Austin Clements

Since this is Linux-only, it could use the new runtime/internal/syscall package, rather than implementing it in assembly for all architectures.

For the syscall number constants, I would create new per-arch files in `runtime/internal/syscall` to define those constants. Eventually, we'll want to shift all of the constants out of `sys_linux_xxx.s` and...

@beoran , we are making baby steps toward better modularization in the runtime by moving the per-OS/arch syscall layer into runtime/internal/syscall and reorganizing it to have significantly less per-arch code....

There's still a distinction of what *semantics* we're exporting. The semantics of `go:notinheap` have changed a few times, but since it's internal it hasn't been a problem to just shape...

We discussed this in the runtime/compiler meeting today (sorry you weren't there @mdempsky !) and everyone's fine with the general idea of replacing the type pragma with an embedded type....

At least for my minor concern, I don't need to see a whole prototype, I just wanted confirmation that it's not a problem to convert the non-struct types to struct...

Duplicate of #54885

@klauspost > I may be lacking imagination, but I don't see the scope of this going much beyond inline decisions. To add to what others have said, we've actually been...

@apparentlymart, you're right that shrink-wrap software generally isn't our target for this. Perhaps we should mention that in the proposal and call out simplifying it as potential future work. There...

I find the arguments that @rhysh and @josharian have been making about debuggability (and bug reportability) pretty persuasive, but I also strongly believe that using pprof profiles as the source...