pyfrs writer occasionally failing

[Junting Chen]

Hello all,

I am experiencing failures from the writer function occasionally. For instance, once I have -

[soln-plugin-writer]

dt-out = 50

basedir = .

basename = concaveTower_{t:04.0f}

- in the ini file, and this simulation started from t=1800. It supposed to write out a file at t=1850 but nothing was written.

Another time I had dt-out = 20, the simulation only wrote out properly until t=60. After that simulation carried on but no pyfrs file wrote out. 

In both cases I am sure the simulation runs ok because the velocity and pressure probes wrote out properly. 

Is anyone experiencing the same issue?

Junting Chen

[Niki Loppi]

Hi Junting,

With dual integrator, in certain circumstances the physical time counter can accumulate round-off error and if the error exceeds 5e-12 at an output time, the writer plugin does not realise it needs to write. As you said, the simulation should run ok and this bug only affects the writer plugin with dual integrator.

For now please use the attached patch which forces the output at the next physical time-step. I will submit a fix to this asap.

Thanks,

Niki

--
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 pyfrmailingli...@googlegroups.com.
To post to this group, send email to pyfrmai...@googlegroups.com.
Visit this group at https://groups.google.com/group/pyfrmailinglist.
To view this discussion on the web, visit https://groups.google.com/d/msgid/pyfrmailinglist/41b781ac-9aca-4ca0-b7f0-892a65cfb72d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Junting Chen]

Thanks for the quick response!

Junting Chen

On Wednesday, July 3, 2019 at 1:09:40 PM UTC-4, Niki Loppi wrote:

Hi Junting,

With dual integrator, in certain circumstances the physical time counter can accumulate round-off error and if the error exceeds 5e-12 at an output time, the writer plugin does not realise it needs to write. As you said, the simulation should run ok and this bug only affects the writer plugin with dual integrator.

For now please use the attached patch which forces the output at the next physical time-step. I will submit a fix to this asap.

Thanks,

Niki

On 03/07/19 15:50, Junting Chen wrote:

Hello all,

I am experiencing failures from the writer function occasionally. For instance, once I have -

[soln-plugin-writer]

dt-out = 50

basedir = .

basename = concaveTower_{t:04.0f}

- in the ini file, and this simulation started from t=1800. It supposed to write out a file at t=1850 but nothing was written.

Another time I had dt-out = 20, the simulation only wrote out properly until t=60. After that simulation carried on but no pyfrs file wrote out. 

In both cases I am sure the simulation runs ok because the velocity and pressure probes wrote out properly. 

Is anyone experiencing the same issue?

Junting Chen

--
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 pyfrmai...@googlegroups.com.

[Douglas Fontes]

Hello Junting,

Did you implement the patch proposed by Niki. I am trying to do this, but I am not sure which folders/files I have to change. After the changes, is it necessary to recompile pyFR? Can you help me?

Thank you.

[Junting Chen]

Hello Douglas,

Yes I did implement the patch.

No, you dont need to recompile pyfr. I think what i did was dropping the patch file in pyfr directory and simply type:

patch < 0001-Writer-patch.patch 

Try it, everything works great so far. Thanks Niki.

Best regards,

Junting Chen