Sized Hierarchy and Scalable Vectors

@davidtwco

Metadata
Contact	@davidtwco
Team(s)	compiler (@davidtwco), lang (@nikomatsakis), libs (@Amanieu), types (@lcnr)
Goal document	2026/scalable-vectors

Summary

Over the next year, we will build on the foundational work from 2025 to stabilize
the Sized trait hierarchy and continue nightly support for scalable vectors:

Stabilize the refined Sized trait hierarchy (without constness), unblocking extern types
Propose and implement const Sized to support scalable vectors
Achieve RFC acceptance for [RFC #3838] (Scalable Vectors)
Land SVE types and intrinsics in stdarch for nightly experimentation
Continue addressing stabilization blockers for SVE itself
Begin design work for supporting the Scalable Matrix Extension (SME)

The const Sized work (Part II of [RFC #3729]) is deferred to a future goal,
allowing us to deliver value sooner through the trait hierarchy stabilization.
This future work interacts with ongoing const generics
efforts, as const Sized depends on progress in const traits.

This issue is intended for status updates only.

For general questions or comments, please contact a goal owner or post in the Zulip topic.

This issue is intended for status updates only.

For general questions or comments, please contact the owner(s) directly.

@davidtwco

An initial update on what we've been up to and some background:

This goal is submitted on behalf of the Rust team at Arm, but primarily worked on by myself (@davidtwco) and @JamieCunliffe. Anyone interested can always contact me for updates and I'll keep this issue up-to-date.
@JamieCunliffe been working on supporting Arm's scalable vector extension (SVE) for a couple years - primarily in RFC: Add a scalable representation to allow support for scalable vectors rfcs#3268 and its implementation Support for a scalable simd representation rust#118917.
- Through this work, we've discovered other changes to the language necessary to be able to support these types without special cases in the type system, which we're also working on (see below).
- Jamie is still resolving feedback on this RFC and its implementation, and keeping it rebased. We hope that it can be landed experimentally now that there's a feasible path to remove the special cases in the type system (see below).
- The next steps for this RFC and implementation are..
  - ..to continue to respond to feedback on the RFC and implementation.
I've (@davidtwco) been working on Hierarchy of Sized traits rfcs#3729 which improves Rust's support for exotically sized types, and would allow scalable vectors to be represented in the type system without special cases.
- We've had two design meetings with the language team about the RFC and had a broadly positive reception.
  - There is a non-strict dependency on const traits (Make trait methods callable in const contexts rfcs#3762) which has created uncertainty as to whether this RFC could be accepted without the specifics of const traits being nailed down.
- I've been working on implementing the RFC: an initial implementation of the non-const traits has been completed and adding the const traits is in-progress.
  - The language team have indicated interest in seeing this land experimentally, but this will depend on whether the implementors of const traits are okay with this, as it would add to the work they need to do to make any syntactic changes requested by the language team in Make trait methods callable in const contexts rfcs#3762.
- I'm continuing to respond to feedback on the RFC, but as this has largely trailed off, the next steps for this RFC are..
  - ..for the language team to decide to accept, reject, or request further changes to the RFC.
  - ..for progress on the implementation to continue.

A small update, we've opened a draft PR for the initial implementation of this - rust-lang/rust#137944. Otherwise, just continued to address feedback on the RFCs.

This is a continuing project goal, and the updates below this comment will be for the new period 2026

@nikomatsakis

Since our last update, on the Scalable Vector half of the goal:

We've landed SVE intrinsics on nightly (initial SVE intrinsics stdarch#2071), enabling users to experiment with SVE for the first time. We've since been working on a bunch of follow-ups:
- Enabling load/store tests on MSVC where we were observing spurious failures (Re-enables & Fixes AArch64 SVE ld/st tests for MSVC target stdarch#2142)
- Disabling SVE intrinsics for big endian where they aren't available (Adds cfg to ensure little endian for SVE stdarch#2145)
- We've investigated whether SVE intrinsics should be available on arm64ec
- We're looking into what we need to do to get the svpfalse intrinsic emit a pfalse intrinsic - either adding a compiler intrinsic for that (Create a sve_zeroinitializer compiler intrinsic rust#157110) or using mem::zeroed
- We're also looking into inline asm support for scalable vectors
We're also following up with intrinsic-test support for SVE in intrinsic-test: sve support stdarch#2160
- We've landed some compiler patches in support of this (codegen_ssa: peel trans. wrappers on scalable vecs rust#157768 and scalable vecs size incl. num vecs + no sret for scalable vecs rust#157915)
- We've also landed stdarch prerequisites and refactorings in support of this (intrinsic-test: run rustfmt on generated sources stdarch#2122, intrinsic-test: add lots of documentation stdarch#2124, intrinsic-test: sve type parsing and constraints stdarch#2121, intrinsic-test: remove arm::argument module stdarch#2127, intrinsic-test: refactor populate_random and related code stdarch#2126, update arm_intrinsics.json for svset and svget stdarch#2162, intrinsic-test: fwd args in intrinsic-test.sh stdarch#2164 and core_arch: redefine svrev, svzip and svuzp stdarch#2163)

On the Sized Hierarchy half of the goal:

We've been working with @nikomatsakis to clarify some of the open questions that came out of our last design meeting on "only bounds", to which Niko wrote a blog post. We've since started a crater run to understand the potential impact of adding a contextual only keyword (crater: contextual implementation of only trait bound modifier rust#157696)
We had a session at the All Hands, primarily with t-libs, on the naming of the traits in the Sized Hierarchy proposal and believe we've come to some early consensus on Sized, MetadataSized and MaybeSized
We had a session at the All Hands and a follow-up call with @lcnr and @yoshuawuyts to understand the migration strategy for the Sized Hierarchy traits and new default auto traits, and how we could work together with the other ongoing project goals
- We're going to start implementing the FCW-approach so we can start understanding what challenges the implementation side of that work will have
We plan to follow-up on our open PR to add a redundant sizedness bound lint since discussions around this work in general have advanced (Uplift and extend clippy::needless-maybe-sized into rustc rust#145924)
We've been working on advancing the next solver work in general, assisting @lcnr and the types team, as we'll need it eventually for const sizedness

We plan to update the RFC to reflect all of the above once we've made a little bit more progress.

@lqd

Since our last update, on the Scalable Vector half of the goal:

We finished landing intrinsic-test support for SVE in intrinsic-test: sve support stdarch#2160
- Since the last update, this involved the following rustc prerequisites:
  - codegen_ssa: no dbginfo for scalable vec local w/ -O0 rust#158088
  - codegen_ssa: multiply scalable vec size by vscale rust#158253
- ..and the following stdarch prerequisites:
- We intend to follow-up on this to re-enable some tests once LLVM is updated to include [LLVM][InstSimplify][SVE] Only transform sve.eorv(all_active, A)->0 when A is the result of a splat. llvm/llvm-project#203954 and we'll also be able to revert codegen_ssa: no dbginfo for scalable vec local w/ -O0 rust#158088
- We'll also re-enable GCC testing once we can upgrade stdarch CI to one of GCC 17, 16.2, 15.4, 14.5 or 13.5
We've opened a patch for inline assembly support for SVE in Adds support for AArch64 SVE to inline assembly rust#158312
We're aware of LLVM ICEs that can be triggered when using SVE, which we'll focus on looking into in the next month or two

On the Sized Hierarchy half of the goal:

@lqd has continued working on an implementation of the FCW that we plan for the migration
We concluded that there wouldn't be any major issues with adding a contextual only keyword, just the usual edition churn, once we've confirmed that's what t-lang want, we can go ahead with it
We had some calls with @tmandry, @nikomatsakis and @the8472 to discuss trait naming, syntax, etc.
- @davidtwco to schedule a follow-up call to that
- @nikomatsakis to write up his conclusions from the call
- @tmandry started a thread to discuss "standard patterns" that we'd want to make enabling easy/concise - #t-lang/implicit-trait-bounds > expressing common patterns concisely
@lqd has continued working to advance the stabilisation of the next solver (see Stabilize the next-generation trait solver #113) which we'll need for const sizedness

@lqd

Since our last update, on the Scalable Vector half of the goal:

We revisited our patch for inline assembly support for SVE - Adds support for AArch64 SVE to inline assembly rust#158312 - and that should land soon
Following the LLVM 23 upgrade, we landed Revert "codegen_ssa: no dbginfo for scalable vec local w/ -O0" rust#160558 and sve: re-enable sveorv testing after LLVM upgrade stdarch#2204 to remove a workaround and enable more testing respectively
We landed rust-lang#rust#160642 to prevent projections into SVE types
We landed rust-lang#rust#161231 to prevent the dead code lint for triggering for the field of a scalable vector
We're aware of a couple bugs related to the ABI checks/handling for SVE types and plan to look into those next (e.g. Scalable vector ABI check seems incomplete rust#161753)

On the Sized Hierarchy half of the goal:

@lqd continues working on the FCW for migration, but with the stabilisation of Polonius on nightly ( 🎉 ), we've not made as much progress here as planned
We've not made much progress discussing syntax and naming with lang due to holidays/relocations/conferences/etc. I expect more progress on this front by the next update.

@Jamesbarford

Since our last update, on the Scalable Vector half of the goal:

@Jamesbarford has started working on f16b SVE vectors and intrinsics (see Tracking Issue for f16b rust#160630 for f16b)

And on the Sized Hierarchy half of the goal:

@lqd continues working on the FCW for migration
We had a constructive meeting with @nikomatsakis in our regular sync where we agreed on a path forward for the open syntax/naming questions from lang, which we'll update our RFC with (perhaps once we know more about the migration part from the FCW work) and the implementation with

Navigation Menu

Repository navigation

Sized Hierarchy and Scalable Vectors #270

Summary

35 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Sized Hierarchy and Scalable Vectors #270

Description

Summary

Activity

nikomatsakis commented on Feb 18, 2025

davidtwco commented on Feb 19, 2025

davidtwco commented on Mar 3, 2025

35 remaining items

tomassedovic commented on May 5, 2026

davidtwco commented on Jun 16, 2026

davidtwco commented on Jul 15, 2026

davidtwco commented on Sep 3, 2026

davidtwco commented on Oct 15, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions