-
Notifications
You must be signed in to change notification settings - Fork 694
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
64-bit Widening Multiplication #1495
Comments
Do you mean the Arm64 equivalent of |
@dtig Thank you for your comment! Yes, For the architecture support, I can give a simple summary here.
About adding them to the spec, there are several considerations.
Considering those all, I suggest adding them to wasm spec as an extension. It could be expected that some implementations will happily implement it while others can ignore the extension. |
Shall we draft a proposal for it? |
Sorry I lost track of this issue, I think drafting a proposal is a good next step - though it might also be a good question for the CG or the SIMD subgroup to see if there's other operations that we can include as we don't have a precedent for introducing one-off operations, but there's probably a class of operations we could consider. Here is some documentation on how to get a proposal started, and in parallel this could also be a good topic for discussion at a future SIMD subgroup meeting. @penzn are the subgroup meetings still running? |
Thank you for your time. Yeah, it could be better to discuss it somewhere. I would like to hear some opinions from other people. I saw some people misunderstood this as trying to add the instruction directly into the core instr set, but it's not correct. Actually, I'm proposing this as an extension. Because it might be hard to do the 64-bit widening multiplication on some platforms. |
Currently, wasm doesn't have instructions to calculate the high bits of 64-bit widening multiplication. Is it possible to add it in the future?
For your information, LLVM emits a very slow routine to simulate it. And it seems unlikely to be optimized easily for those wasm JIT & AOT compilers.
The text was updated successfully, but these errors were encountered: