Easy and Featureful Parallelism with Dagger.jl - JuliaCon 2021

Explore advanced parallelism techniques using Dagger.jl in this JuliaCon2021 talk. Learn why Dagger.jl surpasses Distributed.jl for complex problems and large-scale parallelization, offering GPU execution and fault tolerance. Follow along as the speaker builds a real-world application, demonstrating Dagger's powerful scheduler and resource utilization. Discover key features like distributed arrays, checkpointing, and GPU programming with DaggerGPU.jl. Gain insights into Dagger's architecture, its relationship with Distributed.jl, and upcoming features such as distributed tables, task and data scopes, and mutable chunks. Perfect for Julia developers looking to enhance their parallel computing skills and leverage Dagger.jl's full potential in their projects.

- Introduction.
- Overview.
- When and why use Dagger instead of Distributed?.
- Coding example.
- A note on Dagger.delayed.
- How does Dagger scheduler work?.
- How is Dagger different from what Distributed.pmap and friends do?.
- Dagger vs Distributed - Dagger is built on top of Distributed.
- Real application demo built on Dagger - security camera monitoring system.
- Dagger feature: distributed arrays.
- Dagger feature: checkpointing.
- Dagger feature: GPU programming with DaggerGPU.jl.
- Future Dagger feature: Distributed tables.
- Future Dagger feature: Task and data scopes.
- Future Dagger feature: Mutable chunks.
- Check the Issues on the Dagger repository for more.