 Think it's time to start with the next talk Which is me talking about Debbie cams that are most important announcement as we finally have a logo. I Plucked this together last year. I think what you use ago So Debbie cam is a Something I started about 12 years ago now I got into Debbie in because I studied chemistry and They were Basically no chemistry packages in Debbie and around the turn of the century I was back then a student and there was one Program called cam tool which you can use to draw chemical to these structures Which you need for for your studies And I was doing a bit of upstream work on it and I thought okay. I'll package that for Debbie For five or six years why not get into it and then just when I got about it some guy called dr Günther Bechler or something ITP did and uploaded it to Debbie and I was like damn that was my one chance to become a deviant developer and It's gone. It's the only chemistry package. I know which is open source And then a couple of months later. He actually resigned from Damian for personal reasons so I decided to take it over and at that time There was almost no Free software and about chemistry so as a as a 100-foot overview Debbie cam is the W improve pure plant blend It's officially started 2006 But I mean the the thing itself has been going on since I started packaging chemistry packages around 2000 so right now we have 30 persons with commit access, but it's mostly a couple of people actually doing a lot of work and We have now 100 source packages I'm mostly I've since been focusing on computational chemistry So that's the packages. I mostly maintain, but I've also not been actually Employed as a chemist for the last nine years. So this is only a volunteer thing for me and I guess for most other people except for Grammings who works at University of South Africa was last year's Japconf So he's actually made a deep an administrator at the chemistry department there or chemical engineering I guess and he he gets to work quite a bit on your boon to endeavor and packages for for that field so Just takes a while. So that's one of the graphs that Andreas I don't know. Did you show no? Well, this is the upload of statistics for for Debbie cam But Andreas is doing those statistics. So you can see I'm the orange person. So we're for the first couple of years I did the all by myself And around 2006 we started the the subversion Repository because other people wanted to participate and that was mostly Daniel Lydard who did a lot of work from 2008 to 2013 roughly But then in the last couple of years he kind of disappeared. So he showed up again a bit Two years ago and we actually he met what he came to Heidelberg But I actually haven't heard from him for a while So he has he has unfortunately kind of disappeared and then Philippa Rosconi. He's doing mass spectrometry Packages so but he's mostly working on his own on the camera. So he has one or two people But they're doing their stuff and I'm doing the computation chemistry stuff. So we're a bit disjoint but he's been doing a lot of work and then Nicolas Breen He's packaging grow max, which is the Molyka Dynamics Software, but he's that's basically the only package. He's maintaining but he's doing it all the time and The next one is then Waymings who has been working We started to do a lot of work this year last year. So he sees where quite active and then it's Andreas already so As you can see is there were one or two people and it's mostly me there and it's mostly Daniel for some reason last year. I did apparently a lot of Uploads or maybe that was the best tone. No, there's not possible 12. So yeah, I did a lot of uploads last year I don't remember that actually because I thought I didn't have a lot of time But maybe it's because of the source only uploads now that I tend to do more uploads these days so as I said the original mission statement for Debbie cam was Package all chemistry related free and open source software. There's not so few of them. Actually, there was only Like four or five in 2001 Camtula already mentioned mpqc, which is the massively parallel quantum chemistry That means that meant like 16 cores In parallel or something in 2000 and a few other Viewers garlic and chemical garlic a chemical was a GTK or is a GTK to based visual 3d viewer So though that was it there was these five things and we packaged it and there was not a lot There was a bit more coming on after a couple of years in 2006 We had 20 packages. I checked that and now it's 100 so it slowly got more and more and Nowadays actually the mission statement. I would say it's packaged the most useful versatile projects in each field because It turned out that there's five or six packages to to write Chemical structures and there's like 10 or 20 different ones to three-dimensionally visualized molecules And it doesn't make sense to actually package them all Because they're a huge overlap, but on the other end it's also quite difficult to figure out which ones are the ones which are useful and also actually I found out that Even more so in computation chemistry. There's a huge difference in performance. So we had one project called the QM speed test by no lab oil Who was just running? Optimizations of of molecule. I think was aspirin or something on different codes and and actually there's a factor of 50 or Something in the runtime. It's huge like one some packages are done in two minutes and another one took three hours for the same stuff so There's a huge difference between code bases, which are Fortran which have been optimized since the 70s versus some guys doing a C++ project over the summer It's the same algorithms, but the implementation is is very different So right now we try to do that for example just last week. There was another article in Journal of computation chemistry about another thing pirate is a well-done medium-sized python library for rary wins So that's another open source thing. So every there's a new one coming up every week. I have the feeling and Actually, I have another 100 packages on a list of prospective packages Which are all free software There's a few really big ones siesta for example is a very well-known code and then the periodic Or crystal community which got open source a while ago. We should package it big DFT is pretty pretty well-known and Pi SCF Couple of those are really should be packaged, but we don't get around to it that much But there's 100 at least that we know of so at least twice the size that we already have so we should really make a Point in only packaging the more important ones So what I want to do next is give a few examples of really innovative things because at the beginning There was open source in the in the let's say knots Trying to catch up like give a the same functionality as proprietary software But nowadays actually there is open source codes especially in computation chemistry which are even more advanced than the best proprietary software codes or at least They caught up to it very much. So so one of them is bagel Which is done by Toro Shiozaki in Northwestern and he basically wrote it over the last couple of three years with a couple of grad students Himself, but they're doing pretty awesome stuff. For example on the fly cast p2 to surface hopping dynamics And nobody did that actually not even So this is a paper from from a month ago or something I don't think there's even a proprietary part package that does that or just two weeks ago They published a new part of NW cam, which is a huge code base from the Northwestern laboratory and in the US pushing configuration Interaction to the limit It's as you can maybe not read they have one trillion Orbitals or whatever going on in there in their calculation So it's massive on huge clusters and then this is really kind of competitive code or another one from Zurich They they're doing CP 2k, which is a that we can package for a for a long time and they are actually simulating They have simulations with millions of atoms in one simulation and each atom is actually Actually modeled on the quantum scale. It's not just the sphere or something as you have a molecular dynamics. Those are really quantum Molecular dynamics with real quantum chemistry behind it but on the other hand these Advances in in open source got people quite a bit unhappy all these proprietary people and so for example a while ago There was this letter or this view point they say in the journal of chemistry a physical chemistry letter saying What a surprise of open source software and it was from a lot There was like 10 or 20 authors and they're all the main authors of the proprietary Software packages in the field of computation chemistry and they say yeah, you know Everybody says you should open source everything, but really we have such a complicated code base So for example Q cam has 300 papers and we have like five million or two and a half million lines of code and we have so many developers and It's really really hard work to do so we cannot just open source it and people should continue funding our proprietary model of innovation and actually what I've Saw in the last couple of years is that there is a very big change in the sense that funding agencies like the That case the German funding agency DFG are not funding Well, what they're doing is they're funding research for new Algorithms and stuff, but they won't fund proprietary software just for the maintenance of it So for example, there's this code from Stuttgart Called IMD and what they write is like IMD has now been placed under the GNU GPL as a consequence of moving the repository to get up Which that's bit stupid because that's not a requirement, but okay and the requirements of the funding agency DFG So the DFG told them okay if you want money from us you should make it open source and they did so that's quite a Huge impact, I guess that open source has there and I also heard from Privately from a colleague of mine. There is a huge code in the Fitzhaber Institute in Berlin and they also have a lot of trouble because it's proprietary to get grants these days and What they're doing now is they're open sourcing their underlying Lining linear algebra Libraries and and trying to make new projects out of them So in order to get some funding at least yeah, because they they cannot get funding for their proprietary parts So and there has been a couple of previous open sourcing and I'll be okay. I already mentioned Used to be proprietary, but they open sourced it about 10 years ago or something side three side four these are Standard packages were for students to simulate molecules and They're rewrite so psi one two used to be proprietary those are now GPL and Yes, I just mentioned and there's any CI which just got Open source a couple of months ago in a sense it was on github and then suddenly somebody put the GPL on it Well one of the office, so it's legit, but now it's now it's really open source so That's as an introduction. I want to just mention what what Debbie came does these days These are the tasks Andreas talked about it earlier, so we split it up a bit more fine-grained in these days I actually did some work on that. I think it didn't land in Jesse back then, but now it's in stretch at least so we have 12 tasks, I would say we have a crystallography and Analytical biochemistry, this is what Rosconi is mostly doing and I'm mostly working on molecule app initial Molecular dynamics molecular modeling and periodic I mean issue stuff and then they have lots of packages which produce or input output stuff There's some major packages Which I don't want to get into because I don't have a lot of time I just want to mention that yeah There's also Yeah, for visualization. For example, we have pymol. We which got mentioned earlier Very grateful that he got picked up because I didn't have the time and j-mole got also picked up by Infinity zero I guess So this was finally upgraded. So there's some some cooperation happening there So those are all the packages, but there's also a lot of interdependencies on the packages So I decided to make a graph of all the built dependencies at least for the computational chemistry packages So that's probably pretty bit difficult to read But those are the libraries and there are lots of interdependencies for example mpqc3 Has a lot of libraries and lip xc def is used in a lot of packages Same as lip int these are just the chemistry packages a lot of them Of course use the linear linear algebra packages like blast lapak or the MPI packages like open MPI and And this one is how the input output things work So we have a couple of packages like open bubble python ase arogadro and Travis and cc lip that parse the output of packages to either Compute properties or just display properties of them or to transform those outputs into other File formats and what they also can do is produce inputs because most of these packages Actually, they they still rely on text input and output. So it's a you have an input file You run the input file on some executable and then you get an output file But this is getting going even further for example, there was a recent paper for the PSI for from the PSI for developers and they This is a actually a graph from their paper, which is in the journal of chemical theory and computation and they have all these External projects which they integrate and they all open source except for MRCC I think And we haven't packed. Well, we have packaged lip int but all and cam MPS to those two are packaged But all the other ones are still not packaged. So we have quite some work to do here They can also be used by by sci-4 and in order to Make their work better Couple packages that are new and stretched I don't want to get into too much detail here But as I said Jmo got updated. That's that's great sci-4 is finally at 1.0 This is a very standard package for people to work on CP2k got updated and Grammings brought a seed atomic system simulation environment into that end Finally, I wanted to talk about it about outreach that I tried to do at least so in 2015 I gave a talk about CP2k and depicam in Zurich at the CP2k Developer meeting which was quite nice. There was also a guy from fedora there So they invited us giving a talk about how we package stuff And I also presented two posters in the summer of 2015 at the 10th European Conference on computation chemistry in folder and a few weeks later in the symposium for theoretical chemistry in Potsdam So I made this debicam poster which was also actually in in I actually also presented it in or it got printed for South Africa last year and hang up at the open day, but I made this poster of debicam all the capabilities of the computation chemistry packages and Got into some quite some interesting talks a lot of those people knew Debbie in but Not really Debbie Cam has always the same problem with the pure plants thing But that gave some some outreach I guess so finally I want to Give let's talk a little bit about what's what's left to do. So one thing I think we should really do in a bit Not happy that it's not happening upstream is that there should be more input output pausers for for the debicam packages so we have a Avogadro, which is a Graphical KDE application or Qt application which can read them and generate input But it's not doing it for all the packages we have an open bubble is is also in CC lip these are packages Which we could extend? For example CC lip is just Python. So it's rather easy So a couple of packages like CP to K Begu and aces are not really represented there So maybe there is a GSOC program Google some of code project looking there We should switch finally from subversion to get on Daniel Leiter like subversion very much so We didn't force it, but he's not really active then more and then if new people want to work on something they can just Convert stuff to get that's happened with Pymol. I think for example and some others There's just a question of whether we just use the debi directory or the full project It's easier. I guess if you do the switch to just use the debi in Directory, but we'll see that I would like to get more serious on back ports I Have been back porting one or two packages, but not as well as the neuro deviant team does they unfortunately not here But they are really doing great work on that and Then there's the other things like what to do about the MPI implementation There was quite a few problems with MP open MPI in the stretch 3d cycle and I heard from some upstreams that Open MPI is not working with us It's not it's not good if you actually have both multiple threats and multiple processes Which nowadays is the case? So you would have a cluster where you run the program on several nodes several computers on the cluster But each cluster would have 16 cores, so you don't want to run one threat on each node you want to run 16 threads and then 16 cores and For that kind of thing apparently open MPI had a lot of problems, but they are supposedly fixed So we'll see but I think one thing that it's going forward as we saw on the deviant science list I think that at least some packages like scalar pack will be compiled for both MPI CH and open MPI so you can choose your your Stack so for example for bagel I use MPI CH because it was more stable The other thing is the DH Fortran Fortran 90 module thing that's looming over us all the time It's again stuck every depth conf is going a bit forward, but it's again stuck. I guess so maybe we I or somebody should start working on that. I would also have things about more a Bit bit about optimizing packages for Production use so right now we have to go with the lowest common denominator which means MTU native or as much as not possible So it would be good if we could integrate it that you can just recompile your package And it's it's running at machine speed if that's actually useful. We should also benchmark it It's a bit difficult to to say anything about it without running proper benchmarks And then we should remove the packages which are outdated So there's quite a few ones which are 90s technology and there's been better ones in the same sins And we should look at them and actually remove them Or maybe if there's just slow if they're doing computational stuff and then other packages are way faster Just remove them, but we haven't really start doing that yet. So that's it basically for me That's some contacts. We have Debbie Camde well mailing list. There's an ISE channel, but it's not used really and The subversion and get repo and if you're interested in chemistry packaging Always happy to collaborate or to help or to just give you access. So Thanks, and is there any questions? Anyway overruns anyway, so I guess It's fine if there's no questions and we'll be continuing with the science Both in about nine minutes at half past three. Okay. Thanks