Hi Andrew,

Thanks for pointing this out.  I fixed it, though it actually had no effect 
because COPROC_INTEL::correlate() is an empty routine, as there is nothing to 
correlate the Intel OpenCL instances with.  

To explain this a bit better: NVIDIA GPUs can perform computation using either 
the OpenCL or CUDA framework, so NVIDIA OpenCL instances must be correlated 
(matched) with their corresponding CUDA instances, to prevent scheduling the 
same GPU for two tasks at the same time.  Likewise, AMD GPUs can perform 
computation using either the OpenCL or CAL framework, so AMD OpenCL instances 
must be correlated with their corresponding CAL instances.  But Intel GPUs can 
perform computation only using the openCL framework.

Cheers,
--Charlie

--
Charlie Fenton                        [email protected]
BOINC / SETI@home Macintosh & Windows Programmer
Space Sciences Laboratory
UC Berkeley



On Dec 14, 2013, at 2:32 AM, Andrew Dicker wrote:

> Hi all,
> Line 168 of gpu_detect seems odd to me, so i thought i'd raise a query
> about it. sorry if this isn't the correct place, i couldn't find anywhere
> on the forums discussing code.
> 
> 
> 
> 166
>    nvidia.correlate(use_all, ignore_gpu_instance[PROC_TYPE_NVIDIA_GPU]);
> 167     ati.correlate(use_all, ignore_gpu_instance[PROC_TYPE_AMD_GPU]);
> 168
>    intel_gpu.correlate(use_all, ignore_gpu_instance[PROC_TYPE_AMD_GPU]);
> 169     correlate_opencl(use_all, ignore_gpu_instance);
> 
> Shouldn't it be PROC_TYPE_INTEL_GPU ?
> 
> Regards,
> Andrew Dicker
> _______________________________________________
> boinc_dev mailing list
> [email protected]
> http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
> To unsubscribe, visit the above URL and
> (near bottom of page) enter your email address.
> 

_______________________________________________
boinc_dev mailing list
[email protected]
http://lists.ssl.berkeley.edu/mailman/listinfo/boinc_dev
To unsubscribe, visit the above URL and
(near bottom of page) enter your email address.

Reply via email to