Is there any way to speed up the PDE solver with the parallel computing toolbox?

Question

Steven 2018-9-25

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/420739-is-there-any-way-to-speed-up-the-pde-solver-with-the-parallel-computing-toolbox

编辑： Steven 2018-9-26

Can I speed up this code by paralleling it with the parallel computing toolbox?

thermalmodel = createpde('thermal','transient');
gm = multicuboid(1,1,0.5,'ZOffset',0.5);%if this doesn't work choose your own geometry
thermalmodel.Geometry = gm;
%%%%%%%%%%%%%%%%%%%%%%%GENERATE MESH %%%%%%%%%%%%%%%
% meshsize = 0.4;%mm                           
meshsize = 4;%mm                           
generateMesh(thermalmodel,'Hmax',meshsize/1000);
thermalmodel.Mesh
pdemesh(thermalmodel)                           
htc = vpa(25) *u.Watt/(u.m^2*u.Kelvin);
htc = rewrite(htc,u.Watt/(u.in^2*u.Kelvin));
htc = double(separateUnits(htc));
thermalBC(thermalmodel,'Face',1:thermalmodel.Geometry.NumFaces,'ConvectionCoefficient',htc,'AmbientTemperature',288.15);
thermalBC(thermalmodel,'Face',1,'Temperature',50+273);%bottom face
mybc = @(region,state) sin(350*state.time).^2;
thermalBC(thermalmodel,'Face',12,'Temperature',mybc);%top face
tfinal = 6;
tlist = 0:3:tfinal;
thermalIC(thermalmodel,0);
thermalmodel.SolverOptions.ReportStatistics = 'on';
result = solve(thermalmodel,tlist);
figure; pdeplot3D(thermalmodel)
pdeplot3D(thermalmodel,'XYData',T(:,end),'Contour','on','ColorMapData',results.Temperature); 
title('Temperature in K')
colorbar
pdeplot3D(thermalmodel,'ColorMapData',results.Temperature); 
axis equal
title(sprintf('Transient Temperature at Final Time (%g seconds)',tfinal));

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Alan Weiss 2018-9-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/420739-is-there-any-way-to-speed-up-the-pde-solver-with-the-parallel-computing-toolbox#answer_338518

Sorry, currently Partial Differential Equation Toolbox™ does not use parallel computing.

Alan Weiss

MATLAB mathematical toolbox documentation

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

Answer 2

Ravi Kumar 2018-9-26

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/420739-is-there-any-way-to-speed-up-the-pde-solver-with-the-parallel-computing-toolbox#answer_338531

Your mesh density is too high. I could not get the mesh generated on my machine in a reasonable time. This means, there are large number of DoFs. Can you run your problem first with the default mesh, by using:

generateMesh(thermalmodel);

instead of:

generateMesh(thermalmodel,'Hmax',meshsize/1000);

I am sure you will be able to solve this on a single machine. Then see how much mesh resolution you really need.

Regards, Ravi

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Steven 2018-9-26

编辑：Steven 2018-9-26

it works fine on my machine even with mesh densities 100x this, the mesh density doesn't matter, the transient solution is not working. I'm more concerned about how to use the PDE toolbox with the parallel computing toolbox

请先登录，再进行评论。