CuLifter: Lifting GPU Binaries to Typed IR

TL;DR

CuLifter is a framework that accurately lifts GPU binaries to typed LLVM IR by recovering register types, enabling better binary analysis and optimization.

Contribution

It introduces a novel type recovery method for GPU binary lifting that achieves near-perfect lifting success across diverse benchmarks.

Findings

CuLifter successfully lifts 99.98% of GPU functions to valid LLVM IR.

Type recovery is essential; disabling it drops success rate from 73.8% to 0%.

The framework works across open-source, vendor, and ML runtime applications.

Abstract

GPU compilers merge all data types into a single unified register file, erasing the type information that binary-analysis tools rely on. We show that type recovery from this untyped register file is the central challenge of GPU binary lifting. We present CuLifter, a SASS-to-LLVM IR lifting framework that recovers register types via constraint propagation with conflict detection, reconstructs explicit control flow, and aggregates multi-instruction patterns. Across eight benchmark suites (24,437 GPU functions in 919 cubins) spanning open-source applications, vendor libraries, and optimized ML runtimes, CuLifter successfully lifts 99.98% of functions to valid LLVM IR. An ablation study confirms that type recovery is the only step required to produce semantically correct IR: disabling it drops the x86 pass rate from 73.8% to 0%, a 73.8 percentage-point drop.