--- Log opened Sun Sep 18 00:00:44 2011 | ||
-!- blackburn [~blackburn@188.168.2.76] has left #shogun [] | 01:46 | |
-!- blackburn [~blackburn@188.168.5.156] has joined #shogun | 09:26 | |
CIA-3 | shogun: Sergey Lisitsyn master * r828a0f4 / (11 files): Various structure improvements for DR preprocessors - http://git.io/xKcPyg | 11:55 |
---|---|---|
CIA-3 | shogun: Sergey Lisitsyn master * r1f6bf22 / src/shogun/preprocessor/DimensionReductionPreprocessor.h : Added doc for detect_dim routine - http://git.io/sok0Vg | 12:01 |
blackburn | sonney2k: superlu is such a pain in ass! | 19:12 |
@sonney2k | blackburn, :) | 19:25 |
blackburn | they have no doc | 19:25 |
blackburn | even less than we have | 19:25 |
blackburn | sonney2k: how to turn on debug prints? | 19:44 |
@sonney2k | obj.io.set_loglevel(MSG_DEBUG) | 19:46 |
blackburn | thanks | 19:53 |
blackburn | sonney2k: btw my parallelization have impact on high-dimensional data | 20:01 |
blackburn | yesterday I've played with images a little | 20:01 |
blackburn | and 2 threads makes it 2 times faster | 20:01 |
@sonney2k | very good | 20:28 |
-!- shogun-buildbot [~shogun-bu@7nn.de] has joined #shogun | 20:42 | |
shogun-buildbot | build #232 of cmdline_static is complete: Success [build successful] Build details are at http://www.shogun-toolbox.org/buildbot/builders/cmdline_static/builds/232 | 20:47 |
@sonney2k | blackburn, btw I have managed to move the templated stuff from the .h to the .cpp files | 21:36 |
blackburn | sonney2k: and how it is? | 21:36 |
@sonney2k | blackburn, I didn't really do any benchmarks - did you notice any speed diff? | 21:36 |
blackburn | sonney2k: not really | 21:36 |
@sonney2k | blackburn, yeah - I am not really sure if it will change anything in the swig wrapper | 21:37 |
@sonney2k | I mean in the end it will only look at *.h files so it must be slightly faster | 21:37 |
@sonney2k | but the generated wrapper should be as big as it used to be | 21:37 |
blackburn | I see | 21:37 |
@sonney2k | blackburn, http://www.parashift.com/c++-faq-lite/templates.html#faq-35.15 | 21:38 |
@sonney2k | however, I think it is much more readable this way | 21:38 |
@sonney2k | so I will continue this effort for other templated classes | 21:38 |
@sonney2k | blackburn, just look at SimpleFeatures.h now | 21:39 |
blackburn | sonney2k: I've been wondering if we can use some swig custom things | 21:39 |
@sonney2k | really nice now :) | 21:39 |
@sonney2k | swig custom? | 21:39 |
blackburn | I mean there are a lot of options | 21:39 |
blackburn | may be we can use some of them | 21:39 |
@sonney2k | options for what? | 21:39 |
blackburn | don't know, to ignore unnecessary things, improve something | 21:40 |
@sonney2k | blackburn, I think we can can currently only improve compile speed by making certain types configurable | 21:40 |
@sonney2k | I mean we could have a flag ./configure --supported-data-types=int8,uint8,...,floatmax_t | 21:41 |
blackburn | ah | 21:41 |
@sonney2k | and then it will only generate wrappers for the ones in this list | 21:41 |
blackburn | I see | 21:41 |
@sonney2k | this way we could just create say 3 of these by default - it doesn't help overall though | 21:42 |
@sonney2k | blackburn, the other thing that could help a lot is removing all the Array.h Array2.h Array3.h and replacing it with SGVector,SGMatrix, SGNDArray | 21:50 |
blackburn | sonney2k: yeah I really don't understand why to use ArrayX | 21:51 |
@sonney2k | blackburn, well it is used - so we just need to transfer some of its functions (like resize support, get/set functions w/ error checks) | 21:52 |
blackburn | aha | 21:52 |
@sonney2k | and then it can be replaced... | 21:56 |
@sonney2k | blackburn, heh the pirate party got 9% of the votes in the Berlin elections | 21:57 |
blackburn | sonney2k: hehe | 22:24 |
blackburn | you have truly elections.. | 22:24 |
--- Log closed Mon Sep 19 00:00:48 2011 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!