Add InstanceStruct test data generation helper and unit test

[iomaganaris]

• Taken over work from https://github.com/BlueBrain/nmodl/tree/pramodk/test-helper
• Added mechanism to instantiate the InstanceStruct with deterministic data
• Compare those in a unit test
• Check if typecasting works in the unit test
• Review and cleanup by Pramod

[pramodk]

@georgemitenkov : going to do some cleanup but this overall gives an idea of what creating data data. See test/unit/codegen/codegen_llvm_instance_struct.cpp.

[georgemitenkov]

@pramodk @iomaganaris looks great! On LLVM/JIT side, I am working on the following:

1. Generating a wrapper LLVM function that takes a void pointer and casts it to the struct LLVM pointer type.
2. Calling a kernel.

Additionally I am thinking of creating functions that can be lowered to LLVM that use CodegenInstanceData to print/compare the results.

[pramodk]

Additionally I am thinking of creating functions that can be lowered to LLVM that use CodegenInstanceData to print/compare the results.

I am curious what you meant by lowered to LLVM here. If we add such functions in CodegenInstanceData, would that fit the purpose?

[georgemitenkov]

@pramodk actually you are right, and we do not strictly need it I think. We can always define these on the higher level in CodegenInstanceData. I will make changes to the JIT first, and then we will see if we really need something like this:)

[georgemitenkov]

@pramodk @iomaganaris I added a sample test case in #549 to show how this would look on LLVM side. I think I have all functions I need in this PR - for the data generation and for member comparisons.

The printing functions are not really necessary - I guess we may need them for visual interpretation only :) So i would say that printfs are sufficient for that now!

[georgemitenkov]

@iomaganaris @pramodk I have noticed that when the NEURON does not have current, SOLVE states METHOD cnexp block is added to the kernel for loop:

for ( ... ) {
   SOLVE states METHOD cnexp
   ...
}

This does not affect code generation at all, but I am wondering if we should raise a ticket for this :)

[pramodk]

@georgemitenkov : I also so that somewhere and its a bug. Could you add ticket with some details?

[pramodk]

@iomaganaris : could you look at CI fixes?

[iomaganaris]

@iomaganaris : could you look at CI fixes?

Yy. I am looking at that atm 👍