Implement the table64 extension to the memory64 proposal#1531
Implement the table64 extension to the memory64 proposal#1531fitzgen merged 5 commits intobytecodealliance:mainfrom
Conversation
Fix validation of `local.tee` to push the local's type, not the actual type.
This commit implements validation, fuzzing, printing, etc, for the extension of the memory64 proposal that adds 64-bit tables. These tables work similarly to memories where the index type is statically listed and is required for all various operations. The bulk of the changes here were to wasm-smith to generate modules with these new table instructions.
|
I'll note that this was mostly motivated by keeping up to date with the testsuite submodule which now has table64 tests, coupled with the fact that implementing it wasn't too too bad. In updating the spec testsuite submodule that uncovered a minor issue with the gc proposal's validation of I'm running the |
There was a problem hiding this comment.
We really gotta bit the bullet and get this crate using wasm-encoder one of these days... Wish I did it years ago...
| config.allow_start_export = false; | ||
|
|
||
| // NB: just added "table64" support to this and wasmtime doens't | ||
| // implement that yet |
Pointed out in bytecodealliance#1531
Pointed out in #1531
Brings wasm-tools and all the various bits and pieces here up-to-date with the latest version of the spec. This includes support for a toggle between using i32 or i64 as the index type for tables along with all the consequences therein.