Porting Rosetta - Unix, Linux etc
Message boards : Number crunching : Porting Rosetta - Unix, Linux etc
| Author | Message |
|---|---|
Dotsch Send message Joined: 12 Feb 06 Posts: 111 Credit: 241,803 RAC: 0 |
10) Runs on Mac and Linux too. Everyone can help. I disagree you in this point. I would also do more work for Rosetta, but there a no binaries for my systems available.... There are a lot of other platforms (HPUX, Solaris, AIX, IRIX, TRU64, FreeBSD, OpenBSD, NetBSD, MAC OS X on Intel,...) out there, which do a lot of work for the other projects (SETI, SETI Enhanced, Einstein, SZAKI). I have asked in different places in the forums for ports for the most common Unixes, but there was never a awnser by the project officals. So I can not commit the always told good comunications between the project and the partipicants (your point 4). |
 Dimitris HatzopoulosSend message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
10) Runs on Mac and Linux too. Everyone can help. I think this particular issue is quite technical. Considering that the project has been releasing new executables every week or so recently, I think they might want to let things stabilize a bit. There are much more pressing priorities right now imo, like tweaking BOINC server feeder/scheduler to send large WUs to PCs suited for them. Anyway, although I haven't tried it myself, I'm pretty sure one can right now run the Linux ELF binary of Rosetta under *BSD (FreeBSD, OpenBSD, NetBSD) using the well-known technique of ELF branding brandelf -t Linux rosetta_5.07_i686-pc-linux-gnu the executable is statically linked afterall: # file ~boinc/BOINC/projects/boinc.bakerlab.org_rosetta/rosetta_5.07_i686-pc-linux-gnu /home/boinc/BOINC/projects/boinc.bakerlab.org_rosetta/rosetta_5.07_i686-pc-linux-gnu: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), for GNU/Linux 2.2.5, statically linked, stripped The other OSes (HPUX, Solaris, AIX, Irix etc) represent just too small a marketshare in the DC world, so I doubt most projects will officially support them in-house (i.e. compiling executables for them in-house). Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
Feet1st Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
10) Runs on Mac and Linux too. Everyone can help. I was trying to emphasize support for Mac's and Linux, because there are other projects that don't support those. As you've shown my word choice of "everyone" was a bit too broad. Could I ask you to bring this up in the thread on why some people avoid BOINC projects? Perhaps someone there can offer some specific details to get you on your feet with those other boxes. It might be helpful if you identify in your list which boxes you actually have, and which exist out in the field, and are not supported by Rosetta. Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
|
Dotsch Send message Joined: 12 Feb 06 Posts: 111 Credit: 241,803 RAC: 0 |
Anyway, although I haven't tried it myself, I'm pretty sure one can right now run the Linux ELF binary of Rosetta under *BSD (FreeBSD, OpenBSD, NetBSD) using the well-known technique of ELF branding How much did it impact performance ?
No. There over 50 million Credits for this platforms on SETI (0,7 % of total credits) and about 0.5 million credits on SETI Beta (2,2 %), einstein 0.7 Mio. credits(0,1 %), SZAKI 17000 Credits (0,1 %). But a lot of platforms are not available for a long time, which decreases percental credit calculations... |
|
Dotsch Send message Joined: 12 Feb 06 Posts: 111 Credit: 241,803 RAC: 0 |
10) Runs on Mac and Linux too. Everyone can help. I have still posted it many times to different rosetta forums including my offer to help at the porting work, but did not get any anwsers. The plaform support it not a boinc specifiy problem. There are boinc clients for the most platforms available, some projects offers still binaries of there science apps. I think a broad platform support could boost the throughput of rosetta, and is very good for the publicity. Also, I not only thought for me, I also thought for the other peoples with not supported plaforms... |
|
Feet1st Send message Joined: 30 Dec 05 Posts: 1755 Credit: 4,690,520 RAC: 0 |
I have still posted it many times to different rosetta forums including my offer to help at the porting work, but did not get any anwsers. I appreciate your willingness to help with the porting. The problem though is that they do not release the source code. This is because while they do make the program available to universities and acedemia, they still charge for it to corporations. If they give it to someone to port, they would be losing control of the source code. So, I guess we'll have to wait for the project to support those platforms. At this point, there are much larger targets to growing the project. The checkpointing was a major one. It means everyone already crunching is significantly more productive. My apologies for overstating that everyone can help. I was simply trying to express optimism and enthusiasm for the project. Anyone else have 10 reasons they crunch Rosetta?? Add this signature to your EMail: Running Microsoft's "System Idle Process" will never help cure cancer, AIDS nor Alzheimer's. But running Rosetta@home just might! https://boinc.bakerlab.org/rosetta/ |
|
Dimitris Hatzopoulos Send message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
Anyway, although I haven't tried it myself, I'm pretty sure one can right now run the Linux ELF binary of Rosetta under *BSD (FreeBSD, OpenBSD, NetBSD) using the well-known technique of ELF branding Performance should be very good, like a native a.out app. For more read Setting up Linux compatibility on FreeBSD 6 "Linux compatibility isn't instruction-level emulation or some kind of virtual machine. The Linux ABI is implemented in the FreeBSD kernel, so in most senses the Linux binaries could be considered to be running natively." If I may remind you, even mainstream apps like Adobe Acrobat Reader and RealPlayer and browsers like Opera and Flash-plugin AREN'T available "natively" (a.out binary) for *BSD, but as ELF.
I guess everyone is entitled to his own opinion about what is important and what isn't, but I'm trying to be a realist here and provide ideas to maximize overall TeraFLOPS to Rosetta@home. E.g., you have attached 2 hosts to R@H, one with 384MB and another with 192MB RAM, both underspec'ed wrt the 512MBytes RAM "official" requirement. One of them has a Win98SE O/S (where sometimes BOINC misreports time and claims 0 credits). Now, realistically speaking, if I were making priorities for Rosetta@home, I'd much rather try to better utilise these underspec'ed PCs with less RAM and slower CPUs like yours, by ensuring that I was sending them the smaller WUs. Or at least avoid sending the big CASP WUs to slower PCs or non-24/7 PCs + leave-in-mem=no PCs. Then, I'd look into further optimising the science app, like akosf did for Einstein@home, improving the speed 4x (four-fold). Or take advantage of CPU features like SSE etc. And only AFTER these issues had been taken care of, only then would I even consider adding native support for niche O/Ses, like Solaris (much less HPUX or Irix), or doing ports that even big companies like Adobe don't do. Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
|
Dotsch Send message Joined: 12 Feb 06 Posts: 111 Credit: 241,803 RAC: 0 |
I guess everyone is entitled to his own opinion about what is important and what isn't, but I'm trying to be a realist here and provide ideas to maximize overall TeraFLOPS to Rosetta@home. If you has readen my last posts - it is also my goal to improve and enhance rosetta, too. E.g., you have attached 2 hosts to R@H, one with 384MB and another with 192MB RAM, both underspec'ed wrt the 512MBytes RAM "official" requirement. One of them has a Win98SE O/S (where sometimes BOINC misreports time and claims 0 credits). I know the Windows 98 problems... Now, realistically speaking, if I were making priorities for Rosetta@home, I'd much rather try to better utilise these underspec'ed PCs with less RAM and slower CPUs like yours, by ensuring that I was sending them the smaller WUs. Or at least avoid sending the big CASP WUs to slower PCs or non-24/7 PCs + leave-in-mem=no PCs. Sorry, but i disagree with again. There are several methods to increasse the througput which could also done in paralell, for example the enhancements you have written, portings,.... Normaly, if you have a linux app you could port it to other Unixes with some work. And it don't mather if it is a Solaris, HPUX or Irix, or... My request was coupled with a offer of help to port it to other platforms. - Feed1st I am well know about the licensing problems, but I meaned with help a lot of different aspects of help. Btw. Dimitris, are you one of the project officals ? - Is this a offical statement ? |
|
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
10) Runs on Mac and Linux too. Everyone can help. Actually I tried to respond to your email just today, and the message came back with an invalid address message. The message I was trying to send out you is this - A the present time the Rosetta source code has not been released to the public. This means that any versions for other platforms would have to be prepared by the Rosetta programmers. At the current time there are not enough people on the Rosetta team to create these Moderator9 ROSETTA@home FAQ Moderator Contact |
|
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
|
|
Dimitris Hatzopoulos Send message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
Now, realistically speaking, if I were making priorities for Rosetta@home, I'd much rather try to better utilise these underspec'ed PCs with less RAM and slower CPUs like yours, by ensuring that I was sending them the smaller WUs. Or at least avoid sending the big CASP WUs to slower PCs or non-24/7 PCs + leave-in-mem=no PCs. I'm not a project official, just a cruncher (that's why I wrote "if I were making priorities for Rosetta@home"). The issues are what Mod9 explained. Don't get me wrong, I agree with you that every little bit helps and having "native" binaries for every platform would be great. I've used most of these other Unices (*BSD, AIX, SunOS, Solaris, HPUX, Irix etc) and I like some of them a lot. I just don't think that porting Rosetta to those O/Ses would make much difference to the project's TeraFLOPS (other than make people like you and me happier to run Rosetta "natively"). Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
|
Beach Bum Send message Joined: 29 Apr 06 Posts: 9 Credit: 196,378 RAC: 0 |
I believe there is a way to keep both sides happy in the offer to port the program to other linux/unix layouts. As a consultant and programmer, most companies I do work for have a non disclosure contract I must sign and adhere too. Now most of these contracts have a monitary punishment for leaking anything. They also have a set years of non-disclosure. Most of them are for 25 to 50 years. If dotsh would like to help the project by porting the code. Then bind him to a said contract. This gives Rosetta a fall back if he is to break the contract. The smallest fine on any of my non-disclosures starts out at 500,000 dollars(U.S.). This does not even include the secondary part of the contract which involves damages done due to the breach. Those could easily run into much more money. I believe if Dotsh is wanting to do the code port, and he is upfront and honest about it. Then the contract should be no problem to him. As we know the code is not released to the public, then any release to the public can be tracked rather easy to the people with access. This contract should be instated with all people/companies that have access to the code if one is not in use now. Fax a contract to him, demand at least two notary seals with signatures of him and the notary on both. Once Rosetta has their copy of the contract, make the code available to him. It would help the project get some vital help from other OS's. We all know there is a strain on most of the projects in the Boinc system. That is why many volunteer their time. Take this spirit into the project, and let it help the project out. As Rosetta is not open source code, I, as well as others understand the need to limit access to the code. But for the sake of the project, use a contract and harness the extra help when it is given. There is a reason why SETI has grown, it harnesses as much of its volunteer staff as possible. They have no money to speak of at SETI. They have next to no funding, they stay afloat mostly on donations/help from its community. They do have issues at times. But they get through them. I do say that the feedback to users here, is a big plus for this project that could easily push it past SETI. Personally, my team and myself have been talking about bringing more of our equipment to crunch here. With that all said, I hope this helps to spread some light on an option.  Come Join us at Hawaiian Beach Bums |
|
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
I am aware that the project staff is looking at a variety of options with regard to additional platform support. But in this case it is not as simple as you describe. For one thing neither the corporate model you present, nor the SETI model are directly applicable. In the first place the Rosetta code is not stable such that it is easy to port as is the case with may other BOINC applications. A big part of what is happening here is research into the computational methods themselves, so the code changes in significant ways on very short time lines. In just the last month there have been 5 deployments of new applications at Rosetta. So the true model is not sending the code out, but instead bringing people in. Second, there is no corporate gain from the release of SETI code or many of the more arcane BOINC project applications. While SETI is a worthwhile research project, nobody can make any money if the code falls into the hands of a large corporate entity. That is not the case with Rosetta. A penalty of $500,000 is a pittance to pay for the potential profits to be made by a large pharmaceuticals firm that might profit from accurate protein structure predictions based on Rosetta code. From what I can tell the project team is committed to avoiding that scenario. Lastly, Rosetta and all its results will be released as a part of the publishing of scientific papers by the research team, and through a broad range of corporative agreements between non profit universities and other outlets of similar nature. What you are asking is the equivalent of having Einstein (the scientist not the BOINC project) release all his notes on relativity, two years before he actually published them. I feel certain the project will consider all its options, and do what is best in the larger picture. In any case as noted previously this thread is not the place for this discussion. Moderator9 ROSETTA@home FAQ Moderator Contact |
|
David Baker Volunteer moderator Project administrator Project developer Project scientist Send message Joined: 17 Sep 05 Posts: 705 Credit: 559,847 RAC: 0 |
I'd just like to point out again that the rosetta source code can be obtained for free from the UW tech transfer office. This version does not have the latest boinc interface built in, but any optimization or compilation issues that are solved with this version should be immediately transferable to the latest boinc version. |
|
Dotsch Send message Joined: 12 Feb 06 Posts: 111 Credit: 241,803 RAC: 0 |
In any case as noted previously this thread is not the place for this discussion.[/color][/b] Where is then a place to discuss this ? - Would you break down and eliminate this discussion again ? Or do you want to move the postings again to a other thread and cut out the thread with its context again so that no other than the posters can follow the discussion ? Sorry, your treatment of other users, and wild thread and posting movement is frustrating. |
adrianxw Send message Joined: 18 Sep 05 Posts: 653 Credit: 11,840,739 RAC: 34 |
A recurrent theme I've come across in a number of places is the somewhat cynical belief that someone, somewhere is "making money/gaining patents/supporting the US military" out of the freely given resource. This was true of course in the United Devices type setup. Wave upon wave of demented avengers march cheerfully out of obscurity into the dream. |
|
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
|
|
Beach Bum Send message Joined: 29 Apr 06 Posts: 9 Credit: 196,378 RAC: 0 |
From Moderator9 message post 16736 While SETI is a worthwhile research project, nobody can make any money if the code falls into the hands of a large corporate entity. That is not the case with Rosetta. A penalty of $500,000 is a pittance to pay for the potential profits to be made by a large pharmaceuticals firm that might profit from accurate protein structure predictions based on Rosetta code. If you read it right, I said the smallest amount started at 500,000. Most of the contracts have much higher penelties. But unless you have worked for Intel, Microsoft, Unisys, or many more of the big hitters, you most likely have not seen that type of contract figures before. Also as I am to read it from your posting, It makes it sound like the code is to be kept a secret, and the rosetta project is to be a commercial venture after we crunch things for rosetta. But yet the following quote by Mr. Baker himself puts a large hole in your theory of a large pharmaceuticals firm making money if they could get ahold of it. From Mr Baker message post 16739 I'd just like to point out again that the rosetta source code can be obtained for free from the UW tech transfer office. This version does not have the latest boinc interface built in, but any optimization or compilation issues that are solved with this version should be immediately transferable to the latest boinc version. Now what I was pointing a finger to was a possible solution to help the project, not to be flamed by a moderator as being a idiot trying to bring a business model into the project. Now as far as the move of my posting. The move made no sense seeing as a large amount of the thread I originally posted in was about the code porting. Which is what got me interested in making the original post in the first place. Anyway, Dotsch I believe Mr. Baker just gave you the place to get the source code, fair be it a bit older, but still the code. That should give you what you need to take a stab at porting it. And last but not least this quote from moderator9 in this thread. From Moderator9 message post 16736 In any case as noted previously this thread is not the place for this discussion. Now how in the heck can this not be the proper thread to discuss porting the app? 1. the name of the thread is Porting Rosetta 2. You Moderator9 moved the posts to this thread. So this statement from you I truely do not understand. And last as I feel this post will most likely be moved or deleted by Moderator9 I am saving a copy of it to send straight to Mr Baker himself when said move or deletion acures. Have a great day all. Come Join us at Hawaiian Beach Bums |
|
Moderator9 Volunteer moderator Send message Joined: 22 Jan 06 Posts: 1014 Credit: 0 RAC: 0 |
"Beach Bum" I never said that some version of the code was not available somewhere. But the plain fact is that the current BOINC version is not. Dr. Baker has confirmed this fact. In no way does that say the project plans to, or even could profit from it in the future. In fact quite the opposite is the actual case. The code will be released in the normal course of the research. All I said was that fines for misuse are not any real protection against misuse if the gains for doing so are high enough. The actual number is irrelevant. People violate non-disclosure agreements every day. I also said that the models you propose are different that than what has been asked in this case. I also pointed out that with the other BOINC projects the situation is very different, and it is. I am open to other views if someone can explain how providing the source code for a computational research project to a single individual in Germany, is comparable to corporate agreements with Microsoft or any other large company, and how this or any other project might act to prevent misuse under those conditions. That is the actual question that was asked. As to the rest of your comments, I do not flame people. That kind of behavior is a waste of time. If you took it that way, you misread it. If you are going to construe views that differ from your own as flaming, then further discussion of this matter will be fruitless. As has been noted, the posts in this thread were moved from another thread where they were in fact viewed as off topic by the owner of the thread. You of course know that to be the case, and if you didn't you do now. This discussion did not even fit the thread title. The posts are not off topic for THIS thread, and neither I nor anyone else has said they are. If you read the posts from the original thread you will see that prior to your posting there, the owner of the thread had asked that this discussion not continue there, as it was off topic. I view that as a strong indication that additional discussion after that point might be viewed in a similar light, and others did as well. To preserve the discussion, but refocus the original thread onto its original topic, the owner asked that ALL the posts above a certain point in the original thread be moved. Let me be very clear on this point. This was done at the request of the owner of the thread. All I did was honor the request as I would for the owner of any thread in these forums. I have done it before and I will do it again. This is apparently the source of your confusion, but it is all well documented. If you have a complaint, please do send it in. Please do not wait for an event that will never come, before you do. Dr. Baker is already aware that this discussion was moved as he received the same message as all the other participants, including you. Since moving things around to keep them in proper order is a big part of what he has asked me to do, I am certain he will welcome your thoughts on mixed topic threads, and my response to user support requests. As a personal note, before you accuse me of flaming or deleting posts such as the one to which I am replying simply because they express disagreement with something I have done, you should know that has never happened, and Dr. Baker knows that to be the case. So let me offer this advice for structuring your complaint. You might want to consider that It will add credibility to your comments if you try to stick to the actual facts. If you lace your letter with claims of actions where history and the facts do not support your position, you will adversely affect the possibility of serious consideration of what you have to say. You should also avoid predictions of what you think I MIGHT do, and stick to what I have ACTUALLY done. This will make your comments seem less like baseless ranting and more like adult exchange of ideas. Let me know if you have trouble sending it in, I can provide contact information if you need it, but there is a link for contacts from the home page. Moderator9 ROSETTA@home FAQ Moderator Contact |
|
Dimitris Hatzopoulos Send message Joined: 5 Jan 06 Posts: 336 Credit: 80,939 RAC: 0 |
Folks, I'm just catching up with previous discussions here. Ofcourse I agree that "the more platforms/OSes supported, the better". But let's also keep in mind that even IBM's World Community Grid (WCG, the grid running Human Proteome Folding and FightAIDS@home) doesn't support platforms besides Win and Linux/x86. And in the case of WCG, we're talking about IBM, the biggest IT company in the world, which has the resources, know-how and *marketing incentive* to support its OWN platforms (AIX and/or POWER Risc) even if they represent <5% of overall TeraFLOPS. PS: I still think the 3 most pressing priorities are: 1/ installer for Win and Linux 2/ code a BigWU flag in BOINC server code and 3/ have some code-wiz to optimise the science app (like akosf did for Einstein speeding it up 4x -four-fold-) Best UFO Resources Wikipedia R@h How-To: Join Distributed Computing projects that benefit humanity |
Message boards :
Number crunching :
Porting Rosetta - Unix, Linux etc
©2024 University of Washington
https://www.bakerlab.org