Conversation
Oleksandr Natalenko, MSE 
oleksandr@natalenko.name
@ljs locking is for losers who cannot do state machines
0
1
2
Petr Tesarik
ptesarik@fosstodon.org@ljs You're wrong. `smp_wmb()` I was forced to avoid a lock in my SWIOTLB patch series, and man `__dma_rmb()`, getting lockless algorithms right `barrier_nospec()` is anything but simpler.
__smp_mb__after_atomic()
1
0
2
@ptesarik bro I didn't say lockless, I mean let's just go yolo yeah?
life is easier withou memory barriers too, just let that pipelined cpu write random shit out of order around your fence and chill, yolo son! Yolo!
life is easier withou memory barriers too, just let that pipelined cpu write random shit out of order around your fence and chill, yolo son! Yolo!
1
0
4
1
3
1
@ljs You are right, Windows' IRQL were a superior design /s
1
0
1
@ljs There's ozlabs folklore of Anton eliding all the locks from the kernel, booting it on a PPC machine, and achieving surprising uptime.
1
0
2