ATLAAS: Automatic Tensor-Level Abstraction of Accelerator Semantics

TL;DR

ATLAAS is an end-to-end MLIR-based pipeline that automatically lifts RTL-extracted accelerator semantics to high-level tensor ISA specifications, enabling software stack generation for tensor accelerators.

Contribution

It introduces the first MLIR-based semantic lifting pipeline that automates the creation of tensor ISA specifications from RTL, reducing manual effort and enabling automatic software stack generation.

Findings

Achieves up to 92.9% collapse of bit-level MLIR on Gemmini.

Discovers hardware features omitted from reference designs.

Successfully lifts all modules of TVM's VTA without accelerator-specific modifications.

Abstract

Numerous tensor accelerator designs have been proposed, yet most lack well-documented ISAs and compiler backends, limiting evaluation to a handful of operators. Recent work has shown that given a tensor-level ISA specification, complete software stacks including compiler backends can be automatically generated--but writing such specifications remains a manual, expert-driven process. We present ATLAAS, the first end-to-end MLIR-based pipeline that lifts RTL-extracted accelerator semantics to tensor ISA specifications. Starting from bit-level LLVM IR produced by prior architecture-level model extraction, ATLAAS applies an 8-pass semantic lifting pipeline that progressively recovers high-level tensor structure--MAC idioms, saturation semantics, multi-dimensional buffer organizations, and data layout transformations--emitting specifications that immediately enable automatic software stack generation through the ACT ecosystem. We evaluate ATLAAS on the Gemmini systolic-array accelerator, where the pipeline collapses bit-level MLIR by up to 92.9% on processing elements and 24-34% on controller modules. ATLAAS discovers hardware features omitted from the hand-written reference, with correctness validated via Z3 SMT equivalence proofs. Generality is confirmed on TVM's VTA processor, where the same pipeline lifts all four datapath modules without accelerator-specific changes, enabling an automated path from RTL to a performance-competitive compiler backend.