Zhantong Qiu

Zhantong Qiu

Computer systems researcher focusing on performance evaluation, simulation, and hardware-software co-design

Nugget: Portable Program Snippets

Published in IEEE International Symposium on High-Performance Computer Architecture (HPCA), 2026

Abstract

Evaluating architectural ideas on realistic workloads is increasingly challenging due to the prohibitive cost of detailed simulation and the lack of portable sampling tools. Existing targeted sampling techniques are often tied to specific binaries, incur significant overhead, and make rapid validation across systems infeasible. To address these limitations, we introduce Nugget, a flexible framework that enables portable sampling across simulators, hardware, architectural differences, and libraries. Nugget leverages LLVM IR to perform binary-independent interval analysis, then generates lightweight, cross-platform executable snippets (nuggets) that can be validated natively on real hardware before use in simulation. This approach decouples samples from specific binaries, dramatically reduces analysis overhead, and allows researchers to iterate on sampling methodologies while efficiently validating samples across diverse systems.

Main Contributions

  1. Efficient and Portable Interval Analysis on Real Hardware: Using LLVM, unconstrained by the hardware’s microarchitecture or ISA, enabling analysis that runs natively on real systems without simulation overhead

  2. LLVM IR-Level Unit of Work: An IR-level abstraction enabling sample selection algorithms to create and locate samples in a cross-platform manner, agnostic of machine-level instructions and architectural-specific optimizations

  3. Cross-Platform Sample Creation: A sample creation methodology that allows samples to run on any platform (e.g., real hardware or simulators) without recompilation or ISA-specific modifications

  4. Efficient Validation Methodology: Validates selected samples for target workloads and inputs on native hardware, integrated into the development workflow for rapid iteration

Problem Statement

Prior targeted sampling techniques have three main drawbacks:

  • Expensive to find: Require running the entire application via simulation to define intervals and identify representative samples
  • Tied to single binary: Generated samples are specific to a particular binary, requiring full resampling on any binary change
  • Infeasible to validate: Require running applications in detailed simulation to verify sample accuracy

Nugget addresses all three limitations by enabling agile sampling development with real hardware validation.

Research Impact

This work significantly improves the efficiency of computer architecture research by enabling researchers to:

  • Rapidly prototype and validate sampling methodologies on real hardware
  • Create portable samples that work across different binaries, ISAs, and microarchitectures
  • Reduce the time and computational cost of sampling methodology development
  • Decouple sampling algorithm development from binary-specific constraints

Read the full paper on arXiv

Artifact Evaluation GitHub Repo

Recommended citation: Qiu, Z., Samani, M., & Lowe-Power, J., "Nugget: Portable Program Snippets." 2026 IEEE International Symposium on High-Performance Computer Architecture (HPCA). Sydney, Australia.
Download Paper