Speaker
Description
ReProspect is a Python framework designed to support reproducible prospecting of CUDA code - that is, the systematic analysis of CUDA-based libraries and software components through API tracing, kernel profiling, and binary analysis.
ReProspect builds on NVIDIA tools: Nsight Systems, Nsight Compute, and the CUDA binary utilities. It streamlines data collection and extraction using these tools, and it complements them with new functionalities for a fully programmatic analysis of these data, thus making it possible to encapsulate the entire prospecting analysis in a single Python script.
ReProspect provides a practical foundation for developing novel use cases of CUDA code prospecting. It supports collaborative code development by enabling developers to share concise, reproducible analyses that motivate design decisions and help reviewers grasp the impact of proposed changes. It also enables new types of tests that go beyond traditional output-correctness validation in CI/CD pipelines, such as validating the presence of instruction patterns in binaries or confirming expected API call sequences for key library functionalities. Additionally, ReProspect can act as a framework for structuring research artifacts and documenting analyses, enabling others to reproduce the work and build upon it more effectively.
The talk will expose the needs that ReProspect addresses, how the community would benefit from using it, and demonstrate it with concrete case studies inspired by contributions to Kokkos.
