Explicitly choose x86 softfloat/hardfloat ABI #136146

RalfJung · 2025-01-27T17:23:03Z

Part of #135408:
Instead of choosing this based on the target features listed in the target spec, make that choice explicit.
All built-in targets are being updated here; custom (JSON-defined) x86 (32bit and 64bit) softfloat targets need to explicitly set rustc-abi to x86-softfloat.

rustbot · 2025-01-27T17:23:11Z

r? @cjgillot

rustbot has assigned @cjgillot.
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

rustbot · 2025-01-27T17:23:13Z

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

RalfJung · 2025-01-27T17:30:51Z

Cc @workingjubilee

cjgillot · 2025-01-28T00:01:51Z

r? compiler

RalfJung · 2025-01-28T01:41:41Z

compiler/rustc_target/src/target_features.rs

-    ("soft-float", Stability::Forbidden { reason: "unsound because it changes float ABI" }, &[]),
+    // This cannot actually be toggled, the ABI always fixes it, so it'd make little sense to
+    // stabilize. It must be in this list for the ABI check to be able to use it.
+    ("soft-float", Stability::Unstable(sym::x87_target_feature), &[]),


Making this "forbidden" just leads to duplicate warnings so there's little point in doing that.

This makes riscv's forced-atomics the only remaining "forbidden" feature.

That makes sense to me.

workingjubilee

My understanding is that while it changes some error outputs, this PR shouldn't have any practical effect on people, correct? Except those using our unstable target-spec.json, I suppose, and they bought the ticket for the ride they're going to get.

r=me with nits addressed

workingjubilee · 2025-01-28T02:47:41Z

compiler/rustc_target/src/spec/json.rs

+                    match s.parse::<super::RustcAbi>() {
+                        Ok(rustc_abi) => base.$key_name = Some(rustc_abi),
+                        _ => return Some(Err(format!(
+                            "'{s}' is not a valid value for rust-abi. \


Suggested change

"'{s}' is not a valid value for rust-abi. \

"'{s}' is not a valid value for rustc-abi. \

workingjubilee · 2025-01-28T02:48:53Z

compiler/rustc_target/src/spec/mod.rs

+/// The Rustc-specific variant of the ABI used for this target.
+#[derive(Clone, Copy, PartialEq, Hash, Debug)]
+pub enum RustcAbi {
+    /// On x86-32/64 only: do not use any FPU or SIMD registers for the ABI/


the "ABI/"?

RalfJung · 2025-01-28T02:59:38Z

My understanding is that while it changes some error outputs, this PR shouldn't have any practical effect on people, correct? Except those using our unstable target-spec.json, I suppose, and they bought the ticket for the ride they're going to get.

Correct, for custom target specs this can add -soft-float if you don't also set rustc-abi: "x86-softfloat". This is probably surprising as there is no warning about this, it just silently happens.

Maybe we should land #136147 first; that would avoid changing behavior for target-spec users and instead just emit a warning.

workingjubilee · 2025-01-28T03:02:20Z

I'm not worried about being maximally convenient for them, but either way. I'll try to review that when I finish my tea.

lcnr · 2025-01-28T10:33:13Z

r? @workingjubilee

rustbot · 2025-01-28T10:33:16Z

Could not assign reviewer from: workingjubilee.
User(s) workingjubilee are either the PR author, already assigned, or on vacation. Please use r? to specify someone else to assign.

bors · 2025-01-29T08:44:17Z

☔ The latest upstream changes (presumably #136225) made this pull request unmergeable. Please resolve the merge conflicts.

RalfJung · 2025-01-29T10:12:36Z

@bors r=workingjubilee

…and use it for x86 softfloat

RalfJung · 2025-02-03T15:58:53Z

I have pulled the docs change into a separate PR: #136493.

@bors r=workingjubilee

bors · 2025-02-03T15:58:55Z

📌 Commit 8596ce1 has been approved by workingjubilee

It is now in the queue for this repository.

RalfJung · 2025-02-03T16:01:20Z

@jackpot51 heads-up that Redox will have to explicitly select the x86 softfloat ABI once this lands, similar to what was needed for Linux.

bors · 2025-02-03T20:02:58Z

⌛ Testing commit 8596ce1 with merge f027438...

bors · 2025-02-03T22:50:28Z

☀️ Test successful - checks-actions
Approved by: workingjubilee
Pushing f027438 to master...

rust-timer · 2025-02-04T00:24:58Z

Finished benchmarking commit (f027438): 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)

Results (primary -0.8%)

This is a less reliable metric that may be of interest but was not used to determine the overall result at the top of this comment.

	mean	range	count
Regressions ❌ (primary)	3.5%	[3.5%, 3.5%]	1
Regressions ❌ (secondary)	-	-	0
Improvements ✅ (primary)	-3.0%	[-4.5%, -1.4%]	2
Improvements ✅ (secondary)	-	-	0
All ❌✅ (primary)	-0.8%	[-4.5%, 3.5%]	3

Cycles

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

Binary size

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

Bootstrap: 779.425s -> 778.951s (-0.06%)
Artifact size: 328.79 MiB -> 328.87 MiB (0.03%)

Required to enable the soft-float feature. See rust-lang/rust#136146

When using Rust on the x86 architecture, we are currently using the unstable target.json feature to specify the compilation target. Rustc is going to change how softfloat is specified in the target.json file on x86, thus update generate_rust_target.rs to specify softfloat using the new option. Note that if you enable this parameter with a compiler that does not recognize it, then that triggers a warning but it does not break the build. Cc: <[email protected]> # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs). Link: rust-lang/rust#136146 Signed-off-by: Alice Ryhl <[email protected]> Link: https://lore.kernel.org/r/[email protected] [ Added 6.13.y too to Cc: stable tag and added reasoning to avoid over-backporting. - Miguel ] Signed-off-by: Miguel Ojeda <[email protected]>

When using Rust on the x86 architecture, we are currently using the unstable target.json feature to specify the compilation target. Rustc is going to change how softfloat is specified in the target.json file on x86, thus update generate_rust_target.rs to specify softfloat using the new option. Note that if you enable this parameter with a compiler that does not recognize it, then that triggers a warning but it does not break the build. [ For future reference, this solves the following error: RUSTC L rust/core.o error: Error loading target specification: target feature `soft-float` is incompatible with the ABI but gets enabled in target spec. Run `rustc --print target-list` for a list of built-in targets - Miguel ] Cc: <[email protected]> # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs). Link: rust-lang/rust#136146 Signed-off-by: Alice Ryhl <[email protected]> Link: https://lore.kernel.org/r/[email protected] [ Added 6.13.y too to Cc: stable tag and added reasoning to avoid over-backporting. - Miguel ] Signed-off-by: Miguel Ojeda <[email protected]>

When using Rust on the x86 architecture, we are currently using the unstable target.json feature to specify the compilation target. Rustc is going to change how softfloat is specified in the target.json file on x86, thus update generate_rust_target.rs to specify softfloat using the new option. Note that if you enable this parameter with a compiler that does not recognize it, then that triggers a warning but it does not break the build. [ For future reference, this solves the following error: RUSTC L rust/core.o error: Error loading target specification: target feature `soft-float` is incompatible with the ABI but gets enabled in target spec. Run `rustc --print target-list` for a list of built-in targets - Miguel ] Cc: <[email protected]> # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs). Link: rust-lang/rust#136146 Signed-off-by: Alice Ryhl <[email protected]> Acked-by: Dave Hansen <[email protected]> # for x86 Link: https://lore.kernel.org/r/[email protected] [ Added 6.13.y too to Cc: stable tag and added reasoning to avoid over-backporting. - Miguel ] Signed-off-by: Miguel Ojeda <[email protected]>

riking · 2025-02-08T23:50:01Z

src/ci/docker/scripts/rfl-build.sh

@@ -28,7 +28,7 @@ rm -rf linux || true
 # Download Linux at a specific commit
 mkdir -p linux
 git -C linux init
-git -C linux remote add origin https://github.com/Rust-for-Linux/linux.git
+git -C linux remote add origin https://github.com/Darksonn/linux.git


this needs a cfg(bootstrap) to get bumped on release

I don't follow, what does the Linux kernel have to do with bootstraping?

I think I did everything mentioned at https://rustc-dev-guide.rust-lang.org/tests/rust-for-linux.html.

Oops, I missed that the maintainers have taken responsibility for flipping this back and don't need extra reminders from Release.

I don't follow, what does the Linux kernel have to do with bootstraping?

I think I did everything mentioned at https://rustc-dev-guide.rust-lang.org/tests/rust-for-linux.html.

I am confused too -- this seems fine.

When v6.14-rc3 releases in a couple weeks, which is the one I am planning to get the fix into, I will send a PR with that tag, and we will be in the normal situation again.

Oops, I missed that the maintainers have taken responsibility for flipping this back and don't need extra reminders from Release.

I don't think there is a technical requirement to flip it back. I mean, the idea is that we do so, but it is orthogonal to the release, no? It is just a CI test. For instance, in an ideal world, we could be testing several supported Linux versions (e.g. the latest LTS).

This additional field is required by the Rust compiler since rust-lang/rust#136146 .

The Rust compiler now requires an explicit `rustc-abi: x86-softfloat` field when using the soft-float target feature. This was added in rust-lang/rust#136146 and is part of the latest nightlies. I updated the post branches in commit 688a21e

commit 6273a05 upstream. When using Rust on the x86 architecture, we are currently using the unstable target.json feature to specify the compilation target. Rustc is going to change how softfloat is specified in the target.json file on x86, thus update generate_rust_target.rs to specify softfloat using the new option. Note that if you enable this parameter with a compiler that does not recognize it, then that triggers a warning but it does not break the build. [ For future reference, this solves the following error: RUSTC L rust/core.o error: Error loading target specification: target feature `soft-float` is incompatible with the ABI but gets enabled in target spec. Run `rustc --print target-list` for a list of built-in targets - Miguel ] Cc: <[email protected]> # Needed in 6.12.y and 6.13.y only (Rust is pinned in older LTSs). Link: rust-lang/rust#136146 Signed-off-by: Alice Ryhl <[email protected]> Acked-by: Dave Hansen <[email protected]> # for x86 Link: https://lore.kernel.org/r/[email protected] [ Added 6.13.y too to Cc: stable tag and added reasoning to avoid over-backporting. - Miguel ] Signed-off-by: Miguel Ojeda <[email protected]> Signed-off-by: Greg Kroah-Hartman <[email protected]>

rustbot assigned cjgillot Jan 27, 2025

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 Jan 27, 2025

RalfJung force-pushed the x86-abi branch from 9a08be7 to 98dc988 Compare January 27, 2025 17:27

RalfJung mentioned this pull request Jan 27, 2025

x86: make SSE2 required for i686 hardfloat targets and use it to pass SIMD types #135408

Open

RalfJung changed the title ~~Explicit choose x86 softfloat/hardfloat ABI~~ Explicitly choose x86 softfloat/hardfloat ABI Jan 27, 2025

This comment has been minimized.

Sign in to view

rustbot assigned lcnr and unassigned cjgillot Jan 28, 2025

workingjubilee assigned workingjubilee and unassigned lcnr Jan 28, 2025

RalfJung force-pushed the x86-abi branch 3 times, most recently from bd35e7c to 89fa12a Compare January 28, 2025 01:40

RalfJung commented Jan 28, 2025

View reviewed changes

RalfJung force-pushed the x86-abi branch from 89fa12a to 438bd1b Compare January 28, 2025 02:51

workingjubilee approved these changes Jan 28, 2025

View reviewed changes

RalfJung force-pushed the x86-abi branch from 438bd1b to 9f1a635 Compare January 28, 2025 02:58

RalfJung force-pushed the x86-abi branch from 9f1a635 to 11c35f3 Compare January 29, 2025 09:59

bors 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 Feb 3, 2025

RalfJung mentioned this pull request Feb 3, 2025

Tracking Issue for high failure rates on Windows MSVC CI with filesystem errors #127883

Open

RalfJung added 2 commits February 3, 2025 16:56

add rustc_abi to control ABI decisions LLVM does not have flags for, …

f755f4c

…and use it for x86 softfloat

patch RfL job

8596ce1

RalfJung force-pushed the x86-abi branch from f60cbda to 8596ce1 Compare February 3, 2025 15:58

bors added the merged-by-bors This PR was explicitly merged by bors. label Feb 3, 2025

bors merged commit f027438 into rust-lang:master Feb 3, 2025
7 checks passed

rustbot added this to the 1.86.0 milestone Feb 3, 2025

This was referenced Feb 4, 2025

change rustc-abi in custom targets to x86-softfloat rust-osdev/bootloader#491

Merged

change rustc-abi in custom targets to x86-softfloat rust-osdev/bootloader#492

Merged

RalfJung deleted the x86-abi branch February 4, 2025 13:31

RalfJung mentioned this pull request Feb 4, 2025

Nightly 2025-02-03 to 2025-02-04: "Error loading target specification: target feature sse2 is required by the ABI but gets disabled in target spec" with JSON target specification #136544

Open

https123456789 added a commit to https123456789/bootloader that referenced this pull request Feb 5, 2025

fix: add rustc-abi to satasfy latest nightly compiler

138efea

Required to enable the soft-float feature. See rust-lang/rust#136146

workingjubilee mentioned this pull request Feb 5, 2025

Add X86 target support for NuttX #136181

Closed

ojeda mentioned this pull request Feb 6, 2025

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

Open

90 tasks

riking reviewed Feb 8, 2025

View reviewed changes

phil-opp added a commit to phil-opp/blog_os that referenced this pull request Feb 10, 2025

Set rustc-abi: x86-softfloat in target JSON

688a21e

This additional field is required by the Rust compiler since rust-lang/rust#136146 .

phil-opp mentioned this pull request Feb 10, 2025

Add rustc-abi field to target spec JSON phil-opp/blog_os#1393

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Explicitly choose x86 softfloat/hardfloat ABI #136146

Explicitly choose x86 softfloat/hardfloat ABI #136146

RalfJung commented Jan 27, 2025 •

edited

Loading

rustbot commented Jan 27, 2025

rustbot commented Jan 27, 2025

RalfJung commented Jan 27, 2025

This comment has been minimized.

This comment has been minimized.

cjgillot commented Jan 28, 2025

RalfJung Jan 28, 2025

workingjubilee Jan 28, 2025

workingjubilee left a comment

workingjubilee Jan 28, 2025

workingjubilee Jan 28, 2025

RalfJung commented Jan 28, 2025 •

edited

Loading

workingjubilee commented Jan 28, 2025

lcnr commented Jan 28, 2025

rustbot commented Jan 28, 2025

bors commented Jan 29, 2025

RalfJung commented Jan 29, 2025

RalfJung commented Feb 3, 2025

bors commented Feb 3, 2025

RalfJung commented Feb 3, 2025

bors commented Feb 3, 2025

bors commented Feb 3, 2025

rust-timer commented Feb 4, 2025

riking Feb 8, 2025

RalfJung Feb 9, 2025

riking Feb 9, 2025

ojeda Feb 9, 2025 •

edited

Loading

ojeda Feb 9, 2025

	"'{s}' is not a valid value for rust-abi. \
	"'{s}' is not a valid value for rustc-abi. \

Explicitly choose x86 softfloat/hardfloat ABI #136146

Explicitly choose x86 softfloat/hardfloat ABI #136146

Conversation

RalfJung commented Jan 27, 2025 • edited Loading

rustbot commented Jan 27, 2025

rustbot commented Jan 27, 2025

RalfJung commented Jan 27, 2025

This comment has been minimized.

This comment has been minimized.

cjgillot commented Jan 28, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

workingjubilee left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

RalfJung commented Jan 28, 2025 • edited Loading

workingjubilee commented Jan 28, 2025

lcnr commented Jan 28, 2025

rustbot commented Jan 28, 2025

bors commented Jan 29, 2025

RalfJung commented Jan 29, 2025

RalfJung commented Feb 3, 2025

bors commented Feb 3, 2025

RalfJung commented Feb 3, 2025

bors commented Feb 3, 2025

bors commented Feb 3, 2025

rust-timer commented Feb 4, 2025

Overall result: no relevant changes - no action needed

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ojeda Feb 9, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

RalfJung commented Jan 27, 2025 •

edited

Loading

RalfJung commented Jan 28, 2025 •

edited

Loading

ojeda Feb 9, 2025 •

edited

Loading