Gonna pile on:
At this point we may need TSMC to make a specialized chip to run Electron.
At this point we may need TSMC to make a specialized chip to run Electron.
If you disassemble some armv8 binaries that aren't dealing with Javascript, you do see still see FJCVTZS.
https://web.archive.org/web/20201119143547/https://twitter.c...
> Uncontended acquire-release atomic operations are basically free on Apple Silicon
While I don't doubt you, the poster, specifically, how is this possible? To be clear, my brain is x86-wired, not ARM-wired, so I may have some things wrong. Most of the expense of atomic inc/dec is "happens before", which essentially says before the current core reads that memory address, it will be guaranteed to be updated to the latest shared value. How can this be avoided? Or is it not avoided, but just much, much faster than x86? If the shared value was updated in a different core, some not-significant CPU cycles are required to update L1 cache on current current with latest shared value. > some not-significant CPU cycles
should say: > some not-insignificant CPU cycles