[Beignet] double precision support

[Zhigang Gong]

On Thu, Mar 05, 2015 at 03:51:21PM +0100, Malcolm Roberts wrote:
> Hello.
> 
> I just installed the latest version of beignet from git
> (74390cabe5d2958fec5806a5099bad44c74798f5) and I notice that
> cl_khr_fp64 is not available.  I am a bit late for the conversation
> in July
> (http://lists.freedesktop.org/archives/beignet/2014-July/003599.html)
> but I thought that I would add my voice.
> 
> I have a project to do scientific computation using the OpenCL
> platform, and double precision is vital for my work.  I understand
> that the beignet teams has limited resources and that one must
> prioritize, but I feel that double-precision would be a great
> addition.

So far, the double precision support is still not a high priority
task for Beignet. The two major reasons we already discussed in
the above thread remain unchanged now. 
1. Don't have many requirement from the community.
2. The double support is not fully supported. For example, all the math
   functions and even the divide instruction is not supported.

That's why we still haven't started to implement double support.

Thanks,
Zhigang.

> 
> Best,
> 
> ~Malcolm Roberts
> http://malcolmiwroberts.com/
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet