Originally posted by darkbasic
View Post
Announcement
Collapse
No announcement yet.
Beignet 1.3 Released With OpenCL 2.0 Support
Collapse
X
-
-
-
Originally posted by illwieckz View Post
It's not “Beginet” it’s “Beignet”, the “i” is before the “g”. See Wikipedia article: https://en.wikipedia.org/wiki/Beignet
Pronounce it “Benyey”, with “Be” like in “Betty”, “ny” like in “nyan cat” and “ey” like in “they”
It's a pastry, there is many kind of beignets: dougnuts are one kind of beignet for example.
 
And I have a great-great-grandfather who was French, and is turning in his grave...
Comment
-
Originally posted by Linuxhippy View PostI am usually a big fan of AMDs open-source efforts, but when it comes to OpenCL on Linux - there have been a lot of promises but no working code (same as with Vulkan).
There is typically a big pile of work that has to happen before code can be open-sourced... most of that work is now done. You can write something from scratch in public open source but you can't take something written to be proprietary and "replace or re-write the stuff you can't publish" in public, so the "open source" part always has to come last when you are dealing with cross-platform proprietary code.
If we had actually done a clean implementation as you suggest then we could have done more of the work in public, but that was not at all the case.
Originally posted by Linuxhippy View PostFrom what I understand, AMD does not plan to open-source their Catalyst OpenCL-2 implementation - but instead the LLVM based RocM approach - which is a clean implementation, and shows completly different characteristics (just have a look at the phoronix benchmarks).
The new Linux implementation is clean as in IP-clean (or at least pretty close now) but I think you are suggesting clean as in new implementation. That is not what we did - same code will be running on Linux & Windows other than the GPU back end (maybe 5-10% of the total stack).
Originally posted by Linuxhippy View PostI really hope AMD will provide a useable and fully-fledged open-source OpenCL-2 implementation soon as promised ... I've been waiting for that to happen for years.
We also said that work on amdgpu was just starting and that the full transition would take a few years (I think we said 2-3 years).
AFAIK the plan is what we are calling "1.2+", which is 1.2 plus a few 2.0 features that people actually seem to be using. Some of the OpenCL 2.0 features do not seem to have really caught on in OpenCL (just on CUDA) so we are addressing those with HCC/HIP.Last edited by bridgman; 21 January 2017, 02:08 PM.Test signature
- Likes 2
Comment
-
Seems like the GPGPU market is divided in two segments: Nvidia, the leader, which can’t be bothered with OpenCL, and all the others, who do support it, but cannot offer the same level of performance.
Is the market simply not big enough to be properly competitive?
Comment
-
Originally posted by Krejzi View Post
Out of those, SNB has no GPU OCL support, IVB and HSW can only support 1.2. Only reminder is BDW, which we can hope either they or someone from the community will tackle.
(^f resource constrain)
As per IVB and HSW, allow me to make a parallel - AMD support OCL 2 on 1st gen GCN onwards (that's a 2011 product line). In comparison, Intel have been actively dragging their GPU feet, because their CPUs are oh-so-great at OCL.Last edited by darkblu; 22 January 2017, 06:23 AM.
Comment
-
Originally posted by bridgman View PostDepends on whether the features get used, I guess. We supported 2.0 for years but most of the applications stayed with 1.2 or lower.
Comment
Comment