Sync the integer unary operators with AstSemantics.md.#44
Conversation
|
LGTM, but I am rather confused why we have neg & abs for floats but not for ints. What's the motivation? |
|
Same question. Shouldn't we just add int neg/abs/not to astsemantics? Is there a reason to omit them? All three seem useful and common |
|
The FP operations cannot be emulated with other FP operations (though they technically can be with integer operations and a healthy assumption that sign is MSB). The int operations are redundant, and we've gone back and forth on which should be in the IR and which shouldn't. |
|
We discussed integer |
|
I think the != discussion led us to decide that it probably wasn't worth being stingy on opcodes, though. It's probably worth opening an issue for these, and basing our decision on data once we have large codebases? |
|
+1 for waiting on larger applications. I think we've got everything we need On Wed, Sep 2, 2015 at 7:16 PM, JF Bastien notifications@github.com wrote:
|
|
So it sounds like for now this is lgtm to merge, and file an issue on the design repo to gather data? |
|
Merging because I believe we basically have consensus here, at least with respect to bringing the spec repo into accordance with the design repo. If people want integer abs/neg/not or other things, they're welcome to pursue them in the design repo first. |
84836e2 to
3c58f3d
Compare
AstSemantics.md presently has no integer neg, abs, or not operators. It does have a popcnt operator.
3c58f3d to
85846ca
Compare
Sync the integer unary operators with AstSemantics.md.
…Interface Accessing SIMD types form JavaScript throws
This change modifies `assert_return`. After the Wasm function is invoked, the result will either be `undefined`, a JS `Number`, or a JS `Array`. It's simpler to treat them all as an array, with `undefined` representing an empty array, and `Number` representing an array with one element. We don't check for `Number` here, since as soon as we add support for the reference-types proposal, a scalar result may be an `Object` too. Fixes issue WebAssembly#44.
Rename new traps.wast test -> traps0.wast
…ebAssembly#44) * [spec] resume_throw should accept multiple handlers (WebAssembly#71) * Add missing * --------- Co-authored-by: Alan L <gh@symb.olic.link>
These proposals have both been merged in the main spec repo.
AstSemantics.md presently has no integer neg, abs, or not operators. It
does have a popcnt operator.