For Now Intel's Beignet Seems Better Off Than Radeon Gallium3D Clover

[bridgman]

And that's a big omission from AMD's side.

I'm guessing you weren't around in the "give us some documentation and the community will write everything for you" days...

[System25]

I'm guessing you weren't around in the "give us some documentation and the community will write everything for you" days...

He-he, documentation is good, but GPUs seems to be specific and complex things. On side note, quality of open driver improved a lot when AMD did it. Looking on nouveau I can admit things can be much worse. So I would consider documentation release as good thing which has actually worked, more or less.

But, still, don't you think it's really WEIRD when AMD invents OpenCL standard, it getting some traction across devs and then ... Intel beats AMD on their own field due to better drivers?! Holy crap.

[liam]

AMD invents OpenCL standard, it getting some traction across devs and then ... Intel beats AMD on their own field due to better drivers?! Holy crap.

AMD didn't invent it. It was a group of companies who worked on a proposal submitted by apple

OpenCL was initially developed by Apple Inc., which holds trademark rights, and refined into an initial proposal in collaboration with technical teams at AMD, IBM, Qualcomm, Intel, and Nvidia.

Also, intel has a far larger team of developers for their driver than does amd for radeon.

[smitty3268]

AMD didn't invent it. It was a group of companies who worked on a proposal submitted by apple

Indeed, AMD had their own CUDA alternative back in the day, I think it came out of Stanford and was just called Stream?

It never really went anywhere, though, and AMD dropped it as soon as OpenCL was announced.

[darkbasic]

I realize you don't need it now, but the Catalyst bug tracker the devs look at is here :

http://ati.cchtml.com/

I do need it, thank you very much!

[bridgman]

The main focus for the "OpenCL + shader compiler team" (aka Tom) for the last several months has been on improving the shader compiler for GCN, which benefits both graphics and compute.

[darkbasic]

"OpenCL + shader compiler team" (aka Tom)

LOL (stupid character limit)

[Veerappan]

The main focus for the "OpenCL + shader compiler team" (aka Tom) for the last several months has been on improving the shader compiler for GCN, which benefits both graphics and compute.

Speaking of the CL+Shader Compiler team... Do you happen to know Matt Arsenault's official role? It seems that he is more focused on the GCN back-end and/or HSA than OpenCL or graphics, but I never really saw it stated plainly.

[Veerappan]

Also, it seems applicable to this thread:

http://www.pcc.me.uk/pipermail/libclc-dev/2014-September/000692.html

Intel's China OTC team is proposing a possible refactoring of beignet to use libclc (and improve libclc where necessary), and one of the POCL devs also expressed interest in doing the same... Maybe there's hope that the various OSS OpenCL implementations will stop re-inventing the wheel.

[blackiwid]

I realize you don't need it now, but the Catalyst bug tracker the devs look at is here :

http://ati.cchtml.com/

why is it then unofficial?

Just hurry to make this same kernel parts for radeon and the blob, to get rid of 50% of the code of the blob, and then work fast to replace the other 50% I dont get this double strategie, every year it gets less understandable, in the beginning radeon could not do much and was slow but every 5% it gets faster and has more feature, it gets more insane to keep that garbage.

Just "get it out of here", seriously who buys amd hardware to use under linux this blob? If I want to use a blob I know nvidias driver under linux is better. It makes not even sense economical.

I get that u want to keep the old drivers for a few cracy people that use this blob, but just move the 100 developers to make radeon better and keep 10 to maintain the blob not the other way around.