中国分布式计算论坛

 找回密码
 新注册用户
搜索
查看: 2213|回复: 0

[新项目] [BOINC] [数学类] RieselSieve [已失效]

[复制链接]
发表于 2007-1-13 08:52:05 | 显示全部楼层 |阅读模式
2007-10-28 01:01:35 UTC   Riesel Sieve Finds its First BOINC Prime!
Congratulations to Mackerel, who had a box turn in our first BOINC Prime this evening. The k is 469949, a relatively light k at about 29,000 k/n pairs removed. However, the test was a missed test on 1st run, highlighting the need for a proper double-check by BOINC.
We will release the full k/n pair after it is submitted to PrimePages, but it does weigh in at just over 496,000 digits. Is there another double-check prime out there? Only way to find out is to run LLR and help us hunt down the remaining 66 k's. Thank you to all the participants for helping us bring down the workload, and a special thanks to Mackerel for killing his second k.
Manual sieveing DAT file will be released in the morning; the new DAT for BOINC will be released as part of our server switch.
恭喜 Mackerel 发现了我们在 BOINC 上的第一个质数。这个指数的 k 是 469949,相对来说比较“轻量级”,只有 29000 对“k/n”对。然而,这个质数在第一次检查的时候溜过去了,这就说明了 BOINC 上的复查是很重要的。
在提交到 Prime Pages 之后,我们会公布这对 k/n 对。但是这个质数只有大约 496000 位。还有没有别的溜过去的指数呢?要想知道,唯一的办法就是帮助我们搞定剩下的 66 个 k。感谢您的参与和帮助我们减低工作量,特别鸣谢 Mackerel,他干掉了他的第二个 k。
手动筛法的 DAT 文件会在早上发布,BOINC 的 DAT 文件会作为服务器升级的一部分发布。

2007-10-21 19:41:34 UTC   Status Update 10-21
Lee and I wanted to post a notice and let everyone know that things are running well - so, we're going to start making some changes.
k=141941 is about 30 tests away from being tested to n=2^22 - far beyond the other k's we are testing. Unfortunately, a prime has not yet been yielded. We are examining where we want LLR to go next. For the time being, the shorter, double-check workunits are being run.
In the next week or so, we will be migrating from PerlBOINC to 'true' BOINC. While we do not anticipate any major problems or items of notice to the users, there may be small periods of downtime while we switch. We'll make every effort to post them in the forums beforehand. Some extra features, like the sieve & llr range pages, may not be updated or available during the transition.
We are also planning to release all new clients when we switch to the new server code. Linux 64-bit sieve will return, and we will be upgrading all the sieve clients to the latest versions of sr2sieve, with the goal of an additional speed improvement. We also will release a new DAT file for sieve at that time. LLR will receive a new wrapper, with the hopes of finally nailing the dreaded queue crash bug.
All in all, it should be a busy (and much more exciting) next couple of weeks - stay tuned for details!
我和 Lee 想在这里放个通知,让大家知道一切顺利。而且我们正在准备做出一些改进的东西。
k=141941 现在只差 30 个测试就会到达 n=2^22 了,这比其它所有的 k 测试得都要远。不走运的是我们还没有发现一个质数。我们正在考虑我们下一步 LLR 测试的计划。现在运行的是复查的比较短的工作包。
在下个星期左右我们就要从 PerlBOINC 移植到真正的 BOINC 了。我们虽然不希望出现什么引起用户注意的问题,但是在我们转移的过程中服务器可能会时断时续。我们会尝试在这种情况发生之前在论坛上通知大家。有一些功能在转移过程中可能会暂时失效,如 sieve & llr 范围页。
在转移后,我们也准备发布新的客户端。Linux 64-bit 筛法程序将会卷土重来,我们也会用最新的 srsieve 程序代码来更新所有的筛法程序,这样会提高速度。到时我们也会发布一个新的 DAT 文件。LLR 会有一个新的 wrapper,希望这次能解决队列崩溃的问题。
总的来说,下几个星期会是忙碌但令人兴奋的。等着我们的消息!

2007-09-07 22:44:07 UTC   Linux Sieve Problem
We are currently experiencing a major problem with the Linux sieve client, which is affecting about 90% of workunits sent out to linux boxes.
We are actively working on the fix; in the interim, we have switched all linux clients to LLR work (regardless of beta app settings), as the linux LLR client has been incredibly stable.
As an aside, we hope to have a new client out for all platforms this weekend (this was planned), as we plan to release an updated DAT file for sieveing. More details will be posted.
For updates on the Linux sieve problem, please see the forums.
我们的 Linux 客户端出了一个严重的问题,影响到大约 90% 的 Linux 工作包。我们正在努力修复这个问题,在这个过程当中我们已经将分配给 Linux 机器的工作包全都调成 LLR 工作包(忽略是否接受测试程序的选项),因为现在 Linux 的 LLR 程序还是稳定的。我们希望在这个周末能够在所有平台推出一个新版本的客户端,顺便带上新的 DAT 文件(这是在计划中的)。我们将会公布更多的细节。在我们的论坛上有这个问题的更新消息。

2007-07-11 11:42:46 UTC   LLR Focus on k=141941
Lee and I are determined to defeat k=141941, our largest k. He feels that there was a missed prime somewhere for this k, so we're going to turn all of our attention - LLRNet AND BOINC LLR to this one k. LLRNet will go up to 5 million, BOINC LLR will do the re-check from the current level (about 1.5M) to 3.0M.
To rally the troops, all Linux BOINC clients will get LLR work (we're confident that the scoring has been balanced finally, but will still monitor closely). We are also encouraging windows users to switch over to LLR if possible.
LLRNet has about 4000 pairs to do before the k=141941 assault truly begins, and BOINC has about 1400. Here's to slaying another k!
我和 Lee 决定了要挫败 k=141941 这个最大的 k 值。他觉得对于这个值应该有一个属于它的质数,所以我们现在将我们 LLRNet 和 BOINC LLR 所有的注意力都集中到这个 k 上面来。LLRNet 会计算到 n=5M,BOINC LLR 会进行从现在的 1.5M 到 3M 这个范围进行复查。为了打赢这场战役,我们决定让所有 Linux BOINC 的客户端都能接收到 LLR 的工作包(我们相信得分已经被平衡好了,但是我们仍在仔细监控着)。我们也鼓励 Windows 用户尽量转向接收 LLR 工作包。LLRNet 现在还有大约 4000 对数才能开始对 k=141941 的攻击,而 BOINC 则还有大约 1400 对。现在让我们来搞定这一个 k 值吧!

2007-06-26 18:26:09 UTC   Linux LLR Now Available
LLR testing is now available on Linux. Make sure that you have "Allow test applications" enabled for your host and you will receive LLR workunits. If you have questions, please visit the forums.
现在 LLR 测试在 Linux 上也可以进行了。你只要选择“允许测试程序”就可以收到 LLR 的 WU 了。如果你有任何问题,请到论坛上提问。

2007-06-10 14:49:47 UTC   Work Problems June 10th
This morning, a bad sieve range was fed into the work generator. This created about a 6-hour period of repeatedly bad work.
We have fixed the work generation, although workunits will now be 1G for sieve, which is 2x as long as the "old" workunits. We have adjusted crediting to account for this.
The latest info regarding the problems that were created and the current status on the servers can be found in the forums.
今天早上,一个错误的筛法范围被扔进了 WU 生成器,然后它就产出了以 6 小时为循环的坏工作。我们已经把它修好了,而且 WU 被延长了 1 倍。我们已经由此修正了得分方法。在论坛中有详细的信息。

Mon, 14 May 2007 23:39:37 GMT  New LLR Client Version
We're pleased to announce the 5.20 series of LLR clients. This is the first in a series of upgrades & updates we plan to make in the next 1-2 weeks. This introduces the new wrapper format to the Windows OS - it has worked extremely well on Linux, so we're hoping to duplicate those successes. Internal testing over the weekend went well. Two main breakthroughs with this new wrapper: (1) no more 0.00 CPU time LLR WU. (2) (hopefully) No more random queue crash for LLR WU. As long as (2) does not creep up this week - please monitor your hosts - we will move forward with our plans to make LLR more of a project-wide priority.The LLR executable itself stays the same, we are only changing the wrapper and introducing a job file.Please make sure that you notify us of any queue troubles or other issues with the wrapper (or just drop in and tell us how great it is). Thanks in advance!
我们很高兴我们发布了 5.20 版的 LLR 客户端。这是在接下来的一两个星期的一系列的更新之中的第一个。这个客户端将在 Linux 上能良好运行的新的 Wrapper 引入了 Windows 系统,我们希望能再次取得成功。整个周末进行的测试很顺利。新的 Wrapper 有两个好处:(1) 不会出现 CPU 时间是 0 的 WU。(2)(我们希望)不再出现随机的崩溃。因为测试中我们没有遇到崩溃,所以如果遇到这种情况请汇报。我们计划将 LLR 提升到与 Sieve 同样的地位。在这次升级中,LLR 程序本身不变,我们只是改变了 Wrapper 和添加了一个新的 job 文件。如果遇到崩溃情况,请通知我们。提前谢谢了!

Thu, 10 May 2007 07:19:13 GMT  New Linux Clients
A new x86 and x86_64 set of linux clients has been sent out to the hosts. It includes the new DAT file, updated sr2sieve, and a new wrapper format that will allow us to interchange what is run in the wrapper more easily. Look for the PPC build to come out this evening. As always - send any questions over to the forums.
新的 x86 和 x86_64 Linux 平台上的计算程序已经投入使用,它包含了新的 DAT 文件,升级版的 sr2sieve 和一个令交换信息更方便的新的 wrapper 格式。PPC 平台上的程序今晚就能出来。如有问题,请到论坛提出。

Mon, 30 Apr 2007 00:32:50 GMT Prime Found - and a New DAT Version
Congratulations to Cipher, who found his 2nd prime with the Riesel Sieve Project. While not a BOINC prime, this LLRNet beast weighed in at #19 in the world - 794,556 digits long. 342673*2^2639439-1 is indeed prime. Thank you to everyone that has made this prime possible. I'm also pleased to announce the 5.35 release of the Windows client. This is a real bugfix, as watzkej has been gracious enough to spend most of the evening in #rieselsieve debug testing things, and Chuck has helped narrow down all my bad coding problems (that he could find). We're on track again with this 5.35 release and should see a speed increase - both from the smaller dat (about 200,000 pairs, between the Sieve effort and removing 1 k) and the use of sr2sieve. Thanks for sticking with us this last week - it has been quite a hectic time for all of us.
恭喜 Cipher 发现了他在这个项目里发现的第二个质数!这个由经典客户端发现的“怪物”有 794556 位,世界排名第 19。它就是 342673*2^2639439-1。感谢所有参与者,是你们使这个质数得以被发现。我在这里顺便宣布一下,我们发布了 Windows 计算程序的 5.35 版。这回真的是一个可用版本了。感谢 watzkej 在帮助调试中度过一个个不眠之夜,感谢 Chuck 找出了他所有能够找出的错误。现在我们正在监控 5.35 版的运行状况。这个程序应该会更快,因为我们有一个更小的 dat 文件(经过筛法和移除一个 k 后大约剩下 200000 对),而且我们使用了更快的 sr2sieve。谢谢在这个星期对我们的支持,这个星期发生的一切的确让我们每个人都很尴尬。

2007-04-28 12:10:08 UTC   Back to 5.31 Sieve Temporarily
Temporarily, we've had to go back to the 5.31 sieve version. We cannot get 5.32-5.34 working correctly, so 5.31 is the "last working" version. It does not checkpoint, and it is actually the checkpoints themselves causing the problems in 5.32-5.34 I've found. Once I have this hammered out, we'll be back on track.
暂时我们得回到 5.31 版的程序了,更新的版本工作不正常,所以 5.13 版是最后一个可工作的版本。这个版本不会存盘,而我发现正式存盘的部分使更新的版本不能工作。只要我能把这东西捣鼓好,我们就第一时间回到正常的轨道上来。

2007-04-24 23:22:18 UTC   New Day, New Client for Sieve + Good News
Version 5.34 of the Sieve has been released. I'm hoping this is the last release required to stomp out this loop bug that is plaguing the sieve. I'm already starting work on a pseudo-wrapper that will only have one executable file, limiting how much has to be updated when we need to make changes. I also hope that it will bring us closer to a true native BOINC app.
On the good news side of things, a possible prime is undergoing internal verification - it would be the 19th largest in the world if it comes back prime. A prime would cause the release of version 5.35, with a new DAT, later this week.
Thank you all for your help and patience these last two weeks - the variety of problems has kept us all very busy. In the end, I promise, things will be for the better.
5.34 版本的程序已经推出了。我希望这次能解决问题。我正在试图将 wrapper 跟筛法程序绑在一起,这样的话我们更新的时候就省事多了,而且也会使我们的程序更接近于原生的 BOINC 程序。
另外有一个好消息。我们发现了一个疑似的质数!我们正在检查它的素性。如果证实的话,它将会是现在第 19 大的质数,我们也会推出新版本,换用新的 DAT 文件,大概就是这星期内。
谢谢你们在这两个星期来的帮助与配合。各种问题的出现把我们弄得团团转,但是到最后,一切东西都会好起来的。

2007-04-24 00:06:30 UTC   Abort 5.32 Sieve WU on Windows
Users with 5.32 series WU on Windows should abort the WU. The 5.32 client does not properly report its status as "finished", which causes an endless loop for the sieve. Version 5.33 has been released which fixes this problem. Please visit the forums for more information.
Windows 用户请终止掉所有用 5.32 版程序计算的 WU。5.32 版本的程序不能正确地报告它的状态为“完成”,这会导致一个死循环。5.33 版本的程序已经推出,用以修复这个问题。详情请访问我们的论坛。

2007-04-22 23:01:07 UTC   New Sieve App Release for Windows
-------Windows/x86 & Windows/x86_64 Verison 5.32-------

Known Bugs: We don't have the same checkpoint logic as the linux apps, as Windows doesn't support process signals. We are working on a true solution. Right now, checkpoints will occur every 5 minutes.
Having "leave apps in memory while suspended" caused the app to hang in 1 trial (out of 4, on the same box). I'm looking into it. Similarly, in certain cases, the sieve app did not always get properly closed when boinc was being closed.
Bugs Fixed: Problems associated with the 5.31 release - including no progress reporting, some app suspension handling, etc. Fixes handling of 1000T+ WU properly. Note: this is the first windows app to include sr2sieve. We will be putting out a new set of optimized sieve apps, catered to different architectures. This is the athlon-tuned, SSE2-enabled (when CPU has it) edition.
A true 64 bit client is coming - I have to get mingw running 64-bit first.
我们有面向 Windows 的,32 位和 64 位的新的筛法程序了!
已知 Bug:这个程序的检查点逻辑跟 Linux 版本的不同,这是因为 Windows 不支持进程信号。我们正在努力尝试解决这个问题。现在,程序每五分钟做一个检查点。
BOINC 的“暂停时驻留在内存”选项会让程序有大约 1/4 的概率挂掉。我正在检查是什么问题。同样,有时 BOINC 关闭时程序的行为也不大正常
修复的 Bug:所有 5.13 版的问题,如没有进度、暂停问题等都已经解决了。程序已经可以正常处理 1000T+ 的任务了。注:这是首个包含 sr2sieve 的 Windows 程序。我们会放出新的一系列的优化程序。现在的版本是为 Athlon 调节过的支持 SSE2 的版本。
真正的 64 位程序就要来了——等到我将 mingw 搞得能运行在 64 位模式下就可以了。

2007-02-26 22:11:25 UTC   Connectivity to Asia & Australia Restored
We've been receiving confirmation from users that connectivity to Asia, Australia, and the Pacific have been restored as of this afternoon/evening. Thank you to the numerous people, especially AMDave, Webmaster Yoda, and others, who helped with tracerts, pings, and figuring out where exactly our link to the Internet was having problems. Sorry about the long delay, but it took a lot of testing/convincing to get things fixed.
我们已经收到亚洲、澳大利亚和太平洋用户的确定能够重新访问我们的消息。感谢许多许多人,特别是 AMDave、Webmaster Yoda,还有别的人帮忙探测故障处。我们对这个延误深感抱歉,但我们的确需要很多测试才能解决这个问题。

2007-02-21 00:25:09 UTC   New x86_64 Application Version
Due to some bad compilation flags, the 5.33 application version for x86_64 is excruciatingly slow (~20 hours/WU). We've released a new version that has the correct flags in the sr2sieve app, and the speed that should be there has returned. Updates are pushing out to the x86_64 clients now. This did not affect our other clients.
由于搞错了编译开关,x86_64 的 Linux 版程序极其缓慢(~20 小时/WU)。我们已经用修正过的新版本替代了原来的程序,这样的话速度就可以重新上去了。现在这个新版本的程序正在被发放到每台 x86_64 的机器上。其它版本的程序不受影响。

2007-02-19 19:29:13 UTC   Another Linux & PPC Release + x86_64 (Linux) App
A bug has been found in the Linux & PPC apps. So, we've released new version of the Linux/x86 app + the Linux/ppc64 app, which will correct this error.
We also have, thanks to Bok from Free-DC, an x86_64 Linux app for the first time. Instead of a 32-bit app, we're sending a 64-bit version of sr2sieve + the wrapper.
As always, please direct questions/comments to the forums.
在 Linux 和 PPC 程序中发现了一个错误,所以我们发放了新版本的 Linux/x86 和 Linux/ppc64 的计算程序,这将会修正这些错误。
感谢来自 Free-DC 的 Bok,我们终于有了第一个 x86_64 的 Linux 计算程序。我们现在已经用这个 64 位的程序替代了原先的 32 位的程序了。
如果有意见或者建议的话,请到论坛提出。

2007-02-17 02:04:58 UTC   New Linux Sieve Version + PPC/Linux Support Added
It's been a busy, but exciting, night at RieselSieve. We finally have a new Linux/x86 version that correctly reports its timing & status to the BOINC client. It also implements sr2sieve, rather than proth_sieve, so users should see a HUGE speedup.
We also have introduced a Linux/ppc64 client. This will download from our server automatically, as long as you have compiled BOINC from source on your box. We plan to have a binary available soon. The new, from-source build is necessary because of the version string, powerpc64-unknown-linux-gnu, that it uses. Our server matches that & sends out a TRUE PPC sr2sieve version & wrapper. This also has the corrected linux progress & checkpointing capabilities. Thanks to BlisteringSheep for his help & access to a PPC box for us.
As always, please visit the forums to notify us of bugs/changes.
对于 RieselSieve,今晚是个忙碌而又令人激动的晚上。我们终于得到了一个新的可以向 BOINC 客户端准确汇报时间和状态的 Linux/x86 的客户端。我们还在其中用 sr2sieve 替代了 proth_sieve,得到了一个巨大的速度提升。
我们也写了一个 Linux/ppc64 的客户端,不过要在你自己的机器上编译。很快就会直接有能自动下载的程序了。这个版本也包括了关于进度和检查点的改进。感谢 BlisteringSheep 的帮助和提供的 PPC。
如果有意见或者建议的话,请到论坛提出。

2007-02-04 11:20:56 UTC   New LLR Beta Application Version Released
We've brought out a new version of the LLR Beta App that should permanently prevent the "popup window" that has surfaced when a WU crashes in certain situations.
Because of an error in the XML for the 5.18 version, please abort any 5.18 version WU you get, but allow the 5.19 version to crunch. This should stop the signature verification errors that happened overnight. If you have any questions, please stop by the forums.
新的LLR程序已经发布,相信它可以一劳永逸地解决WU计算失败时弹出窗口的Bug。由于在5.18版本的XML中有一个错误,所以请终止所有用5.18版本计算的WU而去计算5.19版本的WU。这样的话应该可以避免像昨晚发生的的签名验证错误。如果有什么问题的话,请到我们的论坛。

2006-12-02 17:19:10 UTC   A New Prime & 5.30 Sieve Released
Congratulations to DarkStar, who found the 742167-digit prime 26773*2^2465343-1 using our "classic" LLR application on Friday. We have verified and submitted the prime, which ranks #20 in the world.
As a result, a new, lean, mean riesel.dat is available for download, which requires the release of a new wrapper file as well. 5.30 for windows is being sent out as we speak. There's about 150,000 fewer pairs in the DAT file, thanks to the removal of 72,000 from the k and another 75,000+ from BOINC.
Thanks to everyone supporting Riesel Sieve, and here's to another prime!
祝贺 DarkStar,他用“经典”客户端发现了 26773*2^2465343-1 这个 742167 位的质数!我们已经确认并申报了这个质数,这个质数是暂时发现的第 20 大的质数。这样,我们就通过清理约 150000 对 k/n 对而得到一个更精炼的 riesle.dat 文件,其中有约 72000 对归功于“经典”客户端,另 75000+ 对归功于 BOINC。谢谢大家!

2006-11-19 21:49:35 UTC   LLR Test Application Now Available
We have enabled a test application using LLR. In order to crunch LLR workunits, you will need to edit your preferences for Riesel Sieve and "Allow test applications". You can set only P4 clients to run the test application using a venue (home, work, or school), and then they will run LLR as long as work is available. We plan to have processor detection/selection in the near future, but the venues will serve as a temporary workaround.
If you have any questions/problems with the new LLR application, please visit the forums.
我们已经启用了一个 LLR 的测试程序。为了启用此程序,请在参数设置页面上设置为“允许测试程序”。您只能用 P4 来运行这个程序,所以请做好设置,建议利用“计算机位置”的分开的设置做到这一点。我们将来会加入识别处理器的部分。如果您在使用新的 LLR 程序的时候遇到问题的话,请访问论坛。

2006-10-28 13:17:40 UTC   Optimized Sieve Applications Now Available
We have now published a list of EXE files that are official optimized sieve applications depending upon CPU features - CMOV5, CMOV7, SSE2, or non-CMOV. The basic application, CMOV6, is most common, which is why we have ran it so far.
These applications will speed up the WU processing, allowing a user to complete more WU per day, and in turn increasing their overall production/score. For details on how to install and the EXE downloads, please see this forum thread: Official Optimized Clients
我们已经发放了一系列针对不同 CPU 特性的正式优化程序:CMOV5, CMOV7, SSE2 和 非-CMOV 版本。我们现在使用的默认版本 CMOV6 是可以最普遍地提升性能的,所以我们直到现在都是利用这个程序进行计算。这些优化程序可以提高计算机处理 WU 的速度,从而增加获得的分数。关于如何安装优化程序的细节,请参看:Official Optimized Clients

评分

参与人数 1基本分 +200 维基拼图 +100 收起 理由
霊烏路 空 + 200 + 100

查看全部评分

您需要登录后才可以回帖 登录 | 新注册用户

本版积分规则

论坛官方淘宝店开业啦~
欢迎大家多多支持基金会~

小黑屋|手机版|Archiver|中国分布式计算总站 ( 沪ICP备05042587号 )

GMT+8, 2021-9-21 14:01

Powered by Discuz! X3.4

© 2001-2017 Comsenz Inc.

快速回复 返回顶部 返回列表