feat(riscv): basic datalayout support

[regehr]

this only does the RISC-V side, but we also want datalayout in the LLVM interpreter since it interacts with memory objects as well

[naveen-seth]

Otherwise, LGTM! I’m still new to the project/this area, so I’ll leave the approval to someone else.

[regehr]

Otherwise, LGTM! I’m still new to the project/this area, so I’ll leave the approval to someone else.

fine to approve (or not) as you choose! we typically wait for all reviewers to approve, or at least most, depending on the complexity of the PR

[math-fehr]

I think that's good to me, I don't know how it's done in MLIR/LLVM right now, but you know better!

[regehr]

I think that's good to me, I don't know how it's done in MLIR/LLVM right now, but you know better!

you can find MLIR's version here:

https://mlir.llvm.org/docs/DataLayout/

the goal in both cases is to provide an interface that answers questions about how objects want to be represented in RAM. there isn't any question that we need to be able to answer these questions, but of course we might disagree on how to best answer them.

the answer here is "these are hard-coded, but then exposed through a generic API." this is far less ambitious than what MLIR provides, which includes things like parsing and printing LLVM-style datalayout strings.

anyway I'm happy to talk about this, discuss alternatives, etc. but for now, this solves practical problems that we've already encountered -- and solved in ad hoc ways.

[luisacicolini]

lgtm, thank you :)