[WIP] Timestep limiter for hydrodynamics

This implements #196 (closed).

We create an extra loop after the second kick where we verify that particles do not have neighbours with vastly different signal velocities. If they do, we wake up the inactive particle.

Currently works with all the hydro schemes as well as with gravity.

Missing for now:

• Deal with MPI.

@bvandenbroucke if you confirm that it all makes sense also for gizmo, I'll work on the last two points.

added 1 commit

• b28ddc5a - Moved vmax initialization to hydro_prepare_force to accommodate timestep limiter.

Compare with previous version

added 1 commit

• a600c531 - Make the 1D Sedov explosion ICs properly symmetric.

Compare with previous version

added 1 commit

• 4bf5348d - Make the time-step limiter factor a constant defined in const.h

Compare with previous version

Ok. So what I think I will do is put the 1D hydro problem aside for now. I'll make sure this works with gravity and MPI so that it can be merged in. Otherwise it will remain open forever and no progress will be made.

added 1 commit

• a44d1628 - If a particle was limited in the previous step and gets limited again, the…

Compare with previous version

added 13 commits

• a44d1628...19ca1efd - 9 commits from branch master
• 7ddb98f0 - Merge branch 'master' into timestep_limiter2
• 380eaa8d - Code formatting
• 72d29cfa - Removed debugging code. Added correct weight for the time-step limiter task.
• ed6a1275 - Move the synchronization of accelerations to a separate stand-alone function.

Compare with previous version

added 4 commits

• 2c613a6c - Do not use the time-step limiter for the 1D Sedov blast.
• c2178852 - Write the engine policy to the snapshots.
• bbe47d69 - Removed spurious debugging message
• d2410c23 - Write the value of the time-step limiter ratio to the snapshots.

Compare with previous version

Can you guarantee the problem only affects 1D?

No I cannot because I don't know yet what the issue is exactly. But I have not seen other examples fail.

added 14 commits

• d2410c23...96c1f4d8 - 12 commits from branch master
• 9dcfcec3 - Merge branch 'master' into timestep_limiter2
• 4e401357 - Missing parameters in the dependency creation function.

Compare with previous version

added 2 commits

• 38db1ebb - When gravity is active, also update the gpart counter-part's time-bin.
• cb2cf4e7 - Remove the use of the double-limiting for now.

Compare with previous version

We can now run with hydro+gravity+limiter. Last step: MPI.

changed the description

added 1 commit

• 5178a66e - Create a dependency between time-step and kick1 for the cases where no time-step…

Compare with previous version

added 41 commits

• 5178a66e...4aad7761 - 40 commits from branch master
• bbf00215 - Merge branch 'master' into timestep_limiter2

Compare with previous version

added 1 commit

• 9ef4c069 - Also reinstate the dependency between end-force and kick2 for the case where…

Compare with previous version

added 2 commits

• a8e380a5 - More secure naming convention for the MPI tags.
• 52b17d7e - Created the communication tasks for the timestep-limiter.

Compare with previous version

added 1 commit

• 2fc83521 - Code formatting

Compare with previous version

added 3 commits

• 2fc83521...eb768e46 - 2 commits from branch master
• 85098f84 - Merge branch 'master' into timestep_limiter2

Compare with previous version