True parallelism, with no concept of threads

on YouTube ↗01h 06m

" summary

IncludeOS exposes per-core event loops so code can assign a lambda to a CPU and run without OS context switches; an SMP_add_tasks API wakes cores and the demo boots VMs up to 255 cores successfully. A pthread backend that simply maps threads to cores wastes cores for blocking joins, so a user-space fiber stack switch (assembly save/restore RSP) implements true blocking; stackful fibers work across cores. Coroutine-based tests (many runs and t‑tests) show coroutines slightly faster but lack stackful blocking today.

AI-written summary · about

" chapters (51)

1. 0:00Introduction
2. 0:35What is include OS
3. 1:28Everything is C
4. 1:54Control all CPU cores directly
5. 2:19Do we need threads
6. 3:50Who likes blocking calls
7. 5:01Eventbased
8. 5:54Include OS
9. 6:23What are threads
10. 7:12What are schedulers
11. 7:59Is a thread concrete
12. 9:38POSIX threads
13. 11:54Reasons for using threads
14. 13:01Accessing the CPUs
15. 13:17Bootstrap processor
16. 13:36Programmable interrupt controller
17. 14:02Advanced programmable interrupt controller
18. 14:35Boot loader
19. 14:57Main function
20. 15:30Code
21. 16:11Event loop
22. 16:23Soft trigger
23. 17:22Halt
24. 17:50Sleep
25. 18:30Add Tasks
26. 19:17Add ID
27. 19:39Signal
28. 20:18Boot
29. 20:50Hyperthreading
30. 21:37Memory
31. 23:33Thread scheduler
32. 24:09Mission accomplished
33. 24:43Lets try to use
34. 24:55This is an attempt
35. 25:44Create a thread
36. 26:42Join
37. 27:13POSIX demands
38. 28:15POSIX output
39. 28:56Boost fiber
40. 30:06Preemptive timer
41. 31:19Performance
42. 31:32preemption
43. 34:07jumpstarter
44. 36:38stack switch
45. 39:20release mutex
46. 40:22stackswitch
47. 42:47worker fiber
48. 45:26threadrunner
49. 48:42increase number
50. 51:40quarantines
51. 53:20yield

" related talks

• 01h 02mOverview of Parallel Programming in C++Pablo G Halpern · CppCon 2014
• 48mConsiderations for choosing the parallel computing strategyDori Exterman · Meeting C++ 2015
• 01h 16mPaying for Lunch: C++ in the ManyCore AgeHerb Sutter · CppCon 2014
• 01h 07mAnother fundamental shift in Parallelism Paradigm?Michael S Wong · CppCon 2014
• 01h 05mDecomposing a Problem for Parallel ExecutionPablo G Halpern · CppCon 2014
• 59mC++ on its way to exascale and beyondThomas Heller · Meeting C++ 2015
• 01h 03mThe Landscape of ParallelismMichael S Wong · Meeting C++ 2015
• 01h 06mAsynchronous Computation in C++Hartmut Kaiser · CppCon 2014