verne.freenode.net changed the topic of #mlpack to: http://www.mlpack.org/ -- We don't respond instantly... but we will respond. Give it a few minutes. Or hours. -- Channel logs: http://www.mlpack.org/irc/
apir8181 has joined #mlpack
trapz has quit [Quit: trapz]
sumedhghaisas has quit [Ping timeout: 265 seconds]
< lezorich>
naywhayare: yes! I wanted to run an individual test suite
< lezorich>
thanks!
< naywhayare>
sure
< naywhayare>
that support is provided through the Boost Unit Test Framework, not anything specific in the mlpack code
< naywhayare>
so there may be more cool things the Boost UTF does that I don't know about :)
stephentu has joined #mlpack
< lezorich>
that's ok :) I was trying to run the test suite by it's filename, e.g: mlpack_test -t emst_test
< lezorich>
but now it worked :)
< naywhayare>
hm, maybe it would be a good idea to have some kind of guideline on how to name a test suite with respect to the filename
< naywhayare>
but that would probably be hard to enforce
< stephentu>
naywhayare: i think we should add an option somehow to mlpack_test which lists all possible test cases
< stephentu>
(this woudl also help us to parallelize the test suite)
< naywhayare>
yeah, I agree
< naywhayare>
this may require a contribution to the Boost UTF project (I don't know if we can easily build something to do that nicely downstream)
< naywhayare>
it would be easy to write a simple script that just listed all the test suites and test cases, but I think you're suggesting something more like 'mlpack_test --list-all-test-suites' or --list-all-test-cases or something
< stephentu>
yes
< stephentu>
i was going to write teh script
< stephentu>
and just regex out the BOOST_TEST_BLAH(...) part
< stephentu>
but that seemed so dirty
< stephentu>
that i didnt
< naywhayare>
yeah; if you were going to, probably the way to do it is in CMake (which actually makes it seem even more horrifying)
< stephentu>
ya that sounds even worse
< naywhayare>
so then you can do 'make list-test-suites' or 'make list-test-cases' and it'll work on whatever platforms CMake works on
< naywhayare>
yeah, CMake scripts for stuff like that are nightmare fuel
< stephentu>
but if someobdy does that, then i can flip on the travis CI builds
< stephentu>
actually fornow i could just exclude the test suite and have travis test compliation
< stephentu>
but thats not really useful i assume our PRs at least compile :)
< stephentu>
actually i take it back, since I set up the image to compile on old versions of arma
< stephentu>
it is still useful
< naywhayare>
hm, so the issue with Travis is that the tests currently take too long?
< stephentu>
ya
< naywhayare>
and you want to be able to ignore certain tests?
< stephentu>
you only get 50 min
< naywhayare>
ok
< stephentu>
either we make it faster
< stephentu>
or i exclude certain tests
< stephentu>
but i also want ot parallelize the tests
< stephentu>
so it doesnt take so long
< naywhayare>
do we get multiple cores with a travis build?
< stephentu>
i htink so? i use make -j4 for builds and it seems to make it go faster
< naywhayare>
seems reasonable enough
< stephentu>
but parallelize would hlep us to squeeze more tests in the allocated time
< naywhayare>
definitely
< naywhayare>
so, the slow tests are something I should fix. let me open some tickets so I remember to do that soon
< naywhayare>
(or, I guess anyone can fix them, but in many cases I know what the issue is and why it's slow)
< stephentu>
anyways sorry i've been busy lately so not able to contribute much code
< zoq>
"Travis CI VMs run on 1.5 virtual cores. This is not exactly a concurrency, which allows to parallelize a lot, but it can give a nice speedup depending on your use case."
< naywhayare>
that links to a document that suggests the 1.5 number as out of date, and that there are now 2 cores per build (as of December 2014)
< stephentu>
nice
< stephentu>
and 4GB mem
< naywhayare>
so this is nice because it means that figuring out what 1.5 virtual cores means isn't a necessity, but I'm still baffled by what that term means
prakhar2511 has joined #mlpack
< stephentu>
when you do cat /proc/cpuinfo
< stephentu>
it cuts out int he middle
< stephentu>
of the 2nd cpu
< zoq>
:)
< naywhayare>
haha
sumedhghaisas has quit [Ping timeout: 252 seconds]
kshitijk has joined #mlpack
kshitijk has quit [Ping timeout: 246 seconds]
prakhar2511 has quit [Ping timeout: 246 seconds]
curiousguy13_ has quit [Ping timeout: 264 seconds]