WCG开始准备支持GPU计算!

Blade.Li

LS先捞着了

nekoko

回复 16# Blade.Li


    http://www.worldcommunitygrid.or ... le.do?articleId=195

HCC GPU beta已经放出

xuyongchen

刚发现我似乎也拿到beta包了。不过似乎是cpu项目的

nekoko

回复 18# xuyongchen


好像是……不是所有的是GPU项目~
且，好像用的是opencl而不是cuda，看来作为显卡项目ppd不会太乐观~

xuyongchen

回复 19# nekoko


    wcg就留给cpu吧，显卡子项目就不要提高ppd了～
你的gpu wu算了多久？gpu时间

nekoko

回复 20# xuyongchen

1. <core_client_version>6.12.34</core_client_version>
   
2. <![CDATA[
   
3. <stderr_txt>
   
4. Commandline: projects/www.worldcommunitygrid.org/wcg_beta3_img_6.44_windows_intelx86__cuda_opencl X0000131280911201112141246.jp2 --device 0
   
5. Boinc requested NVIDIA gpu device number 0
   
6. Found compute platform NVIDIA Corporation
   
7. Selecting this platform
   
8. CL_DEVICE_NAME: GeForce GTX 570
   
9. CL_DEVICE_VENDOR: NVIDIA Corporation
   
10. CL_DEVICE_VERSION: 285.62
    
11. CL_DEVICE_MAX_COMPUTE_UNITS: 
    
12. CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS: 3
    
13. CL_DEVICE_MAX_WORK_ITEM_SIZES: 1024 / 1024 / 64
    
14. CL_DEVICE_MAX_WORK_GROUP_SIZE: 1024
    
15. CL_DEVICE_MAX_CLOCK_FREQUENCY: 1600 MHz
    
16. CL_DEVICE_ADDRESS_BITS: 32
    
17. CL_DEVICE_MAX_MEM_ALLOC_SIZE: 320 MByte
    
18. CL_DEVICE_GLOBAL_MEM_SIZE: 1280 MByte
    
19. CL_DEVICE_ERROR_CORRECTION_SUPPORT: no
    
20. CL_DEVICE_LOCAL_MEM_TYPE: local
    
21. CL_DEVICE_LOCAL_MEM_SIZE: 48 KByte
    
22. CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 64 KByte
    
23. CL_DEVICE_QUEUE_PROPERTIES: CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE
    
24. CL_DEVICE_QUEUE_PROPERTIES: CL_QUEUE_PROFILING_ENABLE
    
25. CL_DEVICE_EXTENSIONS:
    
26. cl_khr_byte_addressable_store
    
27. cl_khr_icd
    
28. cl_khr_gl_sharing
    
29. cl_nv_d3d9_sharing
    
30. cl_nv_d3d10_sharing
    
31. cl_khr_d3d10_sharing
    
32. cl_nv_d3d11_sharing
    
33. cl_nv_compiler_options
    
34. cl_nv_device_attribute_query
    
35. cl_nv_pragma_unroll
    
36. cl_khr_global_int32_base_atomics
    
37. cl_khr_global_int32_extended_atomics
    
38. cl_khr_local_int32_base_atomics
    
39. cl_khr_local_int32_extended_atomics
    
40. cl_khr_fp64
    
41. CL_DEVICE_COMPUTE_CAPABILITY_NV: 2.0
    
42. CL_DEVICE_REGISTERS_PER_BLOCK_NV: 32768
    
43. CL_DEVICE_WARP_SIZE_NV: 32
    
44. CL_DEVICE_GPU_OVERLAP_NV: CL_TRUE
    
45. CL_DEVICE_KERNEL_EXEC_TIMEOUT_NV: CL_TRUE
    
46. CL_DEVICE_INTEGRATED_MEMORY_NV: CL_FALSE
    
47. Estimated kernel execution time = 0.34477 [sec]
    
48. Starting analysis of ../../projects/www.worldcommunitygrid.org/BETA_X0000131280911201112141246_X0000131280911201112141246.jp2...
    
49. Extracting GLCM features...
    
50. Total kernel time: 113.800560 (1026 kernel executions)
    
51. Total memory transfer time: 7.951764
    
52. Average kernel time: 0.110917
    
53. Min kernel time: 0.096074 (dx=25 dy=3 sample_dist=24 )
    
54. Max kernel time: 0.132587 dx=2 dy=1 sample_dist=1
    
55. Total time for ../../projects/www.worldcommunitygrid.org/BETA_X0000131280911201112141246_X0000131280911201112141246.jp2: 267 seconds
    
56. Finished Image #0, pctComplete = 1.000000
    
57. CPU time used = 248.478393
    
58. 01:04:32 (11948): called boinc_finish
    
59. 
    
60. </stderr_txt>
    
61. ]]>
    

复制代码

xuyongchen

回复 21# nekoko


    用了一小时？

panda7456

我也收到beta了，不过不是gpu的……

nekoko

回复 22# xuyongchen


248秒……
乃看到的那个是系统时间……

nobodyk

坐等测试结果，看看能不能抢到DRT的饭碗

hngj2002

gpu项目支持a卡吗？

muclemanxb

opencl里面的啥要支持，记不清了，没有GPU的表示亚历山大

nekoko

回复 26# hngj2002


用opencl写的，或许会支持的，不过偶没有A卡……

xuyongchen

回复 24# nekoko


    那还是挺快的啊

H-Dan

那个 A卡和N卡在游戏方面 计算方面 性价比方面有啥区别？