Update LLVM submodule

[icedrocket]

Fixes #105626

[rustbot]

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @cuviper (or someone else) soon.

Please see the contribution instructions for more information.

[rustbot]

⚠️ Warning ⚠️

• These commits modify submodules.

[cuviper]

Can you add a regression test for this? Probably a ui run-pass test with min-system-llvm-version: 16.

[icedrocket]

rust/tests/ui/numbers-arithmetic/saturating-float-casts-impl.rs

Lines 350 to 380 in eebdfb5

	// int -> f32
	assert_eq::<f32>(127i8 as f32, 127.0);
	assert_eq::<f32>(2147483647i32 as f32, 2147483648.0);
	assert_eq::<f32>((-2147483648i32) as f32, -2147483648.0);
	assert_eq::<f32>(1234567890i32 as f32, /*0x1.26580cp+30*/ f32::from_bits(0x4e932c06));
	assert_eq::<f32>(16777217i32 as f32, 16777216.0);
	assert_eq::<f32>((-16777217i32) as f32, -16777216.0);
	assert_eq::<f32>(16777219i32 as f32, 16777220.0);
	assert_eq::<f32>((-16777219i32) as f32, -16777220.0);
	assert_eq::<f32>(
	0x7fffff4000000001i64 as f32,
	/*0x1.fffffep+62*/ f32::from_bits(0x5effffff),
	);
	assert_eq::<f32>(
	0x8000004000000001u64 as i64 as f32,
	/*-0x1.fffffep+62*/ f32::from_bits(0xdeffffff),
	);
	assert_eq::<f32>(
	0x0020000020000001i64 as f32,
	/*0x1.000002p+53*/ f32::from_bits(0x5a000001),
	);
	assert_eq::<f32>(
	0xffdfffffdfffffffu64 as i64 as f32,
	/*-0x1.000002p+53*/ f32::from_bits(0xda000001),
	);
	// FIXME emscripten does not support i128
	#[cfg(not(target_os = "emscripten"))]
	{
	assert_eq::<f32>(i128::MIN as f32, -170141183460469231731687303715884105728.0f32);
	assert_eq::<f32>(u128::MAX as f32, f32::INFINITY); // saturation
	}

It seems that this part doesn't check unsigned to float casting for edge cases. We need to add unsigned to float casting checks and modify assert_eq to check both static and runtime casting.

[cuviper]

It seems that this part doesn't check unsigned to float casting for edge cases. We need to add unsigned to float casting checks and modify assert_eq to check both static and runtime casting.

I think that's fine to add separately, now that there's a targeted test here.

@bors r+

[bors]

📌 Commit 1c97cde6f11dab37312faef4a2460339115ae3c8 has been approved by cuviper

It is now in the queue for this repository.

[bors]

⌛ Testing commit 1c97cde6f11dab37312faef4a2460339115ae3c8 with merge 97abed1fd6352f808f6309387c82f36fcc9959e3...

[bors]

💔 Test failed - checks-actions

[cuviper]

Those values failed equality on i586-unknown-linux-gnu, but print identically -- that might be more x87 excess precision at work on i586. Maybe we should just limit this test to only-i686-pc-windows-msvc?

[cuviper]

@bors r+

[bors]

📌 Commit 1e4ba39b026dff0378b3612d12da1a9f2e02f128 has been approved by cuviper

It is now in the queue for this repository.

[icedrocket]

I force-pushed to parent commit by accident. Now ready for review.

[cuviper]

I don't think black_box is a reliable way to avoid x87 issues. What was wrong with your sse2 version of the test? This regression test is specifically for a particular target platform; it's ok if we don't run that test elsewhere.

[icedrocket]

I don't think black_box is a reliable way to avoid x87 issues. What was wrong with your sse2 version of the test? This regression test is specifically for a particular target platform; it's ok if we don't run that test elsewhere.

The u64 to f32 conversion is currently implemented using x87 even when SSE2 is enabled. If in the future the compiler changes to use SSE2 for that conversion, it will be difficult to know if there is a problem with the x87 implementation as the test will not fail as long as SSE2 is enabled.

[icedrocket]

But I think it's okay for us to just use the SSE2 version for now. As long as SSE2 is enabled, we can generate code that works fine on x86. I will contribute to solving the x87 issue on the LLVM side later.

[cuviper]

The LLVM fix is specific to Windows, and i686-pc-windows-msvc does have SSE2 enabled by default. I'm not sure what else you're hoping to fix in LLVM, but I think we're good here.

@bors r+

[bors]

📌 Commit 313f04f has been approved by cuviper

It is now in the queue for this repository.

[bors]

⌛ Testing commit 313f04f with merge 2f5bc1f0b395c73a727be74e7f1b277a36a386c3...

[bors]

💔 Test failed - checks-actions

[cuviper]

(x86_64-gnu-llvm-14, 1, ubuntu-20.04-xl):

---- config::tests::download_ci_llvm stdout ----
Detected LLVM as non-available: running in CI and modified LLVM in this change
Detected LLVM as non-available: running in CI and modified LLVM in this change
Detected LLVM as non-available: running in CI and modified LLVM in this change
Detected LLVM as non-available: running in CI and modified LLVM in this change
thread 'config::tests::download_ci_llvm' panicked at 'assertion failed: parse_llvm(\"build.build = \\\"x86_64-unknown-linux-gnu\\\"\")', config/tests.rs:22:5
error: test failed, to rerun pass `--lib`

@rust-lang/bootstrap -- this should be expected! We're updating LLVM, but not all CI builders use that.
(though I'm not certain I understand the failure mode here...)

[jyn514]

@cuviper oops, this might be the first time LLVM's been updated since #107234 - I think we should skip this test altogether if LLVM was updated in the PR. Maybe take the code in

rust/src/bootstrap/native.rs

Lines 219 to 230 in ae5473c

	if CiEnv::is_ci() {
	// We assume we have access to git, so it's okay to unconditionally pass
	// `true` here.
	let llvm_sha = detect_llvm_sha(config, true);
	let head_sha = output(config.git().arg("rev-parse").arg("HEAD"));
	let head_sha = head_sha.trim();
	if llvm_sha == head_sha {
	eprintln!(
	"Detected LLVM as non-available: running in CI and modified LLVM in this change"
	);
	return false;
	}

and separate it into a function?

[cuviper]

@jyn514 I gave that a shot, let me know what you think!

[albertlarsan68]

I approve the last commit (91af58ac58dee5794fe54a8f9cb5556d3585688d).
Just a minor nit, if you want.

[cuviper]

@bors r+

[bors]

📌 Commit 565de58 has been approved by cuviper

It is now in the queue for this repository.

[bors]

⌛ Testing commit 565de58 with merge 18caf88...

[bors]

☀️ Test successful - checks-actions
Approved by: cuviper
Pushing 18caf88 to master...

[rust-timer]

Finished benchmarking commit (18caf88): 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)

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

Cycles

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