined #pypy
aboudreault has joined #pypy
nimaje1 has joined #pypy
nimaje1 is now known as nimaje
nimaje has quit [Killed (kornbluth.freenode.net (Nickname regained by services))]
tbodt has quit [Read error: Connection reset by peer]
DragonSA has quit [Quit: Konversation terminated!]
jamescampbell has quit [Remote host closed the connection]
arigato has quit [Quit: Leaving]
jamescampbell has joined #pypy
vkirilichev has joined #pypy
jamescampbell has quit [Remote host closed the connection]
ronan has quit [Ping timeout: 246 seconds]
jamescampbell has joined #pypy
jamescampbell has quit [Remote host closed the connection]
jamescampbell has joined #pypy
aboudreault has quit [Excess Flood]
aboudreault has joined #pypy
jamescampbell has quit [Ping timeout: 240 seconds]
ronan has joined #pypy
vkirilichev has quit [Remote host closed the connection]
vkirilichev has joined #pypy
vkirilichev has quit [Read error: Connection reset by peer]
vkirilichev has joined #pypy
aboudreault has quit [Excess Flood]
aboudreault has joined #pypy
nimaje1 has joined #pypy
nimaje1 is now known as nimaje
antocuni has joined #pypy
<kenaan> antocuni faster-rstruct-2 616c5957da8d /rpython/jit/backend/llsupport/llmodel.py: @enforceargs does not play well with @specialize, use _annenforceargs_ directly
<kenaan> antocuni faster-rstruct-2 a7c92808278f /rpython/translator/backendopt/writeanalyze.py: typo
vkirilichev has quit [Remote host closed the connection]
vkirilichev has joined #pypy
<kenaan> antocuni faster-rstruct-2 ad1105da8697 /rpython/translator/backendopt/writeanalyze.py: fix test_gc_load_indexed_str
jamescampbell has joined #pypy
vkirilichev has quit [Ping timeout: 268 seconds]
<kenaan> antocuni faster-rstruct-2 4d0500d6eced /: hg merge default
jamescampbell has quit [Ping timeout: 268 seconds]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-s390x/builds/549 [antocuni: force build, faster-rstruct-2]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-64/builds/4631 [antocuni: force build, faster-rstruct-2]
<bbot2> Started: http://buildbot.pypy.org/builders/own-linux-x86-64/builds/5954 [antocuni: force build, faster-rstruct-2]
<bbot2> Started: http://buildbot.pypy.org/builders/own-linux-x86-32/builds/5201 [antocuni: force build, faster-rstruct-2]
<bbot2> Started: http://buildbot.pypy.org/builders/pypy-c-jit-linux-x86-32/builds/3963 [antocuni: force build, faster-rstruct-2]
aboudreault has quit [Excess Flood]
aboudreault has joined #pypy