Sandesh <cofsandy@...> writes:
>
> >
> > Sorry, this was broken. It's fixed in git. Thanks for the report.
> >
> > Andreas
> >
>
> Thankyou for the clarification. I shall checkout the latest version from git
> and see its OK.
>
> Sandesh
>
>
So for the latest checkout of git the solve_pkt_with_cg(..) works fine only
for certain use cases. For a 10 x 10 tridiagonal matrix it worked perfectly
,but for 100 x 100 tridiagonal matrix it gave the following error.
The same error for 1000 x 1000 .triadiagonal matrix. In fact uptill 49 x 49
it works perfectly, the problem starts when the size is 50 x 50 or above.
starting...
building...
start solve
Traceback (most recent call last):
File "main_cg.py", line 95, in <module>
main_cg()
File "main_cg.py", line 61, in main_cg
pagelocked_allocator=pagelocked_pool.allocate)
File
"~/dist-packages/pycuda-2013.1.1-py2.7-linux-x86_64.egg/pycuda/sparse/cg.py",
line 230, in solve_pkt_with_cg
debug_callback=debug_callback)
File
"~/dist-packages/pycuda-2013.1.1-py2.7-linux-x86_64.egg/pycuda/sparse/cg.py",
line 199, in run
raise ConvergenceError("cg failed to converge")
pycuda.sparse.cg.ConvergenceError: cg failed to converge
PyCUDA WARNING: a clean-up operation failed (dead context maybe?)
cuMemFreeHost failed: invalid context
PyCUDA WARNING: a clean-up operation failed (dead context maybe?)
cuMemFreeHost failed: invalid context
PyCUDA WARNING: a clean-up operation failed (dead context maybe?)
cuMemFreeHost failed: invalid context
Sandesh
_______________________________________________
PyCUDA mailing list
[email protected]
http://lists.tiker.net/listinfo/pycuda
