下一周就开始放包出来了？

现在已经有任务包了，但数量比较少，而且单个任务包的完成时间比较长，主流机器60-80个小时吧，现在的计算程序是1.13版，之前的1.12在读取计算进度的时候会出错。

现在应该win/linux/mac平台都可以算。

This completes our effort of porting the main applications to the three main platforms. Now it's time for application enhancements, bugs fixing, and regular WUs.

据说最近放出来的几千个都是小包，我的t5600上，7分钟完成了5%，应该最后所需时间比它说的52小时要少很多。

任务包占内存大概63-90兆，目前没屏保。

那为什么我的没接到包呢？
我的可是24小时在线啊

项目还没有服务器状态页面，不清楚还有多少包。

我这边的三个任务包都正常完成了，但看过同时在算的其它用户的计算情况后，似乎win2k,win98,darwin都有过出错的情况。

http://orbit.psi.edu/oah/forum_thread.php?id=64

Posted 20 Apr 2008 2:08:24 UTC

After a batch of ~4000 WUs here's the situation:

在大约4000个任务包的测试完成后：

win32: the SurveySimulator app seems to run correctly on XP, but fails on win98 or earlier; this is due to a compiler/system limitation. We cannot at this moment support win98 or earlier systems, and we're looking for a way to prevent the submission of WUs to these hosts.

项目将不支持win98及更早的系统。

linux: we're still trying to understand if there is any serious problem with the linux app. WUs seem to complete successfully, but sometimes fail to validate and/or get credit.

linux下的计算结果有时会不能通过验证，正在查找原因。

macos: app version 1.14 still doesn't run on Mac OS 10.4 or earlier. An updated version, likely 1.15, it's in the works to try to fix this.

mac平台，1.14版的计算程序不能在macos 1.04或更早的系统上运行，1.15版试图解决这个问题。

homogeneous redundance: we'd rather not enable it. We want our WUs to run in very different environments, because 1) the app is supposed to produce nearly identical results on any platform; 2) differences help us spot bugs and limits of the app. Every problem spotted by non-homogeneous WUs is likely to produce improvements in the app.

不准备引入homogeneous redundance，因为我们希望计算程序在任何平台下得到的计算结果都是相同的，如果有不同，能够帮助我们定位bug和了解程序的一些不足。

error limit: we're increasing it to 5, to give more chances to the same WU to get computed correctly. We might lower this limit in the future, as the app gets more stable.

目前任务包的错误结果上限为5，以后程序变得更稳定之后，应该会适当降低这个限制。

http://orbit.psi.edu/oah/forum_thread.php?id=67

Posted 20 Apr 2008 11:44:57 UTC

We've just released version 1.16 for win32 and macosx/intel, along with more test WUs.

刚发布了1.16版计算程序，用于win32和macos/intel平台。

win32: it looks like support for windows versions older than XP can still be possible. With 1.16 we're testing a quick solution suggested by Der Meister on the boinc_projects mailing list. If this doesn't work, we'll revert to the previous version.

windows平台，貌似有可能支持xp之前的操作系统版本了。

macosx/intel: we continue to appreciate the beautiful complexity of this platform. It's a joy to spend hours trying to track down linking problems between different feline versions of the Apple. And all the compiler and linker options available can only add to this pleasing experience. Version 1.16 should work on Mac OS 10.5, and possibly on older systems.

macosx/intel平台，说这个平台具有beautiful complexity，不知道是不是反话。。。1.16版本应该能工作在10.5版本上，更早的版本也有可能。

http://orbit.psi.edu/oah/forum_thread.php?id=68

Posted 20 Apr 2008 22:14:40 UTC

macosx: app version 1.16 still has problems, and seems to work only only on Mac OS X 10.5. We still have a couple things that we want to try to make it work, but for now we're removing it from the apps list. We'll get back to it later, when time permits. Volunteer developers welcome to take from where we left and test the app on 10.4 or earlier.

macosx平台：1.16版本仍然有问题，似乎只能在osx10.5上工作，这个问题得稍晚些时候再来解决。

linux/win32 validation problems: we're focusing on this now. We've just generated more WUs that are on average 10 times shorter than previous ones, and enabled the generation of more debugging information on the server, to try to track down the problem.

linux和win32平台的验证问题：这是我们当前主要的工作内容，为些特意生成了比之前的任务包还要小上10倍的任务包，以方便查找问题。

win32: older v1.13 seems to perform a bit better than latest v1.16: 78% vs. 68% success rates. But this is raw, we need to get into the details of what goes wrong with each single host. As an example, a single host returning errors can return a lot of errors, biasing the success rate, while it should only count as 1 or so...

win32平台：1.13版本似乎比1.16版本表现更好些，主要是在任务的成功率方面。但这还需要进一步的比较才能确切地判断问题所在。

linux: v1.13 success rate is about 97%, pretty good.

linux平台：1.13版本的任务成功率在97%，还不错。

说错了。。。状态页还是有的：
http://orbit.psi.edu/oah/server_status.php

Posted 22 Apr 2008 11:26:02 UTC  

This update (win32 and linux) should fix the validation problems across different platforms. These problems were mainly due to incorrect checkpointing (some data was not saved for future restore) and a bug into a caching system used to optimize the code. Our local tests show that both problems are now solved, so we've generated more test WUs for more testing. Since the problem is related to checkpointing, the test WUs are not as short as the previous ones, but about 10 times longer, to give more hosts a chance to suspend the computation and test the fix.

1.17解决了不同平台间的结果验证问题，问题的主要来源是存盘操作有遗漏以及优化代码所用的缓存系统的BUG。

因为要测试存盘，所以任务包不会太小。

Posted 23 Apr 2008 2:14:28 UTC

Version 1.18 is out. This fixes one more bug that slipped into 1.17 and previous versions, related to how win32 handles "new lines" (marked by two special characters) as opposed to linux (only one character). Interestingly, Mac OS X does also use one character, but not the same as Linux. This will ensure more fun when we will try again to release an app for macosx.

1.18版本又解决了一个换行符在不同平台上表示不同引入的问题。

Using a smarter validator could have solved this problem as well, but for now we prefer to stick to the bitwise validator and force result files to agree byte by byte on any platform.

Since we've now clearly run out of bugs, 1.18 will also be referred to as the "good one", bringing peace and prosperity among different platforms.

到目前为止，所有发现的bug都已解决。

All previous WUs have been marked as canceled because previous apps clearly had problems, and fresh new WUs are on the way.

所有的旧任务包均已被取消，新任务包也已上路啦~~

新任务包也已上路啦~~
?????

呵呵，就是为1.18版计算程序准备的任务包，消耗得挺快，上午贴新闻时还有好几千个，刚才看就没了，我就算了两个

来迟了、来迟了......

呵呵，不迟，现在项目刚起步，任务包也很少：）

我怎么就分不到呢？