FreeBSD, GPGPU and OpenCL/CUDA

Gautham Ganapathy gautham at lisphacker.org
Wed Sep 22 20:33:05 UTC 2010


On Fri, Aug 20, 2010 at 6:10 PM, b. f. <bf1783 at googlemail.com> wrote:
>
> Okay.  Maybe we'll see a convergence of the two in the medium/long
> term.  Right now, interested parties need to look at the available
> hardware, and then talk to the vendors about whether they would be
> willing to support a port of their software to FreeBSD, and
> _specifically_, what is needed.  For example, we faced a similar
> situation with the newer Nvidia GPUs not so long ago.  Some key
> developers like John Baldwin got involved, and determined what changes
> needed to be made in the FreeBSD base system in order to support the
> newer hardware and graphics drivers.  It would have been nice to get
> an open-source driver, but since Nvidia wasn't willing to do that,
> FreeBSD  chose to meet them half-way.  Probably a similar effort will
> be needed for CUDA.  Someone should look at the requirements, and have
> a _detailed_, _sustained_ discussion with Nvidia and the FreeBSD
> Foundation.  If, for example, KMS is needed, then the Foundation may
> be willing to invest in that, because it will probably also be needed
> for new graphics drivers and Xorg, anyway.  Robert Noland was working
> on it, but he was doing it largely by himself in his spare time, and
> then he got a new job and had to slow down considerably, if not stop
> altogether.
> ...

Hi

I remember seeing a post sometime back (either here or on nvnews)
about someone getting a prebuilt linux-based CUDA application to work
freebsd's linuxulator. Does this mean that the driver is ready, and
just the toolchain has to be ported?

Regards
Gautham Ganapathy


More information about the freebsd-questions mailing list