Stream: t-compiler/wg-rls-2.0

Topic: C++ IR: IPR


matklad (Jun 17 2020 at 08:55, on Zulip):

Here's an interesting paper describing semantic representation for C++

A Principled, Complete, and Efficient Representation of C++

It seems like we might want something like this for our hir/code_model

matklad (Jun 17 2020 at 08:59, on Zulip):

In particular, this 2000 lines header describes the whole of C++ https://github.com/GabrielDosReis/ipr/blob/master/include/ipr/interface#L1

Our code model should look like this, at least in spirit.

Jeremy Kolb (Jun 17 2020 at 13:09, on Zulip):

I'm a little surprised that they can model the language in such an elegant way. I forget sometimes that C++ isn't as crazy as some people would like it to be.

matklad (Jun 17 2020 at 13:13, on Zulip):

The main trick is that the IR is just a data structure. Really, when you can say that things just immutably point to each other, you get a pretty nice API. That's fundamentally incompatible with lazyness though :(

Last update: Sep 27 2020 at 14:15UTC