Add support for generics in `offload`and remove `no_mangle` attribute by Sa4dUs · Pull Request #159566 · rust-lang/rust

Sa4dUs · 2026-07-19T14:04:27Z

implements a 3rd pass where we collect kernel instantiations required by offload to work
also enforces consistent mangling for offload functions between host and device

closes: #150985

r? @ZuseZ4

bjorn3 · 2026-07-19T21:25:21Z

    // core/std/allocators/etc. For example symbols used to hook up allocation
    // are not considered for export
    let codegen_fn_attrs = tcx.codegen_fn_attrs(sym_def_id);
    let is_extern = codegen_fn_attrs.contains_extern_indicator();


Perhaps contains_extern_indicator should return true for offload functions?

View changes since the review

ZuseZ4 · 2026-07-20T05:11:02Z

Great improvement ( I'll review later).
@Flakebi fyi, I think now we only have the abi checks left to implement for more complex types and offload should cover all the bigger things we talked about?

rustbot · 2026-07-24T15:07:47Z

These commits modify the Cargo.lock file. Unintentional changes to Cargo.lock can be introduced when switching branches and rebasing PRs.

If this was unintentional then you should revert the changes before this PR is merged.
Otherwise, you can ignore this comment.

Flakebi · 2026-07-28T12:22:53Z

Very nice!
If I read this correctly, there are two important parts here:

Not adding the hash suffix when mangling offload kernel names makes the mangling deterministic
Compilation first compiles the CPU code, which generates a new artifact – a list of function instances of kernels that are referenced. When compiling the GPU code afterwards, this list is added to the monomorphization roots.

On a sidenote: As due to 1., no more hash suffix is added to kernels, we should ensure that we fail compilation if someone tries to compile a program that defines the same mangled kernel name twice.
E.g. by importing two different versions of the same crate that defines the same kernel.
This might already fail to compile but a test for that would be good.

ZuseZ4

Just some smaller questions, at the end Oli should probably have another look. But other than cleaning up the old Host, can you add some actual tests for the generic functions?

View changes since this review

ZuseZ4 · 2026-07-28T13:03:51Z

+        .iter()
+        .any(|o| matches!(o, rustc_session::config::Offload::HostMetadata(_)));
+    if is_host_metadata {
+        let crate_items = tcx.hir_crate_items(());


What about trait_items? The impl_items note says it also covers traits, but he owners function on ModuleItems still collects both so they might not be exactly the same?

rustbot · 2026-08-05T17:49:25Z

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.

ZuseZ4 · 2026-08-06T20:21:43Z

+        .emit("obj")
+        .run();
+
+    assert!(object_contains_any_symbol_substring("generic.o", &["6kernelfEB2_"]));


Can you explicitely set the mangling version for that? We just moved to v0 so I don't think it will change soon, but just in case.

View changes since the review

ZuseZ4 · 2026-08-06T20:22:30Z

+fn kernel(_x: f32) {} //~ ERROR symbol `_RNvC19collision_kernels_a6kernel` is already defined
+
+fn main() {
+    _RNvC19collision_kernels_a6kernel(0.0);


same here, otherwise this test silently becomes useless in the (tbh probably far) future.

View changes since the review

ZuseZ4 · 2026-08-06T20:44:03Z

+    /// produced by a previous `HostMetadata` pass. An empty manifest
+    /// means all kernel instantiations are discovered via monomorphization.
+    Device(String),
    /// Second step in the offload pipeline, generates the host code to call kernels.


That's the third step now, right?
And maybe describe Device(String) above as the second step.

View changes since the review

ZuseZ4 · 2026-08-06T20:47:19Z

-    Device,
+    /// Reads a manifest of required generic kernel instantiations
+    /// produced by a previous `HostMetadata` pass. An empty manifest
+    /// means all kernel instantiations are discovered via monomorphization.


The last sentence is imho a little too complicated for people not aware of how gpu compilation works. Can you clarify that it means there are either no generic kernels at all, or that generic functions are only called from non-generic gpu entry points, but never from the Host (cpu), so we don't need to track it?

View changes since the review

ZuseZ4 · 2026-08-06T20:48:30Z

                    }
-                    Offload::Device
+                }
+                "Device" => {


This seems risky. Do we throw a good error if there's a generic kernel but a user passes Device without a path? I'd expect he only get's a cryptic error, or?

View changes since the review

ZuseZ4 · 2026-08-06T20:49:35Z

+/// (which differs between host and device passes).
+pub fn mangle_offload_export<'tcx>(tcx: TyCtxt<'tcx>, instance: Instance<'tcx>) -> String {
+    let instantiating_crate = if is_generic(instance) {
+        Some(instance.upstream_monomorphization(tcx).unwrap_or(LOCAL_CRATE))


While we might compile a lot of code for the gpu, few functions will be entry points, so I guess it will be ok-ish. Otherwise we can think of smarter mangling in the future.

View changes since the review

ZuseZ4 · 2026-08-06T20:52:20Z

Lgtm in general and from the offload side, but let's ping @oli-obk since we talked about it at RustWeek. Is that roughly what you had in mind, or do you see any general issues?

ZuseZ4 · 2026-08-06T20:56:51Z

@bors try @rust-timer queue

Add support for generics in `offload`and remove `no_mangle` attribute

rust-bors · 2026-08-06T22:26:34Z

☀️ Try build successful (CI)
Build commit: 427b2a9 (427b2a96dd79c4d6ce4aaca78ba5e82fd45d2b37)
Base parent: 84b36a7 (84b36a78a28a63f134171c670be1932ffa2485f8)

rust-timer · 2026-08-06T23:33:16Z

Finished benchmarking commit (427b2a9): comparison URL.

Overall result: ✅ improvements - no action needed

Benchmarking means the PR may be perf-sensitive. It's automatically marked not fit for rolling up. Overriding is possible but disadvised: it risks changing compiler perf.

@bors rollup=never rustc-perf
@rustbot label: -S-waiting-on-perf -perf-regression

Instruction count

Our most reliable metric. Used to determine the overall result above. However, even this metric can be noisy.

	mean	range	count
Regressions ❌ (primary)	-	-	0
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-0.4%	[-0.4%, -0.4%]	1
Improvements ✅ (secondary)	-0.4%	[-0.5%, -0.2%]	7
All ❌✅ (primary)	-0.4%	[-0.4%, -0.4%]	1

Max RSS (memory usage)

Results (primary -5.5%)

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)	-	-	0
Improvements ✅ (primary)	-5.5%	[-5.5%, -5.5%]	1
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-5.5%	[-5.5%, -5.5%]	1

Cycles

Results (primary 3.0%, secondary -7.5%)

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

	mean	range	count
Regressions ❌ (primary)	3.0%	[3.0%, 3.0%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-	-	0
Improvements ✅ (secondary)	-7.5%	[-12.9%, -2.0%]	2
All ❌✅ (primary)	3.0%	[3.0%, 3.0%]	1

Binary size

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

Bootstrap: 457.798s -> 460.645s (0.62%)
Artifact size: 399.35 MiB -> 398.64 MiB (-0.18%)

rustbot assigned ZuseZ4 Jul 19, 2026

Sa4dUs force-pushed the offload-generics branch from 7fefbf7 to a019aea Compare July 19, 2026 14:10

This comment has been minimized.

Sign in to view

bjorn3 reviewed Jul 19, 2026

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/collector.rs Outdated

bjorn3 reviewed Jul 19, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_ssa/src/base.rs Outdated

This comment has been minimized.

Sign in to view

bjorn3 reviewed Jul 23, 2026

View reviewed changes

Comment thread compiler/rustc_codegen_ssa/src/back/write.rs Outdated

Sa4dUs force-pushed the offload-generics branch from b40ecd1 to cb87656 Compare July 23, 2026 16:44

This comment has been minimized.

Sign in to view

Sa4dUs force-pushed the offload-generics branch from cb87656 to f684946 Compare July 24, 2026 13:07

This comment has been minimized.

Sign in to view

Sa4dUs marked this pull request as ready for review July 24, 2026 15:07

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 Jul 24, 2026

bjorn3 reviewed Jul 24, 2026

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/collector.rs

bjorn3 reviewed Jul 24, 2026

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/offload_manifest.rs

bjorn3 reviewed Jul 24, 2026

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/offload_manifest.rs Outdated

bjorn3 reviewed Jul 24, 2026

View reviewed changes

Comment thread compiler/rustc_monomorphize/src/offload_manifest.rs Outdated

bjorn3 self-requested a review July 24, 2026 20:49

ZuseZ4 reviewed Jul 28, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

Sa4dUs added 2 commits August 5, 2026 20:47

Add 3-pass compilation to support generics and remove no_mangle attr

9637d18

Remove codegen and move manifest out of query

57b115b

Sa4dUs added 3 commits August 5, 2026 20:47

fix

15ec03c

ci fix

7d1f41a

minor fixes

68db0ed

Sa4dUs force-pushed the offload-generics branch from 2c1b875 to 7860122 Compare August 5, 2026 17:49

rustbot added the A-run-make Area: port run-make Makefiles to rmake.rs label Aug 5, 2026

This comment has been minimized.

Sign in to view

Add tests remove extra mode and other fixes

3686d5c

Sa4dUs force-pushed the offload-generics branch from 7860122 to 3686d5c Compare August 5, 2026 18:43

ZuseZ4 reviewed Aug 6, 2026

View reviewed changes

This comment has been minimized.

Sign in to view

rustbot added the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 6, 2026

This comment has been minimized.

Sign in to view

rust-bors Bot pushed a commit that referenced this pull request Aug 6, 2026

Auto merge of #159566 - Sa4dUs:offload-generics, r=<try>

427b2a9

Add support for generics in `offload`and remove `no_mangle` attribute

This comment has been minimized.

Sign in to view

rustbot removed the S-waiting-on-perf Status: Waiting on a perf run to be completed. label Aug 6, 2026

Uh oh!

Uh oh!

Conversation

Sa4dUs commented Jul 19, 2026 • edited by rustbot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

This comment has been minimized.

bjorn3 Jul 19, 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!

Uh oh!

Uh oh!

ZuseZ4 commented Jul 20, 2026

Uh oh!

This comment has been minimized.

Uh oh!

This comment has been minimized.

This comment has been minimized.

rustbot commented Jul 24, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Flakebi commented Jul 28, 2026

Uh oh!

ZuseZ4 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!

ZuseZ4 Jul 28, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

This comment has been minimized.

rustbot commented Aug 5, 2026

Uh oh!

This comment has been minimized.

ZuseZ4 Aug 6, 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!

ZuseZ4 Aug 6, 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!

ZuseZ4 Aug 6, 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!

ZuseZ4 Aug 6, 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!

ZuseZ4 Aug 6, 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!

ZuseZ4 Aug 6, 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!

ZuseZ4 commented Aug 6, 2026

Uh oh!

ZuseZ4 commented Aug 6, 2026

Uh oh!

This comment has been minimized.

This comment has been minimized.

rust-bors Bot commented Aug 6, 2026

Uh oh!

This comment has been minimized.

rust-timer commented Aug 6, 2026

Overall result: ✅ improvements - no action needed

Instruction count

Max RSS (memory usage)

Cycles

Binary size

Uh oh!

Sa4dUs commented Jul 19, 2026 •

edited by rustbot

Loading

bjorn3 Jul 19, 2026 •

edited by rustbot

Loading

ZuseZ4 left a comment •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading

ZuseZ4 Aug 6, 2026 •

edited by rustbot

Loading