Implementing Parallelism

on YouTube ↗19m

" summary

TinyGo now runs goroutines in parallel on Linux, macOS and the RP2040 dual-core by mapping one goroutine to one OS thread, claiming simpler semantics at the cost of more RAM and context switches. It implements synchronization with futexes and a futex wrapper, refactors channels to use locks, changes garbage collection to stop-the-world via inter-core interrupts on bare metal, and uses a global lock for select to avoid deadlocks. RP2040 Mandelbrot rendering shows clear speedup; macOS may hit ~2000 thread limits.

AI-written summary · about

" chapters (11)

1. 0:00Introduction to parallelism and concurrency
2. 1:28TinyGo multi-core support
3. 3:38One thread per goroutine
4. 5:07Parallelism changes in TinyGo
5. 6:08What are futexes
6. 9:19sync.Group with futexes
7. 11:12Channels for parallelism
8. 11:58Select without deadlocks
9. 13:41Bare metal multi-core
10. 15:51Demo on RP2040
11. 17:20TinyGo parallelism future

" related talks

• 23mAn Operating System in GoPatricio Whittingslow · GopherCon 2025
• 35mGoroutines: the dark side of the runtimeRoberto Clapis · GoLab 2018
• 18mApplied concurrency in GoMatt Aimonetti · dotGo 2015
• 47mQueues, Fairness, and The Go SchedulerMadhav Jivrajani · GopherCon 2021
• 35mGo’s Trace Tooling and ConcurrencyBill Kennedy · GopherCon 2025
• 16mIn the land of Gophers from age 3Nicolas Lepage · DevFest Nantes 2018
• 14mGo's work stealing schedulerJBD · dotGo 2017
• 08mA Small Update On TinyGoRon Evans · GopherCon Europe 2022