binrw VS inlinable-dyn-extension-traits

Serde can work but won't be easy, because It wants to deserialize everything. Take a look at the binrw crate(https://github.com/jam1garner/binrw) and define your binary layout with It, then you can actually create a wrapper which reads the file from any reader(be it a Cursor over the mmap'd slice or a file).

Sure! The commit is here (I haven’t pushed the version of this commit I made on binread itself but the actual changes are identical)

In the meantime, you could check out daniel5151/optional-trait-methods for some chicken-scratch notes + exploration on IDETs, including some code samples which you can throw right into godbolt.org to see how well they optimize.

Funny enough, I've also been meaning to write a full blog post about it, but haven't found the time! In the meantime, I've got a rough overview of this technique as part of the gdbstub docs, and I've got a WIP repo of example code and exploratory notes that explores the technique in more detail (along with comparing it with alternative approaches)

Also, check out daniel5151/optional-trait-methods if you haven't already (I point to it near the end of the gdbstub documentation I linked). It's a very rough overview of all the various techniques I considered before landing on IDETs, and includes several simplified examples of how the resulting APIs might work from a implementation and user's perspective.

I've spent quite a bit of time staring at assembly output and performing in-application testing to make sure that optimizations were being triggered as expected.