中国分布式计算论坛

 找回密码
 新注册用户
搜索
查看: 18908|回复: 60

[分享] APB1 子项目 CUDA 版本开始测试!

[复制链接]
发表于 2009-8-11 10:28:52 | 显示全部楼层 |阅读模式
http://einstein.phys.uwm.edu/forum_thread.php?id=7513

Posted 3 Aug 2009 15:50:42 UTC
Last modified: 7 Aug 2009 10:12:27 UTC

The first Einstein@home CUDA App for Windows is available for Beta Test at Beta Test Page.

windows版本的下载页面:http://einstein.phys.uwm.edu/app_test.php#windowsCUDA
linux版本的下载在同一个页面的下面。

This is a big package to download (17MB), as it contains both applications (S5R5 and ABP1).

文件大小17M,包含两个子项目的计算程序。

We are still new to writing app_info.xml files for CUDA applications. The one contained in this package should allow to run one ABP1 task per CUDA device, occupying a full CPU core for support, and run S5R5 tasks on the remaining CPU cores.

GPU用来处理ABP1的任务,同时也占用一个CPU内核,剩下的CPU内核可以同时处理S5R5的任务。

There are no entries for older than the latest App versions 3.05 / 3.07 in the app_info.xml, tasks already assigned to older Apps will error out.

不支持3.05和3.07之前的任务,如果对这些旧的任务使用优化,会直接出错。

The ABP1 CUDA App has undergone limited testing with a BOINC CC 6.6.36 and a 181.20 NVidia Driver on a GTX285 system. We can confirm that the App doesn't crash immediately on such a setup, but not much more. We already ran into some mysteries of the CC scheduler, though.

项目方的测试环境。

Again, this is our very first experience with CUDA Applications, please bear with us and don't expect everything to work right away.

第一个版本,可能问题比较多,使用时做好心理准备。

Please test and report, and please include important information (like the NVIdia Driver and Core Client version) in your posts.

BM
 楼主| 发表于 2009-8-11 10:39:36 | 显示全部楼层
Posted 4 Aug 2009 12:33:55 UTC
Last modified: 4 Aug 2009 12:40:24 UTC

Hi everyone,

I'd like to give you a bit of background information about thisrelease. First of all we follow the "release early, release often"approach WRT CUDA. The reason for this is that it's one thing to getCUDA code running in a known environment, but a significantly differentstory when it comes to the heterogeneous outside world and BOINC.Therefore this beta release serves as a first test of the whole buildand release chain as well as to identify potential BOINC-relatedproblems. Lessons learned can then be used to simplify the deploymentof upcoming, improved CUDA releases of the ABP1 radio pulsar search,but also - even more important - the upcoming CUDA version of the S5hierarchical search.

初始版本问题可能会比较多,得来的测试经验将用来改进ABP1的新CUDA程序,以及更重要的,S5的CUDA程序!

So far we know that...

  • The ABP1 CUDA release yields a performance improvement of factor~2 (our own tests on Windows, Linux and Mac OS X) as we used the mosttrivial approach which still runs 30%-50% on the CPU
  • We claim a full CPU core for the application because BOINCwon't reset the task to idle priority when less than one core is claimed
  • On-board devices typically can't handle this application due to lack of memory
  • Only CUDA 2.1 seems to work reliably (2.2 might cause segfaults, consistent with findings at GPUGrid)
  • This is release is targeted for 32 bit systems
  • The BOINC client's internal task scheduler seems to have problems with CUDA device identification and task distribution
ABP1的CUDA版本大概提速一倍,同时仍需要占用30-50%的CPU。
板载显卡一般是不能用来跑这个程序的,主要是内存不够。
目前似乎只有CUDA 2.1可以可靠地运行,2.2可能会出错。
目前只有32位版本。
BOINC客户端在CUDA设备的识别和任务调度上似乎还有问题。


Thanks for supporting these tests! Please stay tuned for more as we'reworking hard on getting a first beta release of the hierarchical searchas well as improved CPU and GPU versions of the ABP search.


感觉大家的支持!


Cheers,
Oliver
发表于 2009-8-11 14:08:12 | 显示全部楼层
9300M唉
以后有了台机再说吧
发表于 2009-8-11 16:12:29 | 显示全部楼层
一想到 AQUA 用 200 分骗走我 200 小时的 GPU time。。。。。
发表于 2009-8-11 22:22:37 | 显示全部楼层
E@H的CUDA终于出现了!!!
发表于 2009-8-12 00:08:16 | 显示全部楼层
有任务可以下载来做了吗???版主也说一下怎么设置嘛~
 楼主| 发表于 2009-8-12 09:53:32 | 显示全部楼层
我也没有能算的卡。。。

总归是把测试程序解压到项目文件夹吧,然后boinc客户端版本、显卡驱动版本配套好应该就问题不大了
发表于 2009-8-12 14:11:33 | 显示全部楼层
用CUDA算的E@h要占用一个CPU外加CUDA,不爽。seti@home只占用0.04CPU。
 楼主| 发表于 2009-8-12 14:57:57 | 显示全部楼层
项目方也有说明的,目前这个还是一个比较初级的版本,只是把部分计算任务放在了gpu那端,所以对cpu的使用还比较多

主要还是积累gpu的使用经验吧
发表于 2009-8-12 15:28:57 | 显示全部楼层
我看项目方说明,好像对显卡的要求较高,要260+以上的。像我这9800GT好像差点
 楼主| 发表于 2009-8-12 21:26:03 | 显示全部楼层
也不是要求,主要是说项目方只在这个环境下进行过有限的测试。
发表于 2009-8-13 00:23:28 | 显示全部楼层
目前只用CUDA算FFT,应该对显卡要求不高吧~
有空试试我的128M 8600M GT能不能算~
发表于 2009-8-15 01:10:20 | 显示全部楼层
版主说说咋搞啊。。。程序在哪下?我的260+应该可以跑啊
 楼主| 发表于 2009-8-15 10:35:42 | 显示全部楼层

回复 #13 supertoby 的帖子

参见7楼。。。我手头最好的显卡也还是n年前的9550

另外,最近项目服务器停机,也下载不到任务,等恢复正常后再试吧。
发表于 2009-8-16 17:57:02 | 显示全部楼层
可怜的Youth 摸摸
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, 2021-1-24 07:06

Powered by Discuz! X3.4

© 2001-2017 Comsenz Inc.

快速回复 返回顶部 返回列表