Add regression test to ensure optimal compilation

[KevinA-cpu]

closes #81342

[rustbot]

r? @khyperia

rustbot has assigned @khyperia.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 75 candidates
• Random selection from 21 candidates

[khyperia]

poke me with @rustbot review when the test passes and I'll take a look! - no worries that this is still a WIP, just doing a rustbot-author so the automated tooling doesn't have this on my review todo plate 😅

@rustbot author

[rustbot]

Reminder, once the PR becomes ready for a review, use @rustbot ready.

[KevinA-cpu]

@rustbot review

Thank you, @khyperia, for your patience. As I was looking at the part where the pipeline was failing, I noticed a check for building with wasm32 that uses i32 instead of i64 in the generated LLVM. Is it possible for you to share a bit of knowledge on this part about why we have checks like these and whether is it needed to add tests for 32 bit as well?

[khyperia]

Is it possible for you to share a bit of knowledge on this part about why we have checks like these

Hmm, I'm not particularly sure what you're asking. It is true that usize is not necessarily 64 bits, there are many platforms whose pointer size and whatnot is not 64 bits. Testing these platforms is important! :3

Looking at other codegen-llvm tests that deal with usize, looks like most of them have a glob for an arbitrary integer size, e.g. here:

rust/tests/codegen-llvm/transmute-scalar.rs

Lines 50 to 56 in 9bb55c8

	// CHECK: define{{.*}}[[USIZE:i[0-9]+]] @ptr_to_int(ptr %p)
	// CHECK: %_0 = ptrtoint ptr %p to [[USIZE]]
	// CHECK-NEXT: ret [[USIZE]] %_0
	#[no_mangle]
	pub fn ptr_to_int(p: *mut u16) -> usize {
	unsafe { mem::transmute(p) }
	}

or this one that uses a cute little helper to fetch what usize is and then uses it ([[USIZE]]) throughout the test:

rust/tests/codegen-llvm/atomicptr.rs

Lines 14 to 17 in 9bb55c8

	// Portability hack so that we can say [[USIZE]] instead of i64/i32/i16 for usize.
	// CHECK: @helper([[USIZE:i[0-9]+]] noundef %_1)
	#[no_mangle]
	pub fn helper(_: usize) {}

in your case it's probably fine to just use i{{[0-9]+}} (without //@ only-64bit) as the size of usize isn't particularly relevant to this test.

[KevinA-cpu]

Hi @khyperia

It is true that usize is not necessarily 64 bits, there are many platforms whose pointer size and whatnot is not 64 bits. Testing these platforms is important! :3

Was exactly what I wanted to ask. Thank you for the suggestion, I'll make sure to test other platforms as well in the future. PR is updated.

[khyperia]

neat, thanks!

@bors r+ rollup

[rust-bors]

📌 Commit 4a96536 has been approved by khyperia

It is now in the queue for this repository.