search

geodynamics / sw4

SW4 (Seismic Waves, 4th order) implements substantial capabilities for 3-D seismic modeling, with a free surface condition on the top boundary, absorbing super-grid conditions on the far-field boundaries, and an arbitrary number of point force and/or point moment tensor source terms.
Other
133 stars 65 forks source link

Restart issue: time series files don't match on restart #33

Closed drhansj closed 6 years ago

drhansj commented 6 years ago

Using 81660fd, running a version of examples/meshrefine/loh1-slow-mr-order4.in but with restart.

One test that works fine (usgs ascii, sac binary) on restart is here: performance/scec/LOH.1-h400-check.in

I tried a bunch of other tests beyond Artie's mini Hayward, including ones with mesh refinement, attenuation (3 mech, 1 mech).

One I tried with and w/o MR is this one:

If I comment out the refinement line, then everything is correct on restart and matches the baseline.

drhansj commented 6 years ago

@andersp and @bjorn2 would appreciate any ideas you have to investigate this.

Can you think of a place in the code where MR might require extra initialization (ghost cells, materials, etc.) after reading in from a restart file?

drhansj commented 6 years ago

@andersp : I found a second example, loh3 with mr and att (3 mech). When I comment out the mr and att, I get matching time series data. So extra initialization with attenuation is probably an issue, too.

andersp commented 6 years ago

One routine that needs to be called for MR to work properly is setup_MR_coefficients(). It is called from setupRun(). This could only be a problem if you never call that routine on restart.

drhansj commented 6 years ago

Ok - thanks, I will check.

How about for attenuation?

On May 30, 2018, at 1:53 PM, Anders Petersson notifications@github.com wrote:

One routine that needs to be called for MR to work properly is setup_MR_coefficients(). It is called from setupRun(). This could only be a problem if you never call that routine on restart.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.

andersp commented 6 years ago

Let’s first get MR without attenuation going.

From: Hans Johansen notifications@github.com Reply-To: geodynamics/sw4 reply@reply.github.com Date: Wednesday, May 30, 2018 at 2:53 PM To: geodynamics/sw4 sw4@noreply.github.com Cc: Default petersson1@llnl.gov, Mention mention@noreply.github.com Subject: Re: [geodynamics/sw4] Restart issue: time series files don't match on restart (#33)

Ok - thanks, I will check.

How about for attenuation?

On May 30, 2018, at 1:53 PM, Anders Petersson notifications@github.com wrote:

One routine that needs to be called for MR to work properly is setup_MR_coefficients(). It is called from setupRun(). This could only be a problem if you never call that routine on restart.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub, or mute the thread.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/geodynamics/sw4/issues/33#issuecomment-393332150, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AIgih_AhC-Tai5AiwqHznNWbR1Tv25M_ks5t3xThgaJpZM4UT5s7.

drhansj commented 6 years ago

Anders was able to find the problem(s) - the simulation wasn't setup correctly on restart! It needs 1 extra ghost point for mesh refinement, and one for ghost point values for viscoelastic attenuation.