Hi Mahmood,
Defaults for the couette_flow_2d demo are (1) small and (2) double
precision.
If your GPU has good support for double precision (fp64), and lots of
memory, you can use gmsh to generate a bigger mesh for that demo.
E.g., the script below makes a mesh with 25,000 quadrilateral elements.
Adjust params to suit your GPU (and patience).
save as file "cflow2d.geo", then open in gmsh
- - - - - - - -
lc = 1.00; // Q: 25,4532; [X] Recombine all triangular meshes
xmin = -100.0;
ymin = 0.0;
xmax = 100.0;
ymax = 100.0;
zall = 0.0;
Point(1) = {xmin, ymin, zall, lc};
Point(2) = {xmax, ymin, zall, lc};
Point(3) = {xmax, ymax, zall, lc};
Point(4) = {xmin, ymax, zall, lc};
//---------------------------------------
Line(1) = {1,2};
Line(2) = {2,3};
Line(3) = {3,4};
Line(4) = {4,1};
LL=newll; Line Loop(LL) = {1,2,3,4};
PS=news; Plane Surface(PS) = {LL};
// optionally, recombine triangles to quads
Recombine Surface {PS};
Physical Surface("Fluid", 1) = {PS};
Physical Line("periodic_0_r", 2) = {2};
Physical Line("periodic_0_l", 3) = {4};
Physical Line("bcwalllower", 4) = {1};
Physical Line("bcwallupper", 5) = {3};
// - - - - -
// end of file
Good luck!
Nigel
On Thursday, March 22, 2018 at 9:11:47 PM UTC+11, Mahmood Naderan wrote:
>
> Hi
> According to the user guide, I ran the couette_flow_2d example and saw
> an output like below
>
> mahmood@orca:~$ cp -r /opt/PyFR-1.7.5/examples/couette_flow_2d/ .
> mahmood@orca:~$ cd couette_flow_2d/
> mahmood@orca:couette_flow_2d$ pyfr import couette_flow_2d.msh
> couette_flow_2d.pyfrm
> /home/mahmood/.local/lib/python3.6/site-packages/h5py/__init__.py:36:
> FutureWarning: Conversion of the second argument of issubdtype from
> `float` to `np.floating` is deprecated. In future, it will be treated
> as `np.float64 == np.dtype(float).type`.
> from ._conv import register_converters as _register_converters
> mahmood@orca:couette_flow_2d$
> mahmood@orca:couette_flow_2d$ ls
> couette_flow_2d.ini couette_flow_2d.msh couette_flow_2d.pyfrm
> mahmood@orca:couette_flow_2d$ which pyfr
> /usr/local/bin/pyfr
> mahmood@orca:couette_flow_2d$ pyfr run -b cuda -p
> couette_flow_2d.pyfrm couette_flow_2d.ini
> /home/mahmood/.local/lib/python3.6/site-packages/h5py/__init__.py:36:
> FutureWarning: Conversion of the second argument of issubdtype from
> `float` to `np.floating` is deprecated. In future, it will be treated
> as `np.float64 == np.dtype(float).type`.
> from ._conv import register_converters as _register_converters
> 100.0% [=================================================>] 4.00/4.00
> ela: 00:03:35 rem: 00:00:00
>
>
>
> Couple of things:
>
> 1- Only one cpu core was 100%.
> 2- GPU utilization according to nvidia-smi was about 30%.
>
> Is that normal?
>
>
> Regards,
> Mahmood
>
--
You received this message because you are subscribed to the Google Groups "PyFR
Mailing List" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send an email to [email protected].
Visit this group at https://groups.google.com/group/pyfrmailinglist.
For more options, visit https://groups.google.com/d/optout.
