Open in new window / Try shogun cloud
--- Log opened Wed Nov 21 00:00:17 2012
wikingso it seems i can rebase now00:01
wikingso what do i use instead of sg_realloc?00:05
@sonney2kSG_REALLOC /SG_MALLOC / SG_CALLOC everything as usual00:05
wikingstructure/CCSOSVM.cpp:294:7: error: use of undeclared identifier 'SG_REALLOC'00:06
wikingafter rebase00:06
wikingstructure/CCSOSVM.cpp:298:50: error: too few arguments provided to function-like macro invocation00:06
blackburnwiking: it was working before?00:07
wikingit was00:07
wikingbefore rebase to the memory magix00:08
wikingso what the fuck00:08
blackburncan yuo show these 294 and 298?00:08
blackburnwiking: you should provide old size00:10
wikinghave any of you seen man realloc ? :D00:10
wikingi mean sorry but why would u need the oldsize?00:11
blackburnbecause there is not realloc for new/delete00:11
@sonney2kwiking, that is new[00:11
@sonney2kand copy00:11
wikingwell realloc shouldnt behave like that00:11
@sonney2kand no way to figure out how much to copy underneath00:11
@sonney2kwiking, there is no realloc for new[]00:11
blackburnit is not realloc in case of sgvectors00:11
wikingmy 2 cents00:11
@sonney2kso that is why we need that00:11
wikingfuck me in da face00:12
wikingi mean if we want memory managment00:12
blackburnsee better solution?00:13
wikingwhy u just don't take the fucking memory pool from kernel.org00:13
wikingeven better00:13
blackburnhow does it solve your problem?00:13
wikingit was stupid of me to rebase00:13
blackburnyou don't know old size?00:14
blackburnwiking: put 0 here if you don't know00:14
wikingyeah i have to maintain00:14
blackburnit is not referenced anyway00:14
wikinga new variable00:14
wikingthat like fucking tracks00:14
wikingeach fucking REALLOC's00:14
blackburncan't say adding zero is so much issue00:15
wikingno but if i want to use it right00:15
wikingthat's what i should do00:16
blackburnit is used only if you realloc array of vectors00:16
blackburnwiking: so?00:17
wikingi have 1000+ line of patch here00:19
wikingand now trying to merge00:19
blackburnso you want to change everything to use pool with one more 1000+ line patch? :D00:19
wikingno i have some other patch00:20
wikingand now i've rebased00:20
wikingand now i'm broken00:20
wikingso i have to merge now00:20
wikingso that i can do anything00:20
wikinglike continue working00:20
wikinglibbm* shit has been minorly refactored00:21
wikingbut i have a feeling that that code should be rewriten in c++00:22
blackburnwhich code?00:22
wikingoh yeah00:23
wikingcode is breaking00:23
blackburnwell I agree it could be better to write in C++ everywhere00:24
wikingwell there's so much code redundancy in that one00:24
blackburnbut that's legacy - do you want to write and test it once again00:24
wikingwell i've clean it up00:24
wikingas much as i can00:25
blackburndoes it make use of vector operations?00:25
wikinggot like /2 speedup00:25
blackburnbecause of eigen used?00:25
wikingso it's like 2* faster...00:25
blackburnfor dot00:25
wikingwell eigen and cblas00:25
wikingwithin each iteration it does at least 5 dots00:26
blackburnwe should think about alignment actually00:26
wikingand another 3 of vec1_plus_alpha_vec2 stuff...00:26
wikingbut then again00:26
blackburnfeature vectors should be aligned00:26
wikingsgvector.h has to be refactored00:26
wikingi fucking hate to call static functions for a dot product00:26
wikingwell alignement00:28
wikingis os/arch dependent00:28
wikingso some cases u are lucky00:28
wikingthat the mem is alloced00:28
blackburnin case of 3d vectors00:28
blackburnwe are always unlucky00:28
blackburnI mean for eigen00:29
blackburnI don't care about cblas as it as old as sonney2k :D00:29
blackburnI didn't suggest that guy eigen300:32
blackburnto not speed up nuclear weapons development00:32
blackburnoh I should stop joking about that00:32
blackburnwiking: where have you seen that distCum?00:36
wikingof course a matlab code00:38
wikingacademics ;)00:38
blackburnwhat does it mean?00:38
blackburndistance okay00:38
blackburnbut what cum stands for in code?00:38
wikingcumulative distance00:38
blackburnwiking: I received once00:38
wikingthat's another great one00:39
blackburnthat was task for sth like 'mathematical analysis'00:39
wikingcalculs ;)00:39
blackburncalculus right00:39
wikingmath anal00:39
blackburnwiking: just today my neighbor gsomix showed me their teacher sent zad.zip00:40
blackburnmust be 'zadanie' - task/problem00:40
blackburnbut zad is effectively ass you know probably00:41
-!- zxtx [] has quit [Read error: Operation timed out]00:41
shogun-buildbotbuild #149 of nightly_all is complete: Failure [failed compile]  Build details are at
shogun-buildbotbuild #178 of nightly_default is complete: Success [build successful]  Build details are at
@sonney2kwiking, did you stop crying?08:22
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master:
-shogungit:#shogun- shogun/master 0a59171 Soeren Sonnenburg: include lib/memory.h in common.h again since SGVector forward declaration works ok now08:37
sonne|workwiking: b'cause you only have to change one line which is SG_REALLOC for this stuff to work09:11
sonne|workshogun-buildbot: force rebuild nightly_all09:11
shogun-buildbottry 'force build [--branch=BRANCH] [--revision=REVISION] [--props=PROP1=VAL1,PROP2=VAL2...]  <WHICH> <REASON>'09:11
sonne|workshogun-buildbot: force build nightly_all09:11
shogun-buildbotbuild forced [ETA 28m17s]09:11
shogun-buildbotI'll give a shout when the build finishes09:11
-!- blackburn [~blackburn@] has quit [Quit: Leaving.]09:27
shogun-buildbotbuild #150 of nightly_all is complete: Success [build successful]  Build details are at
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has joined #shogun11:03
-!- zxtx [] has joined #shogun12:43
-!- zxtx [] has quit [Ping timeout: 248 seconds]12:50
-!- zxtx [] has joined #shogun13:07
-shogungit:#shogun- [shogun] vigsterkr opened pull request #831: Add template specialization for SGVector's arg_max and max (master...master)
wikinga little bit of eyecandy first ;D13:08
-shogungit:#shogun- [shogun] sonney2k pushed 2 new commits to master:
-shogungit:#shogun- shogun/master 1f02914 Viktor Gal: Add template specialization for SGVector's arg_max and max...13:21
-shogungit:#shogun- shogun/master ba1e9af Soeren Sonnenburg: Merge pull request #831 from vigsterkr/master...13:21
sonne|workwiking: heh13:21
sonne|workwiking: no idea how the discussion yesterday night went on13:22
sonne|workbut what are your complaints exactly?13:22
blackburnsonne|work: he is just like me :D13:22
sonne|workblackburn: which means?13:22
wikingthat realloc api13:22
blackburnsonne|work: marvin13:22
sonne|workI see13:23
wikingah shit13:23
sonne|workwiking: yes I know. I couldn't come up with any standard solution. If you have an idea speak up13:23
wikingi forgot to change sparsematrix13:23
sonne|workwiking: fwiw I don't like it either but whenever we do realloc in shogun we know the old size so this is the least intrusive change13:24
sonne|workthe only other option I see is to use new SGVector<int>[10]() and delete[] throughout the code13:28
blackburnor restrict realloc for sgvectors13:29
sonne|workblackburn: yeah but then we cannot use DynArray for SGvector13:29
sonne|workblackburn: btw you said std::vector works for cases that need inplace constructors?13:30
blackburnsonne|work: I am curious would work for you?13:30
blackburnsonne|work: vector calls constructors yes13:30
blackburnit has no chance to fail on that no matter what class it is13:31
sonne|workblackburn: any idea how it know when to call constructors?13:31
blackburnsonne|work: what do you mean 'when'?13:31
sonne|workI mean vector of double works too right?13:31
sonne|workwell it needs to know that it can call the constructor13:31
blackburnyes why not13:31
blackburnare you sure new double[]() doesn't work?13:32
blackburnI think it does actually13:32
sonne|workI didn't try but then I wonder why plain new[] exists...13:34
blackburnsonne|work: take a look at stl code :)13:36
wikingblackburn: new double[] should work13:37
blackburnwiking: new double[30]();?13:37
sonne|workblackburn: well it compiles...13:38
blackburnhahaha youtube is banned in my country OKAY13:39
wikingblackburn: lol?13:39
sonne|workwhat happened?13:39
blackburnyoutube is illegal :D13:40
wikingblackburn: can u open it still?13:40
wikingso what?13:40
blackburnbut won't be able soon I think13:40
wikingif they catch u watching it, then they'll come and fuck you up? :)13:40
blackburnthey introduced a list of sites that are denied13:40
wikingah yeah13:40
wikingthe blacklist13:40
blackburnso just a few hours ago they put youtube there13:40
wikingi mean hey13:41
blackburnand now ISPs will deny it13:41
wikingyou don't wanna watch that shit anyways13:41
blackburnhaha right13:41
wikingmaybe there's THAT kind of content there :P13:41
blackburnwiking: they tried to ban it before because of 'innocence of muslims'13:42
blackburnfucking muslims they are trouble makers all around the world13:42
wikinganother fixfuck for shogun13:43
-shogungit:#shogun- [shogun] vigsterkr opened pull request #832: Fix SGSparseMatrix ctor to comply with SGMatrix ctor (master...master)
wikingblackburn: what should new double[]() be?13:43
blackburnwiking: call ctor of double13:43
sonne|workwiking: double[]13:44
-shogungit:#shogun- [shogun] lisitsyn pushed 2 new commits to master:
-shogungit:#shogun- shogun/master c7f0a72 Viktor Gal: Fix SGSparseMatrix ctor to comply with SGMatrix ctor...13:44
-shogungit:#shogun- shogun/master 810c9e5 Sergey Lisitsyn: Merge pull request #832 from vigsterkr/master...13:44
sonne|workmaybe it inits with 0's ?13:44
blackburnsonne|work: yes13:44
sonne|workwiking: did you convert all places that call SGSparseMatrix?13:45
sonne|workI mean otherwise stuff is transposed...13:45
blackburnsonne|work: wait but doesn't new Class[] call constructors?13:46
sonne|workblackburn: no13:46
wikingsonne|work: yeah i've tried to cover all of it as you can see in the patch13:46
sonne|workwiking: k13:46
blackburnsonne|work: so you mean new Class[30] is unsafe later? I am unsure about that13:46
wikingblackburn: new should call the default ctor13:47
sonne|workcrazy crazy crazy13:47
sonne|workdidn't know one can write new Class;13:47
blackburnsonne|work: yes sure one can13:47
blackburnbut I was unsure about new[]() and new[]13:47
blackburnso we just can use new[] and it will work nice13:48
sonne|workblackburn: yeah but it has a different meaning13:48
blackburnwhat's different?13:48
sonne|worklook at the above link13:49
wikingsonne|work: well depends  ;)13:49
wikingi mean blackburn :)13:49
wikingit depends13:49
blackburnsonne|work: no, new[] would work in all our cases I think13:49
wikingi mean if u pass a class13:50
sonne|workblackburn: no I think it won't work like in the above link stuff will be uninited13:50
wikingthen it should work as all classes have by default a ctor and even a copy ctor (as c++ automagically defines one) if you dont enforce it to be private13:50
blackburnsonne|work: things like int?13:50
blackburnsonne|work: okay actually new[size]() is ok for me13:51
blackburnit is rather universal13:51
blackburnalways works no matter simple type or class is used13:51
wikingwe always have default ctor for every class anyways for serialization13:51
sonne|workblackburn: yeah for 'higher' types that makes sense13:52
blackburnwiking: sonne|work it took 2 hr to recognize they are doing something wrong and youtube is not banned anymore :D14:14
wikingthat was fast actually14:17
blackburngreat soviet firewall14:33
-!- zxtx [] has quit [Ping timeout: 248 seconds]15:39
-!- zxtx [] has joined #shogun15:42
-!- zxtx [] has quit [Ping timeout: 255 seconds]15:48
-!- blackburn [5bdfb203@gateway/web/freenode/ip.] has quit [Quit: Page closed]15:49
-!- zxtx [] has joined #shogun15:50
-!- zxtx [] has quit [Ping timeout: 246 seconds]17:01
-!- blackburn [~blackburn@] has joined #shogun17:02
-shogungit:#shogun- [shogun] sonney2k pushed 2 new commits to master:
-shogungit:#shogun- shogun/master dfc7d7d Soeren Sonnenburg: move function definitions to .cpp file18:17
-shogungit:#shogun- shogun/master 217af00 Soeren Sonnenburg: define index_t in lib/common.h like all other datatypes18:17
@sonney2kblackburn, now see what you have done - you could have suggested eigen318:19
blackburnsonney2k: hahahaha18:19
-!- shogun-buildbot [] has quit [Remote host closed the connection]18:20
--- Log opened Wed Nov 21 18:21:32 2012
-!- shogun-toolbox [] has joined #shogun18:21
-!- Irssi: #shogun: Total of 5 nicks [0 ops, 0 halfops, 0 voices, 5 normal]18:21
-!- Irssi: Join to #shogun was synced in 6 secs18:21
-!- shogun-buildbot [] has joined #shogun18:21
-!- sonney2k [] has joined #shogun18:21
-!- mode/#shogun [+o sonney2k] by ChanServ18:21
-!- shogun-buildbot [] has quit [Remote host closed the connection]18:27
--- Log opened Wed Nov 21 18:28:27 2012
-!- shogun-toolbox [] has joined #shogun18:28
-!- Irssi: #shogun: Total of 5 nicks [0 ops, 0 halfops, 0 voices, 5 normal]18:28
-!- Irssi: Join to #shogun was synced in 7 secs18:28
-!- shogun-buildbot [] has joined #shogun18:28
-!- sonney2k [] has joined #shogun18:28
-!- mode/#shogun [+o sonney2k] by ChanServ18:29
-!- zxtx [] has joined #shogun19:59
@sonney2kwiking, around?20:51
@sonney2kwiking, I wonder whether the sparse dot product now assumes sorted indices or not.20:51
-shogungit:#shogun- [shogun] sonney2k pushed 1 new commit to master:
-shogungit:#shogun- shogun/master 7528d0e Soeren Sonnenburg: put impl. of sparse vector / matrix functions into .cpp21:06
wikingsonney2k: it does21:26
wikingeven the comment says so21:27
wikingthat we suppose that the indicies are ordered in non descending manner21:27
shogun-buildbotbuild #179 of nightly_default is complete: Failure [failed tutorial]  Build details are at
--- Log closed Thu Nov 22 00:00:15 2012