Skip to content
Book Open access

Wavel: A Fast and Efficient Compilation System for Wafer-Scale Accelerators

Sep 2026 · Proceedings of the ACM SIGOPS 32nd Symposium on Operating Systems Principles · 0 citations · 37 references

TL;DR

Wavel is presented, a fast and efficient compilation system for wafer-scale accelerators that introduces MeshIR to make executable schedules explicit, constructs them through two legal locality-preserving primitives, CUT and TUNE, and uses physical cost evaluation with structured pruning to keep search practical.

Abstract

Wafer-scale accelerators offer a new scaling point for AI infrastructure, but they also create a new compilation regime: communication cost varies sharply with location, and the space of possible placements and execution schedules is enormous. Existing GPU, distributed, and vendor compilation systems largely retain a sharding-oriented view and therefore fail to fully leverage these emerging accelerators, leaving the dominant physical scheduling decisions unresolved. We make a key observation: at wafer scale, the central complexity of compilation shifts from finding a logical shard plan to constructing an executable schedule that fixes placement, layout alignment, and execution ordering. Based on this observation, we present Wavel, a fast and efficient compilation system for wafer-scale accelerators. Wavel introduces MeshIR to make executable schedules explicit, constructs them through two legal locality-preserving primitives, CUT and TUNE, and uses physical cost evaluation with structured pruning to keep search practical. On WSE-3 hardware, Wavel delivers 1.31×-1.35× higher mean per-setting throughput than the expert-optimized WaferLLM and up to 1.85× speedup over the Full-Chip baseline on dense models; calibrated cost-model results predict a 2.97× speedup on MoE workloads over the same baseline.

Read PDF

Similar papers

Preprint Sep 2026

Compiler and Hardware Co-Design for Accelerator Architectures

Heterogeneous accelerator architectures offer an efficient path to performance for compute-intensive workloads. However, full-stack integration remains difficult. We present EAAC (Extensible Accelerator Architecture), a flexible and extensible compiler and hardware architecture designed to lower the overhead of hardwar...

Karl Herman Krause, Emad Jacob Maroun, Martin Schoeberl · 0 citations
Book Open access Sep 2026

MeshRT: Compile-Time Governed Wafer-Scale Runtime for Low-Latency High-Throughput Inference

This work presents MeshRT, the first wafer-scale system to achieve both ultra-low latency and high throughput for LLM inference via a novel system architecture, which is called compile-time governance of runtime dynamism.

Cong-Jie He, Le Xu, Zhan Lu et al. · 0 citations
Book Open access Sep 2026

Schedules Are Solvable Symbols: Tuning-Free Compilation of Tile Programs on Dataflow Architectures

Loom, a tuning-free symbolic compiler framework for tile-based SPMD programs on spatial dataflow architectures, is presented, suggesting that hardware-derived symbolic compilation provides a retargetable alternative to profiling-based tuning for spatial dataflow architectures while remaining interpretable by keeping op...

He-Ru Wang, Wei Li, Zhen-Yu Bai et al. · 0 citations
Review Open access Aug 2026

From Cray to Exascale: A Critical Survey of Supercomputer Architecture Evolution, Heterogeneity, and Interconnect Bottlenecks

A critical survey of supercomputer architecture from early vector machines to contemporary exascale systems, with particular emphasis on three persistent challenges: memory hierarchy design, interconnection network scalability, and thermal management.

Oraye Godspower · 0 citations
Conference Sep 2026

Shared Load-Store Unit for Instruction-Based Accelerators

Modern systems-on-chip (SoCs) rely on heterogeneous accelerators for performance scaling. Memory access is a critical bottleneck, but the complexity of cache coherence and nuances of weak memory consistency models (which may vary from system to system) represent a significant designer burden to every load/store unit th...

Joseph Maheshe, G. Lemieux · 3 citations
Preprint Aug 2026

To Scale Up or To Scale Out: Evaluating Space-Time Costs of Compiled Logical Circuits on Modular Superconducting Quantum Processors

Modular integration has emerged as the main pathway for scaling superconducting quantum processing units (QPUs) beyond the constraints of fabrication yield and physical footprint. Currently, two primary strategies lead this effort. Mirroring the"Scaling Up"and"Scaling Out"approaches in GPU architectures and AI infrastr...

Nikiforos Paraskevopoulos, S. de Bone, Mick Christophersen et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.