-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Allow MIR debuginfo to point to a variable's address #111440
Conversation
r? @jackh726 (rustbot has picked a reviewer for you, use r? to override) |
Some changes occurred to MIR optimizations cc @rust-lang/wg-mir-opt |
7394a57
to
29ff2bc
Compare
r? mir-opt |
r=me with the debuginfo test extended to have a case where |
☔ The latest upstream changes (presumably #111493) made this pull request unmergeable. Please resolve the merge conflicts. |
29ff2bc
to
8fb888d
Compare
That's perfect, thanks! @bors r+ |
☀️ Test successful - checks-actions |
Finished benchmarking commit (bc88895): comparison URL. Overall result: ✅ improvements - no action needed@rustbot label: -perf-regression Instruction countThis is a highly reliable metric that was used to determine the overall result at the top of this comment.
Max RSS (memory usage)ResultsThis is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.
CyclesResultsThis is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.
Binary sizeResultsThis is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.
Bootstrap: 661.349s -> 660.613s (-0.11%) |
MIR optimizations currently do not to operate on borrowed locals.
When enabling #106285, many borrows will be left as-is because they are used in debuginfo. This pass allows to replace this pattern directly in MIR debuginfo:
a => _1 _1 = &raw? mut? _2
becomes
This pass is implemented as a drive-by in ReferencePropagation MIR pass.
This transformation allows following following MIR opts to treat _2 as an unborrowed local, and optimize it as such, even in builds with debuginfo.
In codegen, when encountering
a => &..&_2
, we create a list of allocas:Caveat: this transformation looses the exact type, we do not differentiate
a
as a immutable, mutable reference or a raw pointer. Everything is declared to*mut
to codegen. I'm not convinced this is a blocker.