Stabilize wasm32 memory-related intrinsics #613
Conversation
coresimd/wasm32/memory.rs
Outdated
| @@ -47,9 +48,10 @@ pub unsafe fn size(mem: i32) -> i32 { | |||
| #[inline] | |||
| #[cfg_attr(test, assert_instr("memory.grow", mem = 0))] | |||
| #[rustc_args_required_const(0)] | |||
| pub unsafe fn grow(mem: i32, delta: i32) -> i32 { | |||
| #[stable(feature = "simd_wasm32", since = "1.33.0")] | |||
| pub unsafe fn memory_grow(mem: u32, delta: usize) -> isize { | |||
There was a problem hiding this comment.
memory.grow conceptually returns the value that memory.size would return after the resize, which suggests that it should be usize rather than isize (with the error value being usize::MAX rather than -1). However, I don't think it'll make a big difference in practice, so I don't feel strongly about it.
There was a problem hiding this comment.
Good point. User programs can use as to go either way anyways, and to me it feels that both approaches have downsides - using isize might mean that negative values might indicate success, while using usize will confuse those trying to use -1 here to catch errors like the spec suggests =/
Using usize here and updating the docs to suggest using usize::MAX feels like the lesser evil.
There was a problem hiding this comment.
That sounds reasonable to me! I've switched to usize and documented the return value on failure. Thanks for the suggestions @sunfishcode!
This commit stabilizes the wasm32 memory-related intrinsics, as specified in rust-lang/rust#56292. The old intrinsics were removed and the current intrinsics were updated in place, but it's the last breaking change!
alexcrichton
force-pushed
the
wasm32-memory
branch
from
a4f9689 to
30da169
Compare
This commit stabilizes the wasm32 memory-related intrinsics, as
specified in rust-lang/rust#56292. The old intrinsics were removed and
the current intrinsics were updated in place, but it's the last breaking
change!