Don't compute FnAbi for LLVM intrinsics by bjorn3 · Pull Request #160077 · rust-lang/rust

bjorn3 · 2026-07-28T14:12:56Z

They don't have a sensible FnAbi, so the fact that we still compute an FnAbi for them requires us to make the ABI sanity check more lenient than it should be.

r? @RalfJung as all non-trivial changes are in Miri

rustbot · 2026-07-28T14:12:59Z

Some changes occurred to the CTFE machinery

cc @RalfJung, @oli-obk, @lcnr

These commits modify compiler targets.
(See the Target Tier Policy.)

Some changes occurred to the CTFE / Miri interpreter

cc @rust-lang/miri

rustbot · 2026-07-28T14:13:02Z

RalfJung is not on the review rotation at the moment.
They may take a while to respond.

bjorn3 · 2026-07-28T14:26:38Z

I guess this finishes my quest of getting rid of PassMode::Direct for BackendRepr::Memory that spanned over 5 years and 13 PRs:

There is still cleanup that can be done, but at least this wildly wrong combination is gone.

bjorn3 · 2026-07-29T10:57:35Z

-    fn test_f32x2(a: f32x2);
-    fn test_f32x2_arr(a: f32x2);
-    fn test_simd(a: Simd<i32, 4>);
-    fn test_simd_unaligned(a: Simd<i32, 3>);


There is no way to test this specific case anymore it seems. LLVM doesn't accept PackedSimd on intrinsics, extern "unadjusted" requires LLVM intrinsics and any other ABI doesn't pass non-power-of-2 vectors as { [3 x i32] }.

View changes since the review

bjorn3 · 2026-07-31T12:21:54Z

Looks like compiler-builtins still has some extern "unadjusted" definitions. Will need to replace those.

rustbot · 2026-07-31T13:33:50Z

compiler-builtins is developed in its own repository. If possible, consider making this change to rust-lang/compiler-builtins instead.

cc @tgross35

RalfJung

interpreter changes mostly LGTM.

compiler-builtins changes will need a review by @tgross35 .

View changes since this review

tgross35 · 2026-08-03T01:57:32Z

-    #[cfg_attr(target_os = "uefi", unadjusted_on_win64)]
+    #[cfg(not(all(target_os = "uefi", target_arch = "x86_64")))]
    pub extern "C" fn __floattisf(i: i128) -> f32 {
        int_to_float::signed(i, int_to_float::u128_to_f32_bits)
    }

-    #[cfg_attr(target_os = "uefi", unadjusted_on_win64)]
+    #[cfg(all(target_os = "uefi", target_arch = "x86_64"))]
+    pub extern "C" fn __floattisf(lo: u64, hi: u64) -> f32 {
+        int_to_float::signed((i128::from(hi) << 64) | i128::from(lo), int_to_float::u128_to_f32_bits)
+    }
+
+    #[cfg(not(all(target_os = "uefi", target_arch = "x86_64")))]
    pub extern "C" fn __floattidf(i: i128) -> f64 {
        int_to_float::signed(i, int_to_float::u128_to_f64_bits)
    }

+    #[cfg(all(target_os = "uefi", target_arch = "x86_64"))]
+    pub extern "C" fn __floattidf(lo: u64, hi: u64) -> f64 {
+        int_to_float::signed((i128::from(hi) << 64) | i128::from(lo), int_to_float::u128_to_f64_bits)
+    }


Do we even need the special casing anymore? i128 has gone through some ABI changes on Windows and it doesn't look like https://github.com/llvm/llvm-project/blob/41322057c3af16d75e239ec6679c6c2bf7aec157/compiler-rt/lib/builtins/floattisf.c#L28 is doing anything special.

View changes since the review

Yes, they still have different ABIs: https://rust.godbolt.org/z/454nczzdr u128 is passed in xmm0 with extern "C", while u64 + u64 is passed in rdx/rcx just like u128 with extern "unadjusted": https://rust.godbolt.org/z/av1f38KrW

Looking at https://rust.godbolt.org/z/fYMc6b8MP I think that Windows is in harmony; always pass indirectly, return in xmm0. I couldn't get it to emit __muloti4.

UEFI seems to be the one that's weird - if I'm reading right it's usually passed indirectly and returned in rax,rdx, but for the __floattisf libcall it seems to be loading into rcx and rdx? I have no idea where this comes from, it doesn't appear to be in https://github.com/llvm/llvm-project/blob/f532c2d780d3afe589b43cb920fbc445322d47f7/compiler-rt/lib/builtins/floattisf.c#L28 so I wonder if it's an oversight in LLVM.

RalfJung · 2026-08-03T14:14:15Z

@rustbot author
Open discussions are: the test, and compiler-builtins.

bjorn3 · 2026-08-04T13:44:06Z

@rustbot ready

RalfJung · 2026-08-04T21:17:40Z

r=me on the Miri part and the test.
r? @tgross35
for the compiler-builtins part

rustbot · 2026-08-04T21:17:45Z

tgross35 is currently at their maximum review capacity.
They may take a while to respond.

tgross35 · 2026-08-13T10:25:31Z

-    #[unadjusted_on_win64]
+    #[cfg(not(all(
+        any(windows, target_os = "cygwin", all(target_os = "uefi", target_arch = "x86_64")),
+        target_pointer_width = "64",
+    )))]
    pub extern "C" fn __muloti4(a: i128, b: i128, oflow: &mut i32) -> i128 {
        let (mul, o) = i128_overflowing_mul(a, b);
        *oflow = o as i32;
        mul
    }

+    #[cfg(all(
+        any(windows, target_os = "cygwin", all(target_os = "uefi", target_arch = "x86_64")),
+        target_pointer_width = "64",
+    ))]
+    pub extern "C" fn f(a_lo: u64, a_hi: u64, b_lo: u64, b_hi: u64, oflow: &mut i32) -> i128 {
+        let (mul, o) = i128_overflowing_mul(
+            (i128::from(a_hi) << 64) | i128::from(a_lo),
+            (i128::from(b_hi) << 64) | i128::from(b_lo),
+        );
+        *oflow = o as i32;
+        mul
+    }


This function is just called f

View changes since the review

I think this can just be dropped; as mentioned in #160077 (comment), I can't get LLVM to emit __muloti4 and I'd expect/hope it to use the usual Windows ABI if it did.

Does cranelift need these on these targets with a different ABI?

I've fixed the function name.

cg_clif used to use __muloti4, but stopped using it about 2 years ago. Would you mind if dropping unnecessary intrinsics is moved to a separate PR. That would make it easier to revert if it turns out to be used after all and avoids scope creep for this PR.

I didn't mean entirely, could you just delete the __muloti4(a_lo: u64, a_hi: u64, b_lo: u64, b_hi: u64, oflow: &mut i32) and leave the existing __muloti4(a: i128, b: i128, oflow: &mut i32)? I just ideally don't want to break tests on Windows when this syncs back.

Or cfg to UEFI-only and match the others

Co-authored-by: Ralf Jung <post@ralfj.de>

rustbot · 2026-08-13T12:25:06Z

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

bjorn3 · 2026-08-13T12:25:42Z

Rebased, fixed test for LLVM 23 and changed the ABI cfg for __muloti4 to just x86_64 UEFI to match other 128bit intrinsics.

tgross35

c-b portion LGTM, so I think this is all set with Ralf's review. Though I still don't know why UEFI needs this special ABI in the first place.

Is the eventual goal to get rid of extern "unadjusted"?

View changes since this review

bjorn3 · 2026-08-13T19:14:48Z

With this PR extern "unadjusted" is effectively extern "llvm-intrinsic" just like we used to have extern "rust-intrinsic". Doing that rename would be a pain though.

@bors r=RalfJung,tgross35

rust-bors · 2026-08-13T19:14:52Z

📌 Commit 0417c4e has been approved by RalfJung,tgross35

It is now in the queue for this repository.

tgross35 · 2026-08-13T19:18:53Z

With this PR extern "unadjusted" is effectively extern "llvm-intrinsic" just like we used to have extern "rust-intrinsic". Doing that rename would be a pain though.

That makes way more sense, I never understood what "unadjusted" was referring to. Perhaps both names should be accepted and stdarch can migrate in pieces?

Rollup of 23 pull requests Successful merges: - #157428 (allocator: refactor for stabilisation) - #158918 (x86_64-win: Enable f128 on LLVM 23+) - #160077 (Don't compute FnAbi for LLVM intrinsics) - #160288 (rustdoc: use anonymous constant for primitives/keywords/attribute docs) - #160440 (Couple of misc improvements to the unwind infrastructure) - #160896 (tidy: Update Python version requirements to 3.11) - #160972 (std: map ERROR_NEGATIVE_SEEK to ErrorKind::InvalidInput on Windows) - #160984 (Fix ICE on { _ } const args in bodies by tracking const-only infer args) - #161036 (Use `with_types_for_suggestion!` in `write_struct_like()`) - #161037 (std: don't clean up the main thread's altstack) - #161040 (Optimize CStr backing slice bounds checks) - #161065 (core: Fix a typo in funnel shift documentation) - #158885 (Add `core::num::Complex`) - #160928 (check `mut`-restriction when tuple constructor is used as a value) - #160969 (give layout errors from `size_of_val` and `align_of_val` a span) - #161002 (disallow `#[cold]` on `extern "custom"` functions) - #161016 (Fix invalid suggestion from try unlabled block) - #161020 (io: Use `NonNull` for all `Custom` API related to `Box`, update documentation for `Custom`) - #161044 (rustc-dev-guide subtree update) - #161046 (Enable unrolling feature of bors) - #161054 (rustdoc: Fix link title attribute value when field of enum variants) - #161055 (`offload!` function-like macro) - #161064 (Revert "riscv: promote d, e, and f target_features to CfgStableToggleUnstable")

@RalfJung

Rollup merge of #160077 - bjorn3:no_unadjusted_fn_abi, r=RalfJung,tgross35 Don't compute FnAbi for LLVM intrinsics They don't have a sensible FnAbi, so the fact that we still compute an FnAbi for them requires us to make the ABI sanity check more lenient than it should be. r? @RalfJung as all non-trivial changes are in Miri

@RalfJung

Rollup merge of #160077 - bjorn3:no_unadjusted_fn_abi, r=RalfJung,tgross35 Don't compute FnAbi for LLVM intrinsics They don't have a sensible FnAbi, so the fact that we still compute an FnAbi for them requires us to make the ABI sanity check more lenient than it should be. r? @RalfJung as all non-trivial changes are in Miri

rust-timer · 2026-08-14T17:12:31Z

Finished benchmarking commit (a96bde1): comparison URL.

Overall result: no relevant changes - no action needed

Benchmarking means the PR may be perf-sensitive. Consider adding rollup=never if this change is not fit for rolling up.

@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

This perf run didn't have relevant results for this metric.

Max RSS (memory usage)

This perf run didn't have relevant results for this metric.

Cycles

Results (primary -3.8%, secondary -1.1%)

A less reliable metric. May be of interest, but not used to determine the overall result above.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	2.0%	[2.0%, 2.0%]	1
Improvements ✅ (primary)	-3.8%	[-3.8%, -3.8%]	1
Improvements ✅ (secondary)	-4.2%	[-4.2%, -4.2%]	1
All ❌✅ (primary)	-3.8%	[-3.8%, -3.8%]	1

Binary size

This perf run didn't have relevant results for this metric.

Bootstrap: 455.615s -> 458.245s (0.58%)
Artifact size: 396.47 MiB -> 396.34 MiB (-0.03%)

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. labels Jul 28, 2026

rustbot assigned RalfJung Jul 28, 2026

This comment has been minimized.

Sign in to view

RalfJung reviewed Jul 28, 2026

View reviewed changes

Comment thread compiler/rustc_const_eval/src/interpret/step.rs Outdated

This comment has been minimized.

Sign in to view

bjorn3 commented Jul 29, 2026

View reviewed changes

bjorn3 force-pushed the no_unadjusted_fn_abi branch from 7358ce7 to eacedff Compare July 29, 2026 10:58

bjorn3 force-pushed the no_unadjusted_fn_abi branch 2 times, most recently from ef2938a to e2873a0 Compare July 31, 2026 12:28

rustbot added the A-compiler-builtins Area: compiler-builtins (https://github.com/rust-lang/compiler-builtins) label Jul 31, 2026

RalfJung reviewed Aug 1, 2026

View reviewed changes

tgross35 reviewed Aug 3, 2026

View reviewed changes

rustbot added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 3, 2026

This comment has been minimized.

Sign in to view

bjorn3 force-pushed the no_unadjusted_fn_abi branch from d2588c6 to a457bee Compare August 4, 2026 13:32

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Aug 4, 2026

rustbot assigned tgross35 and unassigned RalfJung Aug 4, 2026

tgross35 reviewed Aug 13, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

bjorn3 and others added 5 commits August 13, 2026 14:17

Don't use extern "unadjusted" outside of LLVM intrinsics in tests

16dd590

Avoid extern "unadjusted" in compiler-builtins

92e11d1

Avoid computing FnAbi for llvm intrinsics in miri

cc73c4d

Don't compute FnAbi for LLVM intrinsics

12ac752

Add comments

0417c4e

Co-authored-by: Ralf Jung <post@ralfj.de>

bjorn3 force-pushed the no_unadjusted_fn_abi branch from 85cb480 to 0417c4e Compare August 13, 2026 12:25

tgross35 approved these changes Aug 13, 2026

View reviewed changes

rust-bors Bot added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Aug 13, 2026

jhpratt mentioned this pull request Aug 13, 2026

Rollup of 23 pull requests #161075

Merged

rust-bors Bot merged commit a4b0909 into rust-lang:main Aug 14, 2026
13 checks passed

rustbot added this to the 1.99.0 milestone Aug 14, 2026

bjorn3 deleted the no_unadjusted_fn_abi branch August 14, 2026 11:09

Uh oh!

Uh oh!

Conversation

bjorn3 commented Jul 28, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

rustbot commented Jul 28, 2026

Uh oh!

rustbot commented Jul 28, 2026

Uh oh!

bjorn3 commented Jul 28, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

Uh oh!

This comment has been minimized.

bjorn3 Jul 29, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bjorn3 commented Jul 31, 2026

Uh oh!

rustbot commented Jul 31, 2026

Uh oh!

RalfJung left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tgross35 Aug 3, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RalfJung commented Aug 3, 2026

Uh oh!

This comment has been minimized.

bjorn3 commented Aug 4, 2026

Uh oh!

RalfJung commented Aug 4, 2026

Uh oh!

rustbot commented Aug 4, 2026

Uh oh!

tgross35 Aug 13, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

rustbot commented Aug 13, 2026

Uh oh!

bjorn3 commented Aug 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tgross35 left a comment • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bjorn3 commented Aug 13, 2026

Uh oh!

rust-bors Bot commented Aug 13, 2026

Uh oh!

tgross35 commented Aug 13, 2026

Uh oh!

Uh oh!

rust-timer commented Aug 14, 2026

Overall result: no relevant changes - no action needed

Uh oh!

bjorn3 commented Jul 28, 2026 •

edited by rustbot

Loading

bjorn3 Jul 29, 2026 •

edited by rustbot

Loading

RalfJung left a comment •

edited by rustbot

Loading

tgross35 Aug 3, 2026 •

edited by rustbot

Loading

tgross35 Aug 13, 2026 •

edited by rustbot

Loading

bjorn3 commented Aug 13, 2026 •

edited

Loading

tgross35 left a comment •

edited by rustbot

Loading