Support using const pointers in asm `const` operand by nbdd0121 · Pull Request #138618 · rust-lang/rust

nbdd0121 · 2025-03-17T18:07:31Z

Implements RFC#3848 with tracking issue #128464

This adds support of const pointers for asm const in addition to plain integers.

The inline asm! support is implemented using i constraint, and the global_asm! and naked_asm! support is implemented by inserting symbol + offset and make symbol compiler-used. For unnamed consts, it will create additional internal & hidden symbols so that they can be referenced by global_asm.

The feature is also implemented for GCC backend but it's untested.

fmease · 2025-03-17T18:09:34Z

r? codegen

compiler-errors · 2025-03-18T04:25:43Z

r? compiler-errors

rustbot · 2025-04-10T14:59:14Z

Some changes occurred in compiler/rustc_codegen_ssa

cc @WaffleLapkin

Some changes occurred in compiler/rustc_codegen_gcc

cc @antoyo, @GuillaumeGomez

bors · 2025-04-29T06:25:48Z

☔ The latest upstream changes (presumably #140415) made this pull request unmergeable. Please resolve the merge conflicts.

traviscross · 2025-05-04T09:26:42Z

@rustbot author

rustbot · 2025-05-04T09:26:46Z

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

rustbot · 2025-05-13T10:00:16Z

Some changes occurred in compiler/rustc_codegen_cranelift

cc @bjorn3

bors · 2025-05-28T05:09:25Z

☔ The latest upstream changes (presumably #141668) made this pull request unmergeable. Please resolve the merge conflicts.

saethlin · 2026-04-10T00:11:15Z

I just have one question, then r=me.

This was not an easy delta review. More fool me 😂

bjorn3 · 2026-04-10T08:35:54Z

With this PR ThinLTO would no longer be able to rename functions when making them public as the inline asm now directly embeds the symbol names, right?

nbdd0121 · 2026-04-10T10:54:30Z

With this PR ThinLTO would no longer be able to rename functions when making them public as the inline asm now directly embeds the symbol names, right?

Symbol names are handled the same way as today's sym operand. I.e. embedded directly inside global_asm! (with compiler.used added), while for inline asm!() it uses s constraint, which should still allow renaming.

bjorn3 · 2026-04-10T12:13:26Z

+            let alloc = alloc.inner();
+            let typ = self.val_ty(init).get_aligned(alloc.align.bytes());
+
+            let global = self.declare_global_with_linkage(sym.name, typ, GlobalKind::Exported);


View changes since the review

Exporting is wrong. That would lead to symbol conflicts if two asm blocks use the same const.

Hmm, I guess changing the Exported to something else is not going to be sufficient and I need to also do some local dedup in case a asm!() is monomorphized twice due to MIR inlining? This is not an issue as the name hint is only used for global_asm/naked_asm but GCC needs injected symbol names for inline asm too.

I guess I really need to build a working version of cg_gcc and test locally.

Could we have a test that catches this? My understanding is that Rust does run its test suite on gcc too.

Yeah, I'll do that. I've already reproduced the issue locally. It looks like it has been much easier to build Rust with cg_gcc now, which isn't the case when I started working on this in 2024.

bjorn3 · 2026-04-10T12:15:50Z

+        &self,
+        global_alloc: GlobalAlloc<'tcx>,
+        name_hint: Option<Instance<'tcx>>,
+    ) -> Result<(Self::Value, Option<Instance<'tcx>>), u64>;


View changes since the review

This method is not used inside cg_ssa, so it shouldn't be defined there.

It's correlated to scalar_to_backend and I feel this is a good place to put the doc-comments to be shared. I could move it out. Do you want me to duplicate the documentation in cg_gcc and cg_llvm?

This is currently a no-op, but will be useful when const in `global_asm!` can be pointers.

Currently global_asm already have symbol names when using v0 scheme, this makes them obtain symbols with legacy scheme too.

This gives the asm-const code the basic ability to deal wiht pointer and provenances, which lays the ground work for asm_const_ptr. Note that `SymStatic` is not removed, as it supports `#[thread_local]` statics where CTFE does not.

With the previous commit, now we can see there are some code duplication for the handling of `GlobalAlloc` inside backends. Do some clean up to unify them.

CTFE pointers created via type ID, `without_provenance` or pointers to const ZSTs can now be codegenned with all 3 backends. These pointers are generated in the same way as integers.

The backend now fully supports codegen of const pointers, remove the block inside typeck behind a new feature gate. Tests are also added.

rustbot · 2026-04-22T18:12:24Z

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.

saethlin · 2026-04-25T20:49:38Z

r? bjorn3

rust-bors · 2026-04-30T07:42:32Z

☔ The latest upstream changes (presumably #155978) made this pull request unmergeable. Please resolve the merge conflicts.

bjorn3 · 2026-04-30T13:13:14Z

                                "asm! and global_asm! sym operands are not yet supported",
                            );
                        }
-


Please restore this newline.

View changes since the review

bjorn3 · 2026-04-30T13:14:18Z

+                                        // current codegen unit
+                                        tcx.symbol_name(instance)
+                                    }
+                                    GlobalAlloc::Static(def_id) => {


What is the difference between GlobalAsmOperandRef::Const with a static and GlobalAsmOperandRef::SymStatic?

View changes since the review

See #138618 (comment)

#[thread_local] works with sym STATIC today, but it cannot be referenced in CTFE. I would actually be in favour of removing that feature.

I would tend to agree with removing #[thread_local] support for sym STATIC. On some platforms (cough Windows cough) #[thread_local] accesses go through a function shim to ensure cross-dylib accesses work, so sym STATIC would probably already be broken for those platforms.

bjorn3 · 2026-04-30T13:14:59Z

+                                                span,
+                                                "asm! and global_asm! sym operands are not yet supported",
+                                            );
+                                        }


This error should apply to all constants that reference a symbol.

View changes since the review

rustbot assigned fmease Mar 17, 2025

This comment was marked as outdated.

Sign in to view

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 Mar 17, 2025

rustbot assigned workingjubilee and unassigned fmease Mar 17, 2025

This comment has been minimized.

Sign in to view

rustbot assigned compiler-errors and unassigned workingjubilee Mar 18, 2025

nikomatsakis mentioned this pull request Mar 26, 2025

Getting Rust for Linux into stable Rust: language features rust-lang/rust-project-goals#116

Open

53 tasks

ojeda mentioned this pull request Apr 9, 2025

Rust unstable features needed for the kernel Rust-for-Linux/linux#2

Open

52 tasks

nbdd0121 force-pushed the asm_const_ptr branch from ffba8e8 to 92c0cc4 Compare April 10, 2025 10:05

nbdd0121 marked this pull request as ready for review April 10, 2025 14:59

tmiasko reviewed Apr 10, 2025

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/collector.rs

compiler-errors reviewed Apr 13, 2025

View reviewed changes

Comment thread compiler/rustc_codegen_ssa/src/traits/asm.rs Outdated

Comment thread compiler/rustc_codegen_ssa/src/traits/asm.rs Outdated

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 May 4, 2025

nbdd0121 force-pushed the asm_const_ptr branch from 92c0cc4 to c7a2301 Compare May 13, 2025 08:27

This comment has been minimized.

Sign in to view

nbdd0121 force-pushed the asm_const_ptr branch from c7a2301 to 4240004 Compare May 13, 2025 10:00

workingjubilee reviewed Jun 4, 2025

View reviewed changes

Comment thread tests/ui/asm/const-refs-to-static.rs

Comment thread tests/assembly-llvm/asm/x86-types.rs Outdated

oli-obk reviewed Jul 9, 2025

View reviewed changes

Comment thread compiler/rustc_codegen_llvm/src/asm.rs Outdated

bjorn3 reviewed Apr 10, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_gcc/src/common.rs

bjorn3 reviewed Apr 10, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_llvm/src/common.rs

nbdd0121 force-pushed the asm_const_ptr branch from f6a193d to 6e3cca5 Compare April 10, 2026 13:01

This comment has been minimized.

Sign in to view

nbdd0121 added 10 commits April 22, 2026 17:48

Collect constants within global_asm! in collector

086e52e

This is currently a no-op, but will be useful when const in `global_asm!` can be pointers.

Add FnDef/Closure -> FnPtr coercion for inline asm const operand

8337efc

Give global_asm symbol names

e7467b6

Currently global_asm already have symbol names when using v0 scheme, this makes them obtain symbols with legacy scheme too.

Delay stringification of const to backend

b60bb73

Unify handling of asm const and sym fn using CTFE

b5bfc59

This gives the asm-const code the basic ability to deal wiht pointer and provenances, which lays the ground work for asm_const_ptr. Note that `SymStatic` is not removed, as it supports `#[thread_local]` statics where CTFE does not.

Unify handling of GlobalAlloc inside backend

9822024

With the previous commit, now we can see there are some code duplication for the handling of `GlobalAlloc` inside backends. Do some clean up to unify them.

Handle pointers with offset for asm const

5916d6e

Support codegen of asm const pointers without provenance

d6b2782

CTFE pointers created via type ID, `without_provenance` or pointers to const ZSTs can now be codegenned with all 3 backends. These pointers are generated in the same way as integers.

Generate unique symbol names if const pointers refer to promoted static

0846b11

Implement asm_const_ptr feature

f4011e8

The backend now fully supports codegen of const pointers, remove the block inside typeck behind a new feature gate. Tests are also added.

nbdd0121 force-pushed the asm_const_ptr branch from 6e3cca5 to f4011e8 Compare April 22, 2026 18:12

Add test if asm const are MIR inlined and monomorphized twice

7c877c9

rustbot assigned bjorn3 and unassigned saethlin Apr 25, 2026

bjorn3 requested changes Apr 30, 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 Apr 30, 2026

Uh oh!

Conversation

nbdd0121 commented Mar 17, 2025 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment was marked as outdated.

fmease commented Mar 17, 2025

Uh oh!

This comment has been minimized.

compiler-errors commented Mar 18, 2025

Uh oh!

rustbot commented Apr 10, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

bors commented Apr 29, 2025

Uh oh!

traviscross commented May 4, 2025

Uh oh!

rustbot commented May 4, 2025

Uh oh!

This comment has been minimized.

rustbot commented May 13, 2025

Uh oh!

bors commented May 28, 2025

Uh oh!

Uh oh!

Uh oh!

Uh oh!

saethlin commented Apr 10, 2026

Uh oh!

bjorn3 commented Apr 10, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nbdd0121 commented Apr 10, 2026

Uh oh!

bjorn3 Apr 10, 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!

nbdd0121 Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Darksonn Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

nbdd0121 Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

nbdd0121 Apr 22, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

bjorn3 Apr 10, 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!

nbdd0121 Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

This comment has been minimized.

This comment has been minimized.

rustbot commented Apr 22, 2026

Uh oh!

saethlin commented Apr 25, 2026

Uh oh!

rust-bors Bot commented Apr 30, 2026

Uh oh!

bjorn3 Apr 30, 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 Apr 30, 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!

nbdd0121 Apr 30, 2026

Choose a reason for hiding this comment

nbdd0121 commented Mar 17, 2025 •

edited by rustbot

Loading

bjorn3 commented Apr 10, 2026 •

edited

Loading

bjorn3 Apr 10, 2026 •

edited by rustbot

Loading

bjorn3 Apr 10, 2026 •

edited by rustbot

Loading

bjorn3 Apr 30, 2026 •

edited by rustbot

Loading

bjorn3 Apr 30, 2026 •

edited by rustbot

Loading

bjorn3 Apr 30, 2026 •

edited by rustbot

Loading