Bug#841863: transition: nvidia-cuda-toolkit
On 24/10/16 00:04, Emilio Pozuelo Monfort wrote:
> On 23/10/16 23:54, Andreas Beckmann wrote:
>> Rdepends as I remember them (coccia is currently missing dak due to the
>> ongoing ftp-master move):
>>
>> eztrace-contrib
>> hwloc-contrib
>> starpu-contrib
>> pycuda
>
> Do they build fine with CUDA 8.0?
The nvidia-cuda-toolkit 8.0 transition was done in Ubuntu recently.
eztrace-contrib 1.1-5-1 - no changes needed
hwloc-contrib 1.11.3-2 - no changes needed
starpu-contrib 1.1.4+dfsg-6 - not in testing due to #837911,
1.2.0+dfsg-1 is in NEW with a fix. Ubuntu's 1.1.5-0 needed no
changes.
pycuda 2016.1.2+git20160809-1 - included patch from upstream (attached)
Regards
Graham
Description: Fix build with CUDA 8
Fix test suite to account for unavailability of
EXCLUSIVE compute mode in CUDA 8
Origin: upstream, https://git.tiker.net/pycuda.git/commit/255644ad802a20191e31bc15f4fd46e6c9d6e38a
Author: Andreas Kloeckner <inform@tiker.net>
Last-Update: 2016-10-10
--- a/test/test_driver.py
+++ b/test/test_driver.py
@@ -624,7 +624,7 @@
for e in range(e0-6, e0-4):
for i in range(100):
- queue.append(pool.allocate(1<<e))
+ queue.append(pool.allocate(1 << e))
if len(queue) > 10:
queue.pop(0)
del queue
@@ -632,9 +632,9 @@
@mark_cuda_test
def test_multi_context(self):
- if drv.get_version() < (2,0,0):
+ if drv.get_version() < (2, 0, 0):
return
- if drv.get_version() >= (2,2,0):
+ if drv.get_version() >= (2, 2, 0) and drv.get_version() < (8,):
if drv.Context.get_device().compute_mode == drv.compute_mode.EXCLUSIVE:
return
Reply to: