57th HiRSE Seminar
On Thursday, 22nd October at 16:00 CEST, Dr. Peter Pirkelbauer from Lawrence Livermore National Laboratory’s Center for Applied Scientific Computing will join us at the HiRSE Seminar Series to talk about “CompilerGPT: Leveraging Large Language Models to Analyze and Act on Compiler Optimization Reports”
Abstract:
Compiler optimization reports contain valuable insights, but their technical complexity often makes them difficult for programmers to interpret and act upon. This talk presents CompilerGPT, a framework that connects compilers, large language models, and a user-definable evaluation harness in an iterative compile, optimize, and evaluate workflow. CompilerGPT uses optimization reports, source code, compiler diagnostics, and performance measurements to guide LLM-generated code transformations while checking correctness and tracking results.
We will describe the framework, its configurable compiler and model interfaces, and experimental results using Clang and GCC, multiple LLMs, and representative sequential and OpenMP benchmarks. We will review demonstrated speedups of up to 6.5x, along with the variability, cost, and user involvement observed across workloads. The talk will also present updated results and examine CompilerGPT in the broader context of coding agents, discussing how compiler feedback, testing, profiling, and performance evaluation can serve as tools within agentic software-development systems.
The announcement and connection details are posted here:
- HiRSE-internal HiRSE Mattermorst channel,
- HIFIS Mattermost channel,
- Helmholtz Open Science Mattermost channel,
- FZJ-internal #rse Rocket.Chat channel,
- de-RSE mailing list and de-RSE Mattermost channel,
- Helmholtz Open Science Pro Mailinglist,
- RSE mailing lists of the HiRSE partners (if any).
If you are unable to join via one of these channels please email hirse@fz-juelich.de.
Location: virtual