Compiler-Assisted Floating-Point Error Analysis and Profiling with FPChecker

Hacker News Top Events
compiler floating-point error-analysis profiling hpc scientific-computing tutorial

Summary

A half-day tutorial at ISC High Performance 2026 on using compiler-assisted tools (FPChecker/LLVM) for floating-point error analysis and profiling in C/C++ scientific codes.

No content available
Original Article
View Cached Full Text

Cached at: 07/01/26, 02:00 PM

# Compiler-Assisted Floating-Point Error Analysis and Profiling with FPChecker · Tutorial Source: [https://fpanalysistools.org/ISC26/](https://fpanalysistools.org/ISC26/) #### ISC High Performance 2026 \(ISC26\), Hamburg, Germany June 22, 2026 Time: 9:00am \- 1pm Europe/Berlin \(half\-day tutorial\) Location: CCH – Congress Center Hamburg, Room: Hall X12 \- 1st Floor ![](https://fpanalysistools.org/ISC26/photos/SC24_tutorial_v4.jpg) ## Description Floating\-point arithmetic is central to scientific computing, but small rounding effects can accumulate into significant numerical errors, especially when applications are ported to lower precision for performance and energy savings\. This tutorial introduces compiler\-assisted analysis tools built on the FPChecker framework \(https://fpchecker\.org/\) and clang/LLVM to instrument C/C\+\+ code and expose floating\-point behavior\. Participants will learn how to evaluate dynamic range and precision requirements, track rounding and relative error propagation across code regions, and detect infinities and NaNs caused by numerical limits\. Hands\-on examples, including linear solvers, finite\-difference methods, and other C/C\+\+ codes, will run on participants’ laptops\. In the advanced session, attendees can apply the tools to applications or widely used numerical libraries\. All tools and examples will be distributed through Conda, with pre\-configured AWS instances available for participants who cannot use it\. ## Presenter - [Ignacio Laguna](https://lagunaresearch.org/), Lawrence Livermore National Laboratory ## AWS Instances [https://fpchecker\.org/usernames](https://fpchecker.org/usernames) ## Presentation Slides The tutorial slides are here:[slides](https://fpanalysistools.org/ISC26/slides/ISC26-fpchecker_tutorial.pdf) ## Repositories: - FPChecker:[https://github\.com/LLNL/FPChecker](https://github.com/LLNL/FPChecker) - [Documentation](https://fpchecker.org/)

Similar Articles

Intermediate Floating-Point Precision

Lobsters Hottest

This article explores how intermediate floating-point precision in C++ code depends on compiler settings, CPU flags, and architecture, particularly on x87 FPU, and how this affects performance and calculation results.

FP8 is All You Need (Part 1): Debunking Hardware FP64 as the HPC Holy Grail

arXiv cs.AI

This paper argues that using FP8 tensor cores with Ozaki Scheme II can replace native FP64 hardware for high-performance scientific computing on AI-optimized GPUs like NVIDIA's B300, achieving full double-precision accuracy at much higher throughput. The authors present a Tensor-Memory Equilibrium model and show that emulated FP64 performance can exceed native FP64 by orders of magnitude across all workloads.

Performance improvements in libffi

Lobsters Hottest

This article details a performance improvement in libffi, where caching argument placement as a flat list of moves (a 'plan') eliminates redundant reclassification on every function call, offering significant speedups without resorting to JIT compilation.