Arrived in Portland for @tokioconf.com
I’m speaking Wednesday about when and how io_uring can make your storage-intensive tokio app better. Check out the full schedule, awesome lineup! www.tokioconf.com/schedule
And I just received this amazing swag drop in the mail: that beanie is going for a morning run right away!
Super excited to be speaking at #p99conf about the evolution of the Neon storage stack! My talk is on Wednesday: www.p99conf.io/agenda/#wedn...
Ok, that clarifies your original post for me, thanks!
So in terms of possible optimizations unlocked by explicit synchronization points, you’d say Rust today already exploits all of them? IIUC most of the Rust compiler and LLVM doesn’t understand the “meaning” of Poll::Pending, right? I’m interested in whether new optimizations were unlocked if it did.
Ok, so, expressivity in the typesystem, maybe reasoning about deadlock-freedom, etc.
Any foreseeable advantages wrt low level optimizations? Like, can I optimize the code *between* the synchronization points better if I know them at the type system level at compile time?
What improvements are unlocked in PL and compilers by marking synchronization points? Not doubting your point, just wondering what the broader context of the post is.