Question about variable order and solve order in PRISMS-PF

[Jorge Barrios]

Hello,

I am implementing a Demange-style ice model in PRISMS-PF using the alloy_solidification application as a template. My fields are:

• xi1
• phi
• u

I want the effective update order to be:

1. xi1
2. phi
3. u

My question is: how does PRISMS-PF determine the order in which these equations are assembled or advanced? Does the order in loadVariableAttributes() matter directly, especially when mixing EXPLICIT_TIME_DEPENDENT and AUXILIARY variables?

More specifically:

1. Does variable index order affect solve/evaluation order?
2. If I want xi1 -> phi -> u, is setting the variables in that order enough?
3. Is there a recommended way to guarantee an auxiliary field like xi1 is evaluated before the explicit updates that depend on it?

I am currently testing different orderings in the meantime, such as:

• xi1(0), phi(1), u(2)

Any clarification would be appreciated. Thanks.

[Alexander Mensah]

Hello,

The order in the variable attribute loader does not matter.

In PRISMS-PF versions < 3.0, the order of the solves is set automatically and cannot be overridden. AUXILIARY variables are solved before EXPLICIT_TIME_DEPENDENT.

In PRISMS-PF version 3.0, the order of the solves is also set automatically, but can be overridden. The way to override this behavior and control the order is by setting the solve block of each field in the order you want them solved in the variable attribute loader.

In PRISMS-PF version 4.0 (which is to be released any day, but available now on the 4.0-dev branch), you always explicitly construct the solve order yourself, avoiding this potential ambiguity.

The alloy solidification application is also still available as an example on 4.0.

Hope this helps. Let me know if there's anything else.

Xander

[Jorge Barrios]

Hello,

Thank you, that clarifies my question and simplifies my approach.

Best,
Jorge Barrios