Macbook Pro CUDA 跑CC总出错

panzerkiller

2010年的MBP，nvidia Geforce 320M，48个core，不算太差；跑PG的cuda包没问题，跑CC的opencl也没问题，CC的cuda包一算就出错。看了下log，貌似红色的就是出错的包，我也只能看出这点东西，哪位大神给看一下什么地方不对劲

Fri Mar  8 22:06:21 2013 | Collatz Conjecture | Requesting new tasks for NVIDIA
Fri Mar  8 22:06:22 2013 | Collatz Conjecture | Scheduler request completed: got 14 new tasks
Fri Mar  8 22:06:24 2013 | Collatz Conjecture | Started download of mini_collatz_3.10_x86_64-apple-darwin__opencl_osx
Fri Mar  8 22:06:24 2013 | Collatz Conjecture | Started download of collatz_2378766611386888595816_103079215104
Fri Mar  8 22:06:25 2013 | Collatz Conjecture | Finished download of collatz_2378766611386888595816_103079215104
Fri Mar  8 22:06:25 2013 | Collatz Conjecture | Started download of collatz_2378766517894040496488_824633720832
Fri Mar  8 22:06:26 2013 | Collatz Conjecture | Finished download of collatz_2378766517894040496488_824633720832
Fri Mar  8 22:06:26 2013 | Collatz Conjecture | Started download of collatz_2378766611077650950504_103079215104
Fri Mar  8 22:06:26 2013 | Collatz Conjecture | Starting task collatz_2378766517894040496488_824633720832_0 using collatz version 202 (cuda) in slot 2
Fri Mar  8 22:06:27 2013 | Collatz Conjecture | Finished download of collatz_2378766611077650950504_103079215104
Fri Mar  8 22:06:27 2013 | Collatz Conjecture | Started download of collatz_2378766505318376253800_824633720832
Fri Mar  8 22:06:27 2013 | Collatz Conjecture | Computation for task collatz_2378766517894040496488_824633720832_0 finished
Fri Mar  8 22:06:27 2013 | Collatz Conjecture | Output file collatz_2378766517894040496488_824633720832_0_0 for task collatz_2378766517894040496488_824633720832_0 absent
Fri Mar  8 22:06:29 2013 | Collatz Conjecture | Finished download of collatz_2378766505318376253800_824633720832
Fri Mar  8 22:06:29 2013 | Collatz Conjecture | Started download of collatz_2378766518924832647528_824633720832
Fri Mar  8 22:06:29 2013 | Collatz Conjecture | Starting task collatz_2378766505318376253800_824633720832_0 using collatz version 202 (cuda) in slot 2
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Finished download of mini_collatz_3.10_x86_64-apple-darwin__opencl_osx
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Finished download of collatz_2378766518924832647528_824633720832
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Started download of collatz_2378766611180730165608_103079215104
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Started download of collatz_2378766496659722185064_824633720832
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Computation for task collatz_2378766505318376253800_824633720832_0 finished
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Output file collatz_2378766505318376253800_824633720832_0_0 for task collatz_2378766505318376253800_824633720832_0 absent
Fri Mar  8 22:06:30 2013 | Collatz Conjecture | Starting task collatz_2378766611386888595816_103079215104_0 using mini_collatz version 310 (opencl_osx) in slot 2
Fri Mar  8 22:06:31 2013 | Collatz Conjecture | Finished download of collatz_2378766611180730165608_103079215104
Fri Mar  8 22:06:31 2013 | Collatz Conjecture | Finished download of collatz_2378766496659722185064_824633720832
Fri Mar  8 22:06:31 2013 | Collatz Conjecture | Started download of collatz_2378766485939483814248_824633720832
Fri Mar  8 22:06:31 2013 | Collatz Conjecture | Started download of collatz_2378766574072212728168_103079215104
Fri Mar  8 22:06:32 2013 | Collatz Conjecture | Finished download of collatz_2378766485939483814248_824633720832
Fri Mar  8 22:06:32 2013 | Collatz Conjecture | Finished download of collatz_2378766574072212728168_103079215104
Fri Mar  8 22:06:32 2013 | Collatz Conjecture | Started download of collatz_2378766479033176402280_824633720832
Fri Mar  8 22:06:32 2013 | Collatz Conjecture | Started download of collatz_2378766611283809380712_103079215104
Fri Mar  8 22:06:33 2013 | Collatz Conjecture | Finished download of collatz_2378766479033176402280_824633720832
Fri Mar  8 22:06:33 2013 | Collatz Conjecture | Finished download of collatz_2378766611283809380712_103079215104
Fri Mar  8 22:06:33 2013 | Collatz Conjecture | Started download of collatz_2378766486764117535080_824633720832
Fri Mar  8 22:06:33 2013 | Collatz Conjecture | Started download of collatz_2378766479857810123112_824633720832
Fri Mar  8 22:06:34 2013 | Collatz Conjecture | Finished download of collatz_2378766486764117535080_824633720832
Fri Mar  8 22:06:34 2013 | Collatz Conjecture | Finished download of collatz_2378766479857810123112_824633720832
Fri Mar  8 22:06:34 2013 | Collatz Conjecture | Started download of collatz_2378766487588751255912_824633720832
Fri Mar  8 22:06:35 2013 | Collatz Conjecture | Finished download of collatz_2378766487588751255912_824633720832

tefantasy

红色的是程序异常退出，连输出都没有。
其他跑得没问题，就不应该是驱动问题。

是不是程序本身缺陷或是系统设置问题？

cuihao

hmm，windows吗？
昨天我用NVS 5400M跑CC，也是接一个错一个，后来换了官网上的测试程序就好了。

acp134

用官方的优化试试？

panda7456

acp134 发表于 2013-3-10 09:58
用官方的优化试试？

apu的CC也error
我在检查驱动和opencl apps

1. Stderr output
   
2. <core_client_version>7.0.28</core_client_version>
   
3. <![CDATA[
   
4. <message>
   
5. - exit code -1073741819 (0xc0000005)
   
6. </message>
   
7. <stderr_txt>
   
8. Collatz Conjecture v3.11 x86_64 for OpenCL
   
9. Based on the AMD Brook+ kernels by Gipsel
   
10. Parameters          I1048576 K64 S1 --device 0
    
11. Start               2378778907397220706664
    
12. Checking            824633720832 numbers
    
13. Device Vendor       Advanced Micro Devices, Inc.
    
14. Name                Scrapper
    
15. Board Name          AMD Radeon HD 7480D
    
16. Compute Units       2
    
17. Driver Version      1084.4 (VM)
    
18. OpenCL Version      OpenCL 1.2 AMD-APP (1084.4)
    
19. Workgroup Size      Auto
    
20. Kernel              amd_bitalign
    
21. 
    
22. 
    
23. Unhandled Exception Detected...
    
24. 
    
25. - Unhandled Exception Record -
    
26. Reason: Access Violation (0xc0000005) at address 0x0000000076DFE4B4 write attempt to address 0x00000024
    
27. 
    
28. Engaging BOINC Windows Runtime Debugger...
    

复制代码

panzerkiller

acp134 发表于 2013-3-9 19:58
用官方的优化试试？

CC的优化？请问哪里能找到？

panzerkiller

cuihao 发表于 2013-3-9 18:10
hmm，windows吗？
昨天我用NVS 5400M跑CC，也是接一个错一个，后来换了官网上的测试程序就好了。 ...

MAX OS 下跑的， OS 10.6.8， 很老了

acp134

panzerkiller 发表于 2013-3-10 21:59
CC的优化？请问哪里能找到？

很显眼的一个位置，貌似在your account的下面，貌似叫optimize program
唔，正好我这条回复下面的AD 就是

panzerkiller

找到了，谢谢