Control: tags -1 fixed-upstream patch
This was probably triggered by python-attrs 24 removing the temporary
tuples from __eq__ (for efficiency):
https://sources.debian.org/src/python-attrs/24.2.0-1/CHANGELOG.md/#L46
This means that comparing an object to itself is no longer automatically
True
Control: reassign -1 xserver-xorg-video-intel
Control: retitle -1 xserver-xorg-video-intel: [Skylake] breaks OpenCL
On 12/04/17 14:14, Jason Riedy wrote:
And Rebecca N. Palmer writes:
Try removing xserver-xorg-video-intel: it isn't needed on anything
recent enough to use beignet, and has
Control: retitle -1 mesa-opencl-icd/libclc: ! does not accept pointers
There appear to be two separate bugs here:
-get_global_id() does not work (the clinfo problem). This has also been
reported as #857591 and upstream, where it has a patch:
https://bugs.freedesktop.org/attachment.cgi?id=1301
* nvidia-kernel-dkms: Switch to Recommends: nvidia-driver | libcuda1
to break the chain libcuda1 -> nvidia-kernel-dkms -> nvidia-driver.
#768185
suggests that nvidia-opencl-icd works without the graphics side (can
someone check that?), making this the more correct place to cut the chain.
Sorr
(Should we merge these bugs? Also, #767803 looks like another instance
of this, though it doesn't have the apt log to confirm it yet)
* nvidia-kernel-dkms: Switch to Recommends: nvidia-driver | libcuda1
to break the chain libcuda1 -> nvidia-kernel-dkms -> nvidia-driver.
...or drop this Recom
a better fix might be for whatever sets nvidia
as default graphics provider to only do so if the hardware is present,
but I don't know whether that's practical.
The package already has a check in
http://sources.debian.net/src/nvidia-graphics-drivers/340.46-5/debian/libgl1-nvidia-glx.preinst.in
Rebecca Palmerr wrote:
The only other [than pyopencl]
Depends or Recommends on opencl-icd in the current archive is bfgminer.
Sorry...only ones found by "path:debian/control opencl-icd" in
sources.debian.net search (apt-cache rdepends doesn't work on virtual
packages), which evidently doesn't s
I think the trigger is nvidia-opencl-icd adding a new dependency on
libcuda1 (changelog: Add libcuda1 dependency to libraries that seem to
be capable of doing dlopen("libcuda.so") or dlopen("libcuda.so.1").),
which pulls in the rest of nvidia-* as libcuda1 Recommends:
nvidia-kernel-dkms which R
Control: fixed -1 beignet/0.9.3~really.0.8+dfsg-1
Control: found -1 beignet/1.0.0-1
Control: tags -1 patch
Adding RTLD_DEEPBIND in ocl-icd doesn't help, but building beignet with
-Bsymbolic (attached, replaces the existing patch with the same name)
does: beignet and pocl then continue to work w
Control: retitle -1 mesa-opencl-icd,beignet: installing together breaks all ICDs
Control: reassign -1 mesa-opencl-icd,beignet
Control: found -1 mesa/10.3.2-1
Control: found -1 beignet/0.9.3~dfsg-1
It appears I forgot to test mesa+pocl without beignet installed: pocl
then works, i.e. the problem o
If the new package is the ICD rather than the loader, why is it called
libopencl1-mesa rather than mesa-opencl-icd? Consistency with the
others would suggest the latter.
Due to unsupported hardware, I can't tell you whether it actually works;
you might want to join the current discussion at
11 matches
Mail list logo
