2025-07-23 –, Main Room 4
Dagger.jl makes your code go fast, now in innovative new ways! Come here about what's new in Dagger since JuliaCon 2024, how Dagger has continued to develop over the past year, and what you have to look forward to by JuliaCon 2026!
Dagger.jl is a Julia-native task runtime, scheduler, and unified parallelism interface for all kinds of applications. Dagger boasts a wide range of interfaces tailored to parallelizing various kinds of problems, whether you're working with arrays, tables, graphs, or your own custom data structures. Dagger also provides a variety of ways to express operations, such as plain one-off tasks, streaming (continuous) tasks, tasks with implicit data dependencies, and much more. Finally, Dagger supports not just CPUs, but also automatically can utilize GPUs from all 4 major vendors (NVIDIA, AMD, Apple, and Intel) near seamlessly, whether on one node or one hundred nodes.
The purpose of this talk is to let you know what's new, and what's to come. We'll cover everything that's changed or been added since JuliaCon 2024, and how that can be a big benefit to you and your use case. We'll also take a look at what kind of work is slated for the road to JuliaCon 2026, and what possibilities this improvements will unlock.
Be sure to also attend the Dagger Birds of a Feather (BoF) to voice your interest and concerns, and speak directly with the Dagger maintainers and other Dagger users like yourself!
RSE at MIT's JuliaLab, author of AMDGPU.jl, and maintainer of Dagger.jl. I want to improve the accessibility of HPC so that everyone can easily and productively scale their code to the max!