[handshake-sim] Added complete modularity for execution models

[Polymeth]

I still have fix to change, which I'm doing this week. Here are my concerns and stuff to do:

• A bit of 'boilerplate' and similar code (struct definition & declaration...)
• Some functions doesn't have an 'execute' function, but removing the virtual status of it makes the struct without 'execute' abstract, so impossible to store in our map/instanciate
• Boring dyn_cast in tryExecute, which is mandatory for correct overriding I think : might keep it otherwise we need a type erasure wrapper in order to store structs which doesnt look really good imo
• Might want can get rid of the tryExecute+execute system of the default simulator which isn't very clear and modular-friendly (Have to redesign a bit more then!)
• Might want to centralize models map and structures so that users only code in one file
• If configuration isn't found, maybe we can automaticaly substitude with the default configuration ?
• Clean the base simulator code which is kinda messy + fix LLVM codestyle from the default code

[lucas-rami]

Left some comments for code style, but more like general recommendations than direct changes.

I did not really look into how the execution models are working internally since I do not have the time (I understood at some point in the past but forgot) but more at the general way you implemented execution models.

All in all I really like it, I think it's going in a great direction! Perhaps sometime early next week we can see the progress and define some kind of milestone at which point we will be able to merge this. Thanks!

[lucas-rami]

A bit of 'boilerplate' and similar code (struct definition & declaration...)

Kind of unavoidable to some extent (I think I see what boilerplate you're referring to), don't worry too much at this point.

If configuration isn't found, maybe we can automaticaly substitude with the default configuration ?

In general I would be relatively opposed to fail-safe behaviors i.e., as the program if you provide me with a configuration file that does not exist I would rather scream at you in an error message than proceed with a default configuration (which the user might believe is the custom config they specified and then be confused about the results, or worse, take the results for granted). However, if you're just saying "if the user provides no explicit config, use the default one", then yes I totally agree.

Clean the base simulator code which is kinda messy + fix LLVM codestyle from the default code

Sounds great!

[lucas-rami]

Couple of nit-picky things to address + a few "real" remarks with easy fixes, but otherwise this looks good!

[lucas-rami]

Thanks for the quick fixes, this looks good to me!