convore.json/groups/scipy-2011/talk-bringing-parallel-performance-to-python-with-domain-specific-selective-embedded-just-in-time-specialization/messages.json

[{"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310570340.5463309, "message": "my google-fu is insufficient", "group_id": 11000, "id": 1629302}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571621.6902709, "message": "nvm?", "group_id": 11000, "id": 1629491}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571132.660661, "message": "ditto", "group_id": 11000, "id": 1629412}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571206.794364, "message": "I mean, not Intel and the hardware guys, who have a vested interest in you using their specific low-level library (e.g. ArBB), but basically everybody else at the higher level", "group_id": 11000, "id": 1629417}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571414.005656, "message": "I like the motivation....  This is exactly what is needed.", "group_id": 11000, "id": 1629441}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571414.339669, "message": "Yeah.   It's definitely a needed Thing", "group_id": 11000, "id": 1629443}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571414.566725, "message": "It doesn't sound crazy Shoaib.  On the contrary....", "group_id": 11000, "id": 1629444}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310570192.0531659, "message": "Interesting talk about constructing assembly in Python.  Have to check out SimASM", "group_id": 11000, "id": 1629258}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571817.4219379, "message": "Anyone else used CodePy?", "group_id": 11000, "id": 1629528}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571839.9285669, "message": "http://mathema.tician.de/software/codepy", "group_id": 11000, "id": 1629532}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310570336.2312601, "message": "Wait this is the talk I'm in https://convore.com/scipy-2011/talk-a-software-framework-in-python-for-generating-optimal-isogeometric-kernels-on-the-powerpc-450/", "group_id": 11000, "id": 1629301}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310570355.8036549, "message": "heh", "group_id": 11000, "id": 1629306}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571448.9062669, "message": "I like the motivation....  This is exactly what is needed.", "group_id": 11000, "id": 1629450}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571167.3783989, "message": "@teoliphant it's absolutely hilarious to me how basically everybody is thinking along the exact same lines in this space", "group_id": 11000, "id": 1629413}, {"user_id": 33183, "stars": [], "topic_id": 41822, "date_created": 1310571509.959105, "message": "Does anyone have a link on the paper associated with this talk?", "group_id": 11000, "id": 1629467}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571626.3143871, "message": "and their AST just relies on those as markers", "group_id": 11000, "id": 1629493}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571629.0404999, "message": "nvm=\"nevermind\"", "group_id": 11000, "id": 1629495}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310570303.7514019, "message": "Do you have a link for SimASM?", "group_id": 11000, "id": 1629289}, {"user_id": 33559, "stars": [], "topic_id": 41822, "date_created": 1310571589.9861691, "message": "Looks like the parlab webpage is down? http://parlab.eecs.berkeley.edu", "group_id": 11000, "id": 1629481}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571811.970026, "message": "It looks like he generates multiple output codes.", "group_id": 11000, "id": 1629525}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571600.131603, "message": "He mentioned they handle a \"subset of Python\"... but I didn't really see how they are enforcing anything at the syntax level", "group_id": 11000, "id": 1629483}, {"user_id": 37933, "stars": [], "topic_id": 41822, "date_created": 1310571833.334028, "message": "andreas doesn't even use it anymore, i think :)", "group_id": 11000, "id": 1629530}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571615.724154, "message": "oh nvm - you always have to use those iterators", "group_id": 11000, "id": 1629489}, {"user_id": 33559, "stars": [], "topic_id": 41822, "date_created": 1310571705.093915, "message": "I haven't received the paper for the proceedings yet, so assume he's talking about another", "group_id": 11000, "id": 1629511}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571770.27636, "message": "ok, so I'm wrong :)", "group_id": 11000, "id": 1629516}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571079.0521381, "message": "Now, I'm listening to this talk by Shoaib Kamil.", "group_id": 11000, "id": 1629406}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571425.7461121, "message": "heh", "group_id": 11000, "id": 1629445}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571388.5054941, "message": "Sometimes I wonder about the feasibility of layering a purely functional layer on top of Python.  Projects like this, BlockCanvas, numexpr, fast_vectorize, etc. would all benefit", "group_id": 11000, "id": 1629435}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571613.8885491, "message": "Good question to ask at the end...", "group_id": 11000, "id": 1629488}, {"user_id": 37933, "stars": [], "topic_id": 41822, "date_created": 1310571873.16432, "message": "just templating engines", "group_id": 11000, "id": 1629536}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571414.1663809, "message": "I would add \"FPGA\" to the list of possible hardware renderings.", "group_id": 11000, "id": 1629442}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571987.768095, "message": "Fortran 90 got to about 300x faster than the pure Python approach in my case.", "group_id": 11000, "id": 1629559}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310572210.5331969, "message": "I'd like to see a project that by-passes the linker as well.", "group_id": 11000, "id": 1629610}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571749.0922301, "message": "I'm not sure.  He hasn't specified what happens to the \"AST in C\"", "group_id": 11000, "id": 1629514}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571823.6787291, "message": "nope, looks slick", "group_id": 11000, "id": 1629529}, {"user_id": 33559, "stars": [], "topic_id": 41822, "date_created": 1310571943.462338, "message": "Paper is here: https://github.com/shoaibkamil/scipy_proceedings/blob/master/papers/shoaib_kamil/shoaib_kamil.rst (Need to clone and build PDF to get the pictures)", "group_id": 11000, "id": 1629549}, {"user_id": 8649, "stars": [], "topic_id": 41822, "date_created": 1310571716.696557, "message": "I guess they are doing AST transformation to avoid having to run an external compiler?  (like, instead of doing code gen a la weave)", "group_id": 11000, "id": 1629512}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571860.1135011, "message": "Andreas is still definitely doing cod-gen.   If he doesn't use CodePy, what does he use/", "group_id": 11000, "id": 1629533}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310571913.2347541, "message": "Should add his laplacian example to my speeding up Python example set.", "group_id": 11000, "id": 1629546}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310572079.1849749, "message": "http://technicaldiscovery.blogspot.com/", "group_id": 11000, "id": 1629575}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310572104.7847891, "message": "I didn't show the \"Fortran-only\" case which was about 2x faster than the outer-loop in Python approach.", "group_id": 11000, "id": 1629584}, {"user_id": 11207, "stars": [{"date_created": 1310572197.417022, "user_id": 8649}], "topic_id": 41822, "date_created": 1310572169.3939209, "message": "Next-gen SciPy/NumPy should be based on stuff like this.", "group_id": 11000, "id": 1629598}, {"user_id": 11207, "stars": [], "topic_id": 41822, "date_created": 1310572177.340925, "message": "How do we get there?", "group_id": 11000, "id": 1629600}, {"user_id": 37933, "stars": [], "topic_id": 41822, "date_created": 1310677548.9071081, "message": "for those interested, here's what Andreas had to say about codepy: \"i was recommending against ASTs for code generation. cgen/AST generation is overkill for most things but just right for really complicated stuff\" -- Andreas Kloeckner", "group_id": 11000, "id": 1640400}, {"user_id": 38068, "stars": [], "topic_id": 41822, "date_created": 1310681125.8603151, "message": "For code generation templates, what about using a parser to create the AST, and then doing template replacement on the AST.  This seems safer than using purely text-based templating.", "group_id": 11000, "id": 1640612}]