12. Parallel Storage Allocation

[@leixun]

*My takeaways:* 1. Review of memory allocation primitives 0:45 2. Virtual memory allocation - mmap() 4:15 - mmap() vs malloc() 8:20 - Address translation 11:20 3. Stack implementations 15:35 4. Basic properties of storage allocators 41:20 5. Fragmentation 44:00 6. Parallel allocation strategies 50:25 7. False sharing 1:03:30 8. Parallel heap allocation 1:07:48

[@richardm5916]

Have gone through this entire course it is amazing, Learnt So so Much, recommended it to everyone

[@keyboard_toucher]

The way I see it, the most obvious and biggest problem with "Strategy 2" (55:29) is missed. It is that the programmer wouldn't be able to transfer ownership (i.e. deallocation responsibility) of objects between threads. The immediate problem with allocating on one thread and freeing on another, as one student suggested, would not be wastage of memory but rather heap corruption due to races in the heap management code. A mutex would have to be used to avoid that, which invalidates the "no locking"/"no synchronization" bullet points. This deficiency is later resolved in "Strategy 3".

[@PetrSkovron]

34:00 -- How can we reuse the memory used by the subproblems? Aren't they running in parallel?

[@pschneider1968]

Hmm, I did not understand anything from slide 19... 🤔 Not enough explanations/illustrations there (for me). Anyone care to explain?