Hi,

Le 22/01/2017 à 06:46, Pan, Xiuli a écrit :

> Hi all,
>
> Thank you for  pay close attention to our latest release. We are very sorry 
> that we are not very thoughtful about the package making things.
>
> We did not have a performance test for OCL2.0 so we did not set it as a 
> default option and we did not think about how to run a binary version on both 
> old 1.2 devices and new 2.0 devices.
> Thanks to Rebecca's patch to change the 1.2 and 2.0 backend on runtime, I 
> have test it on skl and ivb and it works fine. There are few problems like 
> utest for 2.0 and the #include "../src/cl_device_data.h" in backend.
> I think we can add runtime check for 2.0 in utest and refine the function to 
> pass the OCL20 support version from API runtime to backend.
>
> And the default version problem can also be fixed by deleting the default 
> setting to 200 at all, it seems all default OpenCL C version is 1.2 now.
>
> Sorry for this inconvenience of release 1.3. We were planning to have 1.3.1 
> release with some performance patches, but now we may need have fix release 
> 1.3.1, or you may think Rebecca's patch can solve your problem now?
> As we are having the Chinese New Year holiday soon, we may need to decide it 
> very quickly. Looking forward your replies!

Thanks for paying attention to our reports!

I, for one, would prefer an official 1.3.1 “bugfix” release, eventually
just mostly based on Rebecca’s V2 patch (+the others she sent there),
because our policy (ArchLinux one, AFAIK/U it) is to wait for patches to
be at least accepted upstream before incorporating them; and if you’ve
gone all the way to that, making a 1.3.1 release would probably not be a
lot more efforts while it always feels cleaner to build cleanly from a
release than needing to add patches. ;)

Cheers,
Bruno

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Beignet mailing list
Beignet@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/beignet

Reply via email to