rustc_codegen_ssa: introduce MIR VarDebugInfo, but only for codegen.

[eddyb]

These are all the codegen changes necessary for #56231.

The refactors were performed locally to codegen, and in several steps, to ease reviewing and avoid introducing changes in behavior (as I'm not sure our debuginfo tests cover enough).

r? @michaelwoerister cc @nagisa @rkruppe @oli-obk

[eddyb]

As per #56231 (comment):

@bjorn3 Please leave comments on the commit in the PR's commit list, not on the commit itself, because I can't reply to the latter from the PR view, and they might even disappear.

[eddyb]

r? @nikomatsakis

[nikomatsakis]

🎉 for doc-comments

[nikomatsakis]

@bors r+

[bors]

📌 Commit 2994186 has been approved by nikomatsakis

[nikomatsakis]

@bors r-

Until @eddyb addresses comments by @bjorn3

[eddyb]

Oops, I forgot that @bjorn3's comments were all about pre-existing problems, in the code I moved from mod.rs to debuginfo.rs (in src/librustc_codegen_ssa/mir).

@bors r=nikomatsakis rollup=never

[bors]

📌 Commit 2994186 has been approved by nikomatsakis

[bors]

⌛ Testing commit 2994186 with merge 404d42f...

[eddyb]

Weird, why does wasm get no temporaries here? Are we more aggressive somehow?
Do we not put vectors behind pointers in wasm, and do on other targets, by any chance?

EDIT: more exactly, what's happening here is that on all targets %x is an LLVM parameter of the function, but then when the intrinsic is called:

• most targets use %_2, which was created from %x (presumably via load?)
• wasm uses %x directly (so the function signature surely takes it by value)

I can trivially adjust the test so both forms are allowed, but I'm suspicious of the behavior.

cc @alexcrichton @tlively

[alexcrichton]

@eddyb this looks simd related and wasm-specific so it's likely related to this configuration option where simd values are passed by-value in wasm and not by-reference, because wasm has entirely different validation/register rules than traditional elf executables.

Perhaps it's safe to just ignore the test on wasm?

[tlively]

In one of my recent PRs I generalized one of the SIMD tests by replacing the [0-9]+ with [0-9a-z]+ so it would match x as well. Perhaps it would make sense to do the same here?

[eddyb]

@alexcrichton Thanks, that definitely explains it!

@tlively I'll use %{{x|_2}} (and similar), I think, and leave a comment explaining why that is, to avoid any surprises in the future.

[eddyb]

@bors r=nikomatsakis

[bors]

📌 Commit 60a2266 has been approved by nikomatsakis

[bors]

⌛ Testing commit 60a2266 with merge 01e5d91...

[bors]

☀️ Test successful - checks-azure
Approved by: nikomatsakis
Pushing 01e5d91 to master...