Add OwnedTargetMachine to manage llvm:TargetMachine

[nebulark]

LLVMRustDisposeTargetMachine taking a &mut could be undefined behaviour.
Wrapping it with a struct and using pointers instead avoids this problem.
In addition the TargetMachine is now automatically freed via the Wrappers drop impl. This should fix some memory leaks when
create_informational_target_machine was used, e.g.

rust/compiler/rustc_codegen_llvm/src/llvm_util.rs

Lines 291 to 314 in 327e6cf

	pub fn target_features(sess: &Session, allow_unstable: bool) -> Vec<Symbol> {
	let target_machine = create_informational_target_machine(sess);
	supported_target_features(sess)
	.iter()
	.filter_map(|&(feature, gate)| {
	if sess.is_nightly_build() || allow_unstable || gate.is_none() {
	Some(feature)
	} else {
	None
	}
	})
	.filter(|feature| {
	// check that all features in a given smallvec are enabled
	for llvm_feature in to_llvm_features(sess, feature) {
	let cstr = SmallCStr::new(llvm_feature);
	if !unsafe { llvm::LLVMRustHasFeature(target_machine, cstr.as_ptr()) } {
	return false;
	}
	}
	true
	})
	.map(|feature| Symbol::intern(feature))
	.collect()
	}

r? @Nilstrieb

[Nilstrieb]

Thanks! have a few comments for small improvements, but overall this looks good and should improve the code.

[Nilstrieb]

Not sure about the best name for this. I think TargetMachineWrapper isn't the best name, but it's also not bad. TargetMachine would be confusing. If you can think of something better, do rename it, but if you can't that's fine too and we can keep it.

[nebulark]

Some ideas might be

• TargetMachineBox (it kindo of behaves like a box, but its not copyable )
• TargetMachineLLVM (it the the for the associated type TargetMachine for the LLVM backend.
• TargetMachineOwned / OwnedTargetMachine
• UniqueTargetMachine / TargetMachineUnique

[nebulark]

Renamed it to TargetMachineBox. I think that fits what I am going for and should good give a good intuition of what it does. Also clarified this in the comments.

[Nilstrieb]

I like OwnedTargetMachine more than TargetMachineBox, but TargetMachineBox is fine too. If you want to, you can rename it again, or if you think it's fine and don't want to you can just use @bors r=Nilstrieb and approve the PR.

[nebulark]

Left some comments. Will be implementing those tomorrow.

[nebulark]

Squashed commits

@rustbot review

[Nilstrieb]

r=me with or without another rename
@bors delegate+

[bors]

✌️ @nebulark, you can now approve this pull request!

If @Nilstrieb told you to "r=me" after making some further change, please make that change, then do @bors r=@Nilstrieb

[nebulark]

@bors r=@Nilstrieb

[bors]

📌 Commit 3409ca6 has been approved by Nilstrieb

It is now in the queue for this repository.

[bors]

⌛ Testing commit 3409ca6 with merge 37390d6...

[bors]

☀️ Test successful - checks-actions
Approved by: Nilstrieb
Pushing 37390d6 to master...

[rust-timer]

Finished benchmarking commit (37390d6): comparison URL.

Overall result: no relevant changes - no action needed

@rustbot label: -perf-regression

Instruction count

This benchmark run did not return any relevant results for this metric.

Max RSS (memory usage)

Results

This 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.

	mean	range	count
Regressions ❌ (primary)	2.8%	[0.6%, 4.0%]	4
Regressions ❌ (secondary)	5.1%	[0.9%, 9.0%]	4
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-3.3%	[-8.2%, -1.5%]	9
All ❌✅ (primary)	2.8%	[0.6%, 4.0%]	4

Cycles

This benchmark run did not return any relevant results for this metric.

Binary size

This benchmark run did not return any relevant results for this metric.

Bootstrap: 632.214s -> 631.113s (-0.17%)
Artifact size: 317.15 MiB -> 317.17 MiB (0.01%)