librustc_mir: Remove &*x when x has a reference type.

[pcwalton]

This introduces a new InstCombine pass for us to place such peephole
optimizations.

r? @eddyb

[pcwalton]

This should open up some new copy prop opportunities.

[cbreeden]

Does this include *& as well? Something like how Deref is implemented on &'a T?

[eddyb]

@cbreeden No, that's somewhat more complex because it's not self-contained within a statement.

[pcwalton]

How's this?
r? @eddyb

[eddyb]

r=me with comments addressed. It might be good to also explain (in a comment?) why the optimizations are collected ahead of time instead of done in-place.

[eddyb]

Should use FnvHashSet (from rustc::utils::nodemap) in the compiler.

[eddyb]

This can be done via #[derive(Default)].

[eddyb]

Can use remove instead of contains to do both at once.

[eddyb]

Heh, nice placeholder hack.

[eddyb]

Should use bug! in the compiler.

[pcwalton]

Updated. r? @eddyb

[eddyb]

@bors r+

[eddyb]

@bors r+

[bors]

📌 Commit 11a4965 has been approved by eddyb

[arielb1]

@bors r-

This will cause drop elaboration to crash and burn, because the newly-created code moves behind &mut references.

[eddyb]

Ah, right, my bad, I forgot that it had to go after drop elaboration. We need more comments in the pass list.

[arielb1]

I would like to have more understanding on what properties MIR must have at every step before we start doing destructive optimizations.

[pcwalton]

@arielb1 Right now we are generating very bad code. I think it would be unwise to gate needed optimizations on "understanding of what properties MIR must have at every step". We need these optimizations now. As long as everything works, we can do that later.

[eddyb]

r=me with the pass reordered.

[eddyb]

You only need to reorder the pass to go after this line, which should be separated and commented.
This is the point where moves stop existing and you only have raw semantic-less copies.

[pcwalton]

@bors-servo: r=eddyb

[bors]

📌 Commit e8a44d2 has been approved by eddyb

[bors]

⌛ Testing commit e8a44d2 with merge d37e54b...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-cargotest, auto-linux-64-cross-freebsd, auto-linux-64-debug-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-opt-rustbuild, auto-linux-64-x-android-t, auto-linux-cross-opt, auto-linux-musl-64-opt, auto-mac-32-opt, auto-mac-64-opt, auto-mac-64-opt-rustbuild, auto-mac-cross-ios-opt, auto-win-gnu-32-opt, auto-win-gnu-32-opt-rustbuild, auto-win-gnu-64-opt, auto-win-msvc-32-opt, auto-win-msvc-64-cargotest, auto-win-msvc-64-opt, auto-win-msvc-64-opt-rustbuild
Approved by: eddyb
Pushing d37e54b to master...