Do we really want this? Perhaps CUDA should be just disabled, instead
of configure failing? Note that I've not actually asked for a
CUDA-enabled build:
$ cat linux-gnu/conf/reconfigure-linux-gnu.py
#!/usr/bin/python
if __name__ == '__main__':
import sys
import os
sys.path.insert(0,
I think part of the issue is the configure model for compiler detection.
Perhaps we want buildsystem to detect/set compilers as it sees. And
then petsc side of configure to check if this functionality exists -
and then use it? [with --with-cuda=1 flag in cuda.py?]
There was a recent change to
On 5 November 2010 15:18, Satish Balay balay at mcs.anl.gov wrote:
If so - perhaps the fix is to remove this check from Buildsystem
[setCompilers.py - where it only determines cuda attributes like
version - as stashes this info] - and place the version check back in
cuda.py?
Satish
That
On Nov 5, 2010, at 1:18 PM, Satish Balay wrote:
I think part of the issue is the configure model for compiler detection.
Perhaps we want buildsystem to detect/set compilers as it sees. And
then petsc side of configure to check if this functionality exists -
and then use it? [with
I did this already, have you pulled?
Matt
On Fri, Nov 5, 2010 at 1:29 PM, Barry Smith bsmith at mcs.anl.gov wrote:
On Nov 5, 2010, at 1:18 PM, Satish Balay wrote:
I think part of the issue is the configure model for compiler detection.
Perhaps we want buildsystem to detect/set
There are a few changes after your 'feature addition. I'm not sure
where things diverged. I'm checking the rev history to see if I can
piece together why the 'divergence happeed'.
Satish
On Fri, 5 Nov 2010, Matthew Knepley wrote:
I did this already, have you pulled?
Matt
On Fri, Nov