[Bf-committers] Bundling CUDA Kernels on all plattforms
Brecht Van Lommel
brechtvanlommel at pandora.be
Mon Dec 12 23:45:38 CET 2011
The reason to not support 1.2 and lower is mostly so that we don't
have to drop support for them again in one of the next releases when
we add more features. I don't want to add support for an architecture
that we might not be able to support anymore in the next release.
Regarding bundling CUDA kernels, I'll ask Sergey to do it for Linux
builds. For Mac there's not much point I think, as far as I know
they're not selling computers with NVidia cards that have shader model
1.3 or newer, all the high end computers have ATI.
On Sun, Dec 11, 2011 at 9:39 PM, Wolfgang Fähnle <wfaehnle at freenet.de> wrote:
> Hi, i start an initiative on BA to build older Cuda kernels and upload it
> to graphicall.org.
> It would be nice if some builder compile 1.0, 1.1 and 1.2 kernels for all
> I have upload these kernels for linux 64 on my dropbox account.
> You will find the link in the thread.
> Cheers, mib.
> Am 11.12.2011, 20:36 Uhr, schrieb Yousef Hurfoush <bat3a at msn.com>:
>>> I have to agree with this, as compiling the kernel takes 15 or so
>>> and nearly crashes my computer. And after that it didn't even work (It
>>> it was out of memory, probably because my card is so old. It's a geforce
>> i do have the same card, even thought it is 5 years old it renders as
>> 1/2 my this year i5/2500 CPU.
>> so it worth bundling 1.0 1.1 1.2 kernels too, even people with old cards
>> wants to at least test cycles gpu.
>> Yousef Harfoush
>> bat3a at msn.com
>> Bf-committers mailing list
>> Bf-committers at blender.org
> Bf-committers mailing list
> Bf-committers at blender.org
More information about the Bf-committers