The Portland Group to release CUDA F95 and CC compilers

[Idontcare]

The PGI 9.0 release includes the PGI Accelerator? Fortran and C99 compilers supporting x64+NVIDIA Linux systems; PGF95 and PGCC accelerator compilers are supported on all Intel and AMD x64 processor-based systems with CUDA-enabled NVIDIA GPUs.

http://www.pgroup.com/resources/accel.htm

I love PGI :thumbsup: and I think this is awesome news.

 

[Scali]

"PGI Accelerator compilers target all CUDA-enabled NVIDIA GPU accelerators with compute capability 1.2 or higher."

Guess my dusty old 8800GTS320 is out

Do you still use Fortran? Or are you more interested in the C++ portion?

 

[TurtleCrusher]

Linux only?

Boo.

 

[Idontcare]

Originally posted by: Scali
Do you still use Fortran? Or are you more interested in the C++ portion?

Both. For original works I am only interested in authoring via C++ at this time, but I've got fortran-based source which requires recompile any time I feel like rewriting for improvements and I happen to use PGI for that already.

Originally posted by: Scholzpdx
Linux only?

Boo.

If you are the type who creates/compiles source then you are likely to already have a linux box or vm linux setup anyways. I don't know of anyone (speaking personally, from friends and colleagues) who consider this to be a barrier of any kind.

The "wider" sphere of impact here is that this (mainstream integration of CUDA support into a tier-1 compiler) is one of the necessary steps we (end-consumers) need to see occur before we can expect broader mainstream availability of CUDA enabled applications.

 

[Fox5]

This is really nice.
But I think CUDA 1.2 is only on the GTX 260 and higher.

 

[Scali]

Originally posted by: Fox5
This is really nice.
But I think CUDA 1.2 is only on the GTX 260 and higher.

Well, compute capability 1.2 to be exact. The versioning of Cuda itself is independent of the compute capability. Cuda is now at 2.2, but still supports all versions of the hardware.
Compute capability 1.2 isn't actually used in any hardware as far as I know. But you're right, GT200 is 1.3, G92 is 1.1 and G80 is 1.0. So only GTX260 and higher are supported.

 

[Fox5]

Originally posted by: Scali

Originally posted by: Fox5
This is really nice.
But I think CUDA 1.2 is only on the GTX 260 and higher.

Well, compute capability 1.2 to be exact. The versioning of Cuda itself is independent of the compute capability. Cuda is now at 2.2, but still supports all versions of the hardware.
Compute capability 1.2 isn't actually used in any hardware as far as I know. But you're right, GT200 is 1.3, G92 is 1.1 and G80 is 1.0. So only GTX260 and higher are supported.

Well, 1.3 does provide some very important improvements over 1.1, and seems to be the start of what will come to be "true" CUDA support. In the long-run of CUDA, I think support for geforce 8 and 9 will be dropped. Even if it isn't, a GTX 260 can easily have 10x the performance of a 9800GTX+ in CUDA code.

I'm excited for nvidia's next gen though, and what it will bring in terms of CUDA support.