Pith Number

pith:LVSFEATX

pith:2026:LVSFEATXU2QZEJTWUDTAEZNY5D

not attested not anchored not stored refs pending

Scalable Packed Layouts for Vector-Length-Agnostic ML Code Generation

Ege Beysel, Jan Moritz Joseph, Maximilian Bartel

Vector-length-aware packed data layouts enable ML compilers to generate efficient vector-length-agnostic code for scalable vector hardware.

arxiv:2605.12445 v2 · 2026-05-12 · cs.PF

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{LVSFEATXU2QZEJTWUDTAEZNY5D}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

We present an approach for enabling VLA code generation in an end-to-end ML compilation pipeline through vector-length-aware packed data layouts and corresponding compiler extensions. ... Evaluated on real-world ML workloads on Arm CPUs, our approach generates SVE code that is competitive with, and often outperforms, existing NEON-based code generation within IREE, achieving up to 1.45× speedup.

C2weakest assumption

That the vector-length-aware packed data layouts and extensions to tiling, fusion, and vectorization can be integrated into MLIR/IREE without introducing correctness issues, significant runtime overhead, or suboptimal performance for untested workloads, and that the evaluated ML tasks represent general cases.

C3one line summary

Packed layouts and extensions to tiling/fusion/vectorization in MLIR/IREE enable VLA ML code generation for SVE, achieving up to 1.45x speedup over NEON and outperforming PyTorch frameworks while scaling with vector length.

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T00:05:47.343311Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

5d64520277a6a1922676a0e60265b8e8ee2d9813c6e5781fcd358295093df4cb

Aliases

arxiv: 2605.12445 · arxiv_version: 2605.12445v2 · doi: 10.48550/arxiv.2605.12445 · pith_short_12: LVSFEATXU2QZ · pith_short_16: LVSFEATXU2QZEJTW · pith_short_8: LVSFEATX

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/LVSFEATXU2QZEJTWUDTAEZNY5D \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 5d64520277a6a1922676a0e60265b8e8ee2d9813c6e5781fcd358295093df4cb

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "4712656a39dbfdbb863cc365fd31939af29442d16c40b36e4fd3beed1d1ec4a8",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/publicdomain/zero/1.0/",
    "primary_cat": "cs.PF",
    "submitted_at": "2026-05-12T17:39:24Z",
    "title_canon_sha256": "13fe46a5738069243f192a56cf12f8a2f0cb5ef0791621d6b0f81434930c7572"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.12445",
    "kind": "arxiv",
    "version": 2
  }
}