AMD Prepares "Bobcat 2" Compiler Support

[phoronix]

Phoronix: AMD Prepares "Bobcat 2" Compiler Support

While the AMD Linux graphics team is busy proposing an LLVM back-end staging area, the AMD CPU folks have begun work on the GCC compiler and preparing support for the next-generation Bobcat processors...

http://www.phoronix.com/vr.php?view=MTE0Mzc

[dimko]

upgrade

Looks like AMD will force me to make an upgrade.
After all, I am gentoo dude, so GPU acceleration during compilation, or god forbid them, GPU acceleration of runtime is gonna create MAJOR BOOST to AMD platform, that they need so much in their so far unfair battle with Intel.

[Fenrin]

Looks like AMD will force me to make an upgrade.
[...]

the Bobcat 2 will be a low power APU. But I'm curious about this next-generation bobcat as well. Although I will probably prefer a Trinity APU to it.

Maybe Bobcat2 will already contain a ARM chip, with it's trusted computing technology Trinity does not contain this feature, unfortunately also not HSA. I'm not sure what's better a APU with ARM chip and trusted computing plus HSA support or Trinity APU without these technologies.

[johnc]

Is there really GPU accelerated compilation out there? I've been thinking about that for awhile.

[smitty3268]

Is there really GPU accelerated compilation out there? I've been thinking about that for awhile.

I doubt it - i think compilation would be a fairly serialized process which would be difficult to accelerate on the GPU.

But maybe i'm wrong.

[XorEaxEax]

I doubt it - i think compilation would be a fairly serialized process which would be difficult to accelerate on the GPU.

But maybe i'm wrong.

Well given how you can already make excellent use of cpu cores (-jx) to decrease compilation time then I'm not sure the serialization process is a big hindrance, and while GPU in general probably isn't very suited for the type of code used in compilation, CPU/GPU heterogenous frameworks like opencl will likely have a big impact in utilizing GPU's for general purpose computing.

It seems there are some really exciting times ahead in this area as more and more projects start to utilize the possibilities this offers, which will likely be very much fueled by the increased usage of IGPU's which in turn will rely on multiple gpu cores and the maximized utilization of those together with multiple cpu cores to effectively compete with discrete graphics, an area for which solutions like opencl are tailor-made.

[curaga]

Hey, they did share it has double the L2 (1 -> 2mb) and supports AVX, BMI. Decent info.