--- Log opened Sat May 09 00:00:17 2015 | ||
-!- txomon|home [~txomon@unaffiliated/txomon] has quit [Ping timeout: 265 seconds] | 02:35 | |
-!- txomon|home [~txomon@unaffiliated/txomon] has joined #shogun | 02:49 | |
-!- txomon|home [~txomon@unaffiliated/txomon] has quit [Ping timeout: 240 seconds] | 03:12 | |
-!- txomon|home [~txomon@unaffiliated/txomon] has joined #shogun | 12:03 | |
-!- txomon|home [~txomon@unaffiliated/txomon] has quit [Ping timeout: 245 seconds] | 12:09 | |
-!- lambday [6a33f00b@gateway/web/freenode/ip.106.51.240.11] has joined #shogun | 13:29 | |
-!- mode/#shogun [+o lambday] by ChanServ | 13:29 | |
@lambday | lisitsyn: hey! | 13:29 |
---|---|---|
lisitsyn | lambday: hey | 13:30 |
@lambday | lisitsyn: how are things? | 13:30 |
lisitsyn | ah fine what about you? | 13:30 |
@lambday | lisitsyn: just got your comment on that | 13:30 |
@lambday | lisitsyn: didn't get what you meant by "Actually you don't need rewriting operators for basic types - just operators of Parameter and float/int so that any multiplication/addition/etc goes inside your object that represents expression." | 13:31 |
lisitsyn | ok let me explain | 13:31 |
lisitsyn | you have say | 13:31 |
lisitsyn | 2*Parameter(blabla) | 13:31 |
@lambday | ok | 13:31 |
lisitsyn | you need operator*(float, Parameter) | 13:31 |
@lambday | yes | 13:32 |
lisitsyn | and operator*(Parameter, Parameter) | 13:32 |
@lambday | but say I have a non-parametric thing in my expression.. | 13:32 |
lisitsyn | you don't need custom operator*(float, float) | 13:32 |
@lambday | like -2*sin(98235) | 13:32 |
lisitsyn | that sounds like a constant | 13:33 |
lisitsyn | that can be precomputed | 13:33 |
@lambday | hmm yes that can be done | 13:33 |
@lambday | ok I got it | 13:33 |
lisitsyn | lambday: I wouldn't bother that much :) | 13:37 |
lisitsyn | format is good already I think | 13:37 |
@lambday | lisitsyn: yeah this one looks neat as well | 13:37 |
@lambday | lisitsyn: sending PR with that one then | 13:37 |
@lambday | lisitsyn: one interesting thing is that, I tried viennacl/opencl with both cpu and gpu | 13:39 |
@lambday | lisitsyn: and in my machine it turned out that cpu is even faster | 13:39 |
lisitsyn | notebook? | 13:39 |
@lambday | my graphics card is intel ivybridge integrated one | 13:39 |
lisitsyn | ah | 13:39 |
lisitsyn | could happen I guess | 13:40 |
@lambday | Intel(R) HD Graphics IvyBridge M GT2 | 13:40 |
@lambday | ya but why is that? | 13:41 |
@lambday | lisitsyn: for the cpu one I am using intel's sdk for opencl... and for the gpu I am using beignet | 13:41 |
lisitsyn | I don't know, it could be driver or something | 13:42 |
lisitsyn | I wouldn't expect good performance on such card | 13:42 |
lisitsyn | nowhere near to these things like tesla | 13:42 |
@wiking | booon | 13:59 |
-!- lambday [6a33f00b@gateway/web/freenode/ip.106.51.240.11] has quit [Ping timeout: 246 seconds] | 14:32 | |
-!- HeikoS [~heiko@0544f0d5.skybroadband.com] has joined #shogun | 14:50 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 14:50 | |
-!- nszceta [~nszceta@5ED062E5.cm-7-1b.dynamic.ziggo.nl] has joined #shogun | 15:09 | |
@wiking | yozon | 15:14 |
@wiking | HeikoS: so basically we made another step towards nightly packages | 15:14 |
@wiking | last week | 15:14 |
@HeikoS | wiking: yoyo | 15:14 |
@wiking | but dunno what's the status of besser about getting some more stuff running | 15:14 |
@HeikoS | wiking: whats the step? | 15:14 |
@wiking | but i'll see it on monday the latest | 15:14 |
@wiking | HeikoS: well now finally we have only config.h + swig_config.h | 15:15 |
@wiking | no more -D compiler flags | 15:15 |
@HeikoS | not sure whether I understand what that means | 15:15 |
@wiking | which will allow us basically to build a modular interface without building actually the c++ library | 15:15 |
@wiking | if you have a preexisting shogun installation | 15:15 |
@wiking | that is required for being able to build | 15:15 |
@wiking | modular interfaces standalone | 15:16 |
@wiking | without building the library itself | 15:16 |
@wiking | well previously shogun defined all sorts of compile flags | 15:16 |
@HeikoS | I see | 15:16 |
@HeikoS | thats great then | 15:16 |
@wiking | -DHAVE_WHATEVER -DLALLAA | 15:16 |
@wiking | now we dont have this anymore | 15:16 |
@wiking | we purely rely on config.h | 15:16 |
@wiking | this has a good sideeffect of ccache finally being really effective | 15:17 |
@wiking | so now if u use ccache the recompilation of the library will be really fast | 15:17 |
@wiking | because the problem was of couse | 15:17 |
@wiking | *course that ccache was not hitting any cache | 15:17 |
@HeikoS | ah great | 15:17 |
@HeikoS | thats really good news then | 15:17 |
@wiking | because of a slight change of the compiler flags | 15:17 |
@wiking | which might have only effected 2-3 srcs | 15:18 |
@HeikoS | wiking: macos also seems to work better these days right? | 15:18 |
@HeikoS | so we can have nighly ubuntu, nightly mac, and nightly cloud? | 15:18 |
@wiking | but since ccache is sensitive for compiler flags... basically it never hit a cache | 15:18 |
@wiking | yeah we could | 15:18 |
@wiking | but we still need to fix this separation | 15:18 |
@wiking | being able to build a modular interface | 15:18 |
@wiking | based on a preexisting shogun installation | 15:19 |
@HeikoS | okok | 15:19 |
@HeikoS | could we use this in travis to reduce time? | 15:19 |
@wiking | but yeah another problem now is c# modular interface | 15:19 |
@wiking | mmmm we could although that's quite tricky | 15:20 |
@wiking | if we could like distribute and persist somewhere the ccache | 15:20 |
@wiking | then we actually could save compilation time ;) | 15:20 |
@wiking | now travis offers something like a docker based build | 15:21 |
@HeikoS | wiking: well just a thought | 15:22 |
@HeikoS | I think a better way would be to split shogun into modules | 15:22 |
@HeikoS | but later | 15:22 |
@wiking | but that is still shitty atm | 15:22 |
@HeikoS | whats the problem with c#? | 15:22 |
@wiking | well mono 3.x or later | 15:22 |
@wiking | is just not compatible with the current swig setup | 15:22 |
@wiking | so we need to fix some typedefs | 15:22 |
@HeikoS | wiking: ah crap | 15:24 |
@HeikoS | wiking: maybe thats a good point for someone to dive into those typemaps | 15:24 |
@HeikoS | wiking: and then do the matlab ones on the fly ;) | 15:24 |
@wiking | hehehe | 15:25 |
-!- txomon|home [~txomon@unaffiliated/txomon] has joined #shogun | 15:27 | |
-!- besser82 [~besser82@fedora/besser82] has joined #shogun | 15:31 | |
-!- mode/#shogun [+o besser82] by ChanServ | 15:31 | |
-!- HeikoS [~heiko@0544f0d5.skybroadband.com] has quit [Quit: Leaving.] | 15:34 | |
-!- HeikoS [~heiko@0544f0d5.skybroadband.com] has joined #shogun | 15:36 | |
-!- mode/#shogun [+o HeikoS] by ChanServ | 15:36 | |
@wiking | HeikoS: so | 15:40 |
@wiking | what's the plan of the world domination | 15:40 |
@wiking | ? | 15:43 |
-!- nszceta [~nszceta@5ED062E5.cm-7-1b.dynamic.ziggo.nl] has quit [Quit: Textual IRC Client: www.textualapp.com] | 15:49 | |
@HeikoS | wiking: what? :D | 16:23 |
-!- besser82 [~besser82@fedora/besser82] has quit [Ping timeout: 256 seconds] | 20:12 | |
-!- HeikoS [~heiko@0544f0d5.skybroadband.com] has quit [Quit: Leaving.] | 21:57 | |
--- Log closed Sun May 10 00:00:19 2015 |
Generated by irclog2html.py 2.10.0 by Marius Gedminas - find it at mg.pov.lt!