It's Time To Start Thinking About GCC 4.8
Phoronix: It's Time To Start Thinking About GCC 4.8
While GCC 4.8 will likely not see the light of day in 2012, GCC 4.7 was branched today with an imminent release candidate and now it's "trunk" code-base is open for GCC 4.8 development efforts...
I would like to see initial support for OpenACC. I know that it is complicated especially because of the driver situation, but an initial support of the pragmas just using the CPU what be a sufficient first step.
I'd much rather see AMP. OpenACC is incredibly complicated compared to AMP, is heavily tied to NVIDIA's hardware unlike the hardware-agnostic AMP, and MS's engineers are putting a lot of effort into helping any willing vendor to implement AMP on their compiler and OS of choice (yes, even GCC/Linux, if anyone approaches them asking for dev support).
Originally Posted by atcl
I would rather point to PGI and their Accelerators than to nvidia in the case of OpenACC, but how is that better or worse than AMP being tied to MS and probably their DirectX/DirectCompute architecture. On the measure on complexity I wouldn't dare to judge as I am not in the spot of implementing it. I would prefer OpenACC as it doesn't introduce language additions but mere pragmas which would enable me to use GPGPU just like OpenMP. From my point of view OpenACC won't care about the hardware backend, as that is kind of the point of these frameworks accelerating with whatever is available, be it CPU or GPU. I am not seeing MS actually supporting another compiler but their own beyond actually opening up the "standard" describing AMP.
P.S.: And yes I also saw the second row head to head between AMD(AMP) and nvidia(OpenACC).
Last edited by atcl; 03-03-2012 at 03:08 AM.
Tags for this Thread