> But anything that's happening on the GPU timeline is pretty much impossible to do safely. Rust's type system is not sufficiently capable of modeling this stuff without tons of runtime checks, or making the API so awful to use nobody will bother.
I wonder if there's any thinking / research around what the PLT (programming language tech) would look like that could manage this. Depending on what kind of safety is sought, compile-time safety is not necessarily the only way to ensure this.
Of course it depends greatly on what kind of safety we are looking for (guaranteed to run without error vs memory-safe behaviour that might bail out in some cases, etc)
I wonder if there's any thinking / research around what the PLT (programming language tech) would look like that could manage this. Depending on what kind of safety is sought, compile-time safety is not necessarily the only way to ensure this.
Of course it depends greatly on what kind of safety we are looking for (guaranteed to run without error vs memory-safe behaviour that might bail out in some cases, etc)