Open
Conversation
LegNeato
commented
Mar 7, 2026
| _args: &[OperandRef<'tcx, Self::Value>], | ||
| _is_cleanup: bool, | ||
| ) -> Self::Value { | ||
| bug!("LLVM intrinsic call not supported in SPIR-V backend: {instance:?}") |
Collaborator
Author
There was a problem hiding this comment.
I think we want to use this rather than our own home-grown replacements, but didn't want to do it in this PR.
LegNeato
force-pushed
the
pr-542-latest-nightly-refresh
branch
from
14cdf92 to
6434ab8
Compare
Contributor
|
|
Contributor
|
Tried to update with a patched Details |
Open
Collaborator
Author
|
Cool, I'll take a look. There were some changes in this area. |
fee1-dead
requested changes
Mar 9, 2026
LegNeato
force-pushed
the
pr-542-latest-nightly-refresh
branch
from
2fe48e0 to
db063f6
Compare
Collaborator
Author
|
@nazar-pc I think what you saw should be fixed. |
Contributor
|
Yes, clippy is happy now, tests are passing too |
fee1-dead
requested changes
Mar 13, 2026
Comment on lines
+86
to
+108
| @@ -86,6 +101,12 @@ pub(crate) fn provide(providers: &mut Providers) { | |||
| arg.mode = PassMode::Ignore; | |||
| } | |||
|
|
|||
| // SPIR-V backend lowers arguments by-value and cannot handle | |||
| // backend-specific indirection/casts at this layer. | |||
| if matches!(arg.mode, PassMode::Cast { .. } | PassMode::Indirect { .. }) { | |||
| arg.mode = PassMode::Direct(ArgAttributes::new()); | |||
| } | |||
Collaborator
There was a problem hiding this comment.
There's an make_direct_deprecated above this here. Seems to be duplication?
Collaborator
There was a problem hiding this comment.
It doesn't seem to be removed, could you check again?
Also note that make_direct_deprecated will panic if it uses PassMode::Cast, so this appears to be unreachable.
Collaborator
Author
There was a problem hiding this comment.
Whoops, I pushed to my fork rather than the upstream branch. Just pushed the commit. Yeah, this is unreachable anyway.
Update the pinned workspace toolchain to rustc 1.96.0-nightly (1d8897a4e 2026-03-13) and keep REQUIRED_RUST_TOOLCHAIN in sync. Mirror the current rustc_codegen_ssa crate-attribute header in lib.rs, keep rustc_codegen_spirv-specific lint allowances outside that mirrored block, and drop the now-unneeded bootstrap-only header filtering from build.rs.
Use exact rust_gpu attribute paths with get_attrs_by_path, let AggregatedSpirvAttributes parse filtered iterators directly, and replace the manual debug filename arena allocation with DroplessArena::alloc_str. Also port the rustc-private API drift in this nightly by importing assert_matches from std, using rustc_span::Spanned directly, and renaming BackendRepr::ScalableVector to SimdScalableVector.
Keep the existing runtime-array reification path for const pointer bitcasts, but report a specific zombie reason when the backing allocation size is not a multiple of the runtime-array element size instead of silently falling through to the generic const_bitcast error.
This disassembly test only cares about the rotate lowering pattern, not source line tables. Strip OpLine records entirely so line-number drift across toolchains does not require reblessing the test.
LegNeato
changed the title
Collaborator
|
cc @Firestar99 for some testing |
This is handled above.
Collaborator
Author
|
@Firestar99 did you get a chance to test this? |
Add a UI compile-fail test for casting a 3-byte constant allocation to `*const RuntimeArray<u16>`. This exercises the existing trailing-bytes diagnostic for unsupported unsized constants and guards the review concern that this shape must fail instead of silently lowering.
fee1-dead
approved these changes
Mar 23, 2026
Collaborator
fee1-dead
left a comment
fee1-dead
left a comment
There was a problem hiding this comment.
LGTM, had one last nit, otherwise this is ready to go and feel free to merge after CI green and testing (cc @Firestar99?)!
Comment on lines
+356
to
+350
| && let Some(init) = self.try_read_from_const_alloc(alloc, pointee) | ||
| { | ||
| return self.static_addr_of(init, alloc.inner().align, None); | ||
| if let Some(init) = self.try_read_from_const_alloc(alloc, pointee) { |
Collaborator
There was a problem hiding this comment.
nit: should move back to the let chain instead of putting another check inside.
The generic unsized constant reader already handles `RuntimeArray` constants, including the trailing-bytes error case covered by the new compile-fail test. Keep `const_bitcast` on the direct `try_read_from_const_alloc` path instead of carrying a second fallback that no longer changes behavior.
LegNeato
force-pushed
the
pr-542-latest-nightly-refresh
branch
from
ed2c27e to
3cce05f
Compare
Member
|
Looks like cargo-gpu needs an update to work with this branch... I'll have a look tomorrow (and properly review this) |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requires #542