Using alternatives instead of a conflict for libopencl1?

Tomasz Rybak tomasz.rybak at post.pl
Thu Apr 25 18:49:44 UTC 2013


Dnia 2013-04-22, pon o godzinie 12:41 +0200, Simon Richter pisze:
> Hi,
> 
> I've just introduced "Beignet", which is Intel's Open Source OpenCL
> implementation for Ivy Bridge, into Debian. The library can be loaded
> either as a full OpenCL implementation or as an ICD driver, so I'd
> like to avoid two separate packages for both modes, which leaves the
> problem that I need to introduce a Conflicts: if I use the
> libOpenCL.so.1 name.

From my point of view ICD would be preferred.
I wanted to test beignet with PyOpenCL;  I've installed it
on amd64 AMD Phenom and added file to /etc/OpenCL/vendors
pointing it to libocl.so.0.
When I run clinfo to see what is available, I got
"Invalid instruction" signal and core dump.
Should I file the bug? Can beignet run on non-Intel CPU?
If not, as Debian does not allow to depend on particular hardware
(e.g. "Depends: Intel") I guess there would be need to put
some README.Debian or maybe debconf with warning about it.

Best regards.

-- 
Tomasz Rybak  GPG/PGP key ID: 2AD5 9860
Fingerprint A481 824E 7DD3 9C0E C40A  488E C654 FB33 2AD5 9860
http://member.acm.org/~tomaszrybak

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.alioth.debian.org/pipermail/pkg-nvidia-devel/attachments/20130425/069a667c/attachment.pgp>


More information about the pkg-nvidia-devel mailing list