Tutorial Day

On this years ARITH we are organizing a special Tutorial Day before the conference on Saturday, June 27.
The goal of this day is to give an introduction to several different topics to young researchers.
This year the topics are mixed precision deep neural network training & compression, fast, approximate but rigorous floating-point arithmetic, and multi-level compilation for hardware implementation.
Each tutorial is illustrated on actively developed tools.

The tutorial day is free of charge for registered ARITH attendees.
We will start with presentations in the morning and hands-on sessions in the afternoon.
This is an opportunity to learn from experts and engage with peers before the main event.

Time: Saturday 27th June 09:30-17:30

Registration

Registration is mandatory and open at https://forms.gle/JjMdTgj4zdMwXvXdA

Program (Preliminary)

Time (UTC+2)

Saturday June 27th, 2026
9:30 to 10:00 Registration and Welcome
10:00 to 12:00 Silviu Filip: Tools for the study of mixed precision DNN training & compressions
12:00 to 13:00 Lunch
13:00 to 15:00 Christoph Lauter: Fast but wrong? An introduction to fast, approximate but rigorous floating-point arithmetic, from multiplication over addition to elementary functions
15:00 to 15:30 Break
15:30 to 17:30 Samuel Coward: MLIR/CIRCT
Many people have heard of MLIR, so this tutorial will provide a brief overview of how compiler technologies are offering an alternative approach to EDA. Exploring the CIRCT project, we will demonstrate the Verilog front-end, formal verification tooling and the beginnings of a synthesis pipeline. (https://circt.llvm.org/)

This will be an interactive tutorial so participants are welcome to follow along using a docker build that is available here: https://github.com/cowardsa/CIRCT_TUTORIAL_2026

Organization