As a user I would answer in the affirmative. It should be controllable as are the uses of the GPUs (though that would be nice if some of those controls were GUI features instead of cc config only),
The primary advantage to me of OpenCL is that if the projects targeted that we users would have greater / wider availability of usable projects... only time will tell, but I suspect that optimized applications that use CUDA and Stream will likely still be faster in the end ... but, ... at the moment, the ATI users are still lagging in project availability ... but who is to say the situation will not reverse itself in a few years? On May 17, 2010, at 12:59 AM, Bernd Machenschalk wrote: > Charlie Fenton wrote, On 13.05.2010 2:12 Uhr: >> I checked the NVIDIA link and found the following information for the Mac: >> >>> OpenCL drivers are released as part of the Mac OS X Showleopard >>> operating system. >>> >> >> So apparently no additional drivers are needed, though it's unclear >> whether the NVIDIA CUDA drivers are needed for OpenCL. >> >> There is no mention of Macintosh support on the ATI link. >> > > I successfully ran an OpenCL App with an NVidia GPU on Mac OS 10.6 > without having to install an additional driver. > > It even works with ATI support there, though the kernel compiler is > still buggy. > > Which brings me to: > By design OpenCL enables to use multiple different computing devices > such as CPUs, GPUs (NVidia/ATI/Intel), Coprocessors (Cell) > _on_the_same_machine_, and Apples implementation actually does support > this at least for different GPUs. Does BOINC? Should it? _______________________________________________ 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.
