Compute pressure from microphysics eos

[psharda]

This is my idea of how we want to compute pressure from microphysics. But, it might be better to move this function to EOS.hpp so all the EOS functions are together?

Anyway, first, need this to be approved: AMReX-Astro/Microphysics#1301

[BenWibking]

I think it would make sense to move the new code to EOS.hpp, but make it a function that only needs scalar arguments (e.g., amrex::Real rho, amrex::Real Eint). Then the existing HydroSystem::ComputePressure just calls that function for a given cell.

[psharda]

But, for primordial chem, we will need the mass scalars too

[BenWibking]

Yeah, forgot about that for a sec :) The design you have now makes sense to me.

[psharda]

Aha, encountered the first issue. Take the example of HydroHighMach test. For this test, the mean molecular weight is NAN. So, the gamma_law EOS in microphysics returns a NAN pressure, because it calculates the pressure using the temperature. The dependence on the mean molecular weight cancels out in the end (state.mu / state.mu), but because state.mu = NAN, state.mu/state.mu = NAN.

The fix I can think of is to modify microphysics to use the internal energy directly in place of temperature to calculate the pressure.

[psharda]

Aah, can't do that because gamma_law uses pressure to update the internal energy.

[BenWibking]

Aha, encountered the first issue. Take the example of HydroHighMach test. For this test, the mean molecular weight is NAN. So, the gamma_law EOS in microphysics returns a NAN pressure, because it calculates the pressure using the temperature. The dependence on the mean molecular weight cancels out in the end (state.mu / state.mu), but because state.mu = NAN, state.mu/state.mu = NAN.

The fix I can think of is to modify microphysics to use the internal energy directly in place of temperature to calculate the pressure.

You could do that. Another option is to just set the mean molecular weight to some arbitrary but non-zero and non-NAN value.

[psharda]

Aha, encountered the first issue. Take the example of HydroHighMach test. For this test, the mean molecular weight is NAN. So, the gamma_law EOS in microphysics returns a NAN pressure, because it calculates the pressure using the temperature. The dependence on the mean molecular weight cancels out in the end (state.mu / state.mu), but because state.mu = NAN, state.mu/state.mu = NAN.
The fix I can think of is to modify microphysics to use the internal energy directly in place of temperature to calculate the pressure.

You could do that. Another option is to just set the mean molecular weight to some arbitrary but non-zero and non-NAN value.

Oh, if that's possible, then that's the easiest thing to do without changing stuff in microphysics. Okay, I will then set the mean molecular weight to 1.0 for this test (and tests like these that might fail too)

[BenWibking]

Aha, encountered the first issue. Take the example of HydroHighMach test. For this test, the mean molecular weight is NAN. So, the gamma_law EOS in microphysics returns a NAN pressure, because it calculates the pressure using the temperature. The dependence on the mean molecular weight cancels out in the end (state.mu / state.mu), but because state.mu = NAN, state.mu/state.mu = NAN.
The fix I can think of is to modify microphysics to use the internal energy directly in place of temperature to calculate the pressure.

You could do that. Another option is to just set the mean molecular weight to some arbitrary but non-zero and non-NAN value.

Oh, if that's possible, then that's the easiest thing to do without changing stuff in microphysics. Okay, I will then set the mean molecular weight to 1.0 for this test (and tests like these that might fail too)

Inelegant, but probably the easiest solution.

[psharda]

/azp run

[psharda]

@BenWibking some tests fail when -DENABLE_ASAN=ON, and the failure is caused by a division by zero error in ComputePressure in EOS.hpp, on the line where we do state.e = Eint / rho. This error does not appear when the address sanitizer is not enabled. How do I debug this error?

[BenWibking]

@BenWibking some tests fail when -DENABLE_ASAN=ON, and the failure is caused by a division by zero error in ComputePressure in EOS.hpp, on the line where we do state.e = Eint / rho. This error does not appear when the address sanitizer is not enabled. How do I debug this error?

You can add an if statement that checks whether rho == 0. If so, you can then trigger a breakpoint that you can then examine in gdb: https://stackoverflow.com/a/4326474.

[psharda]

@BenWibking the remaining test (debug arm64 gcc) fails because RadForce fails. This test is likely failing because it uses isothermal gamma, and the function ComputePressure (which is what hydro_system.hpp uses everywhere) returns a NAN pressure for isothermal gamma.

What should be the pressure when gamma is isothermal? Its not zero right? Do you find it from the (finite, non-zero) isothermal sound speed, like here:

quokka/src/hydro_system.hpp

Line 482 in aa288b5

Pplus2 = primVar(i + 2, j, k, primDensity_index) * cs_sq;

[BenWibking]

@BenWibking the remaining test (debug arm64 gcc) fails because RadForce fails. This test is likely failing because it uses isothermal gamma, and the function ComputePressure (which is what hydro_system.hpp uses everywhere) returns a NAN pressure for isothermal gamma.

What should be the pressure when gamma is isothermal? Its not zero right? Do you find it from the (finite, non-zero) isothermal sound speed, like here:

quokka/src/hydro_system.hpp

Line 482 in aa288b5

Pplus2 = primVar(i + 2, j, k, primDensity_index) * cs_sq;

Yes, it's $\rho c_s^2$.

[psharda]

/azp run

[psharda]

all tests should pass now

[BenWibking]

It looks like the SphericalCollapse test failed.

[psharda]

/azp run

[psharda]

all done @BenWibking

[BenWibking]

Minor code style changes.

[psharda]

/azp run

resolved