-
Notifications
You must be signed in to change notification settings - Fork 451
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
Any intention add fp16 & i16 native support in webassembly? #804
Comments
fp16 is supported in two broad manners on some HW: some only support it as a storage type (widen loads to 32 or 64, and narrow stores, with all arithmetic in 32 or 64), whereas some architectures also support 16-bit arithmetic. To move any fp16 proposal forward, it would be useful to have a survey of relevant architectures and what they support. When considering arithmetic, it's further useful to understand if full IEEE 754 is supported. |
Yeap, for some target such as MCU they only have i16 support. does i16 are necessary? |
I don't think i16 is necessary: we can expand any f16 conversion to i32 or i64, and do arithmetic on those types. |
i|f bitscope is critical for optimization & range profiling. |
We can use i16 as bitwise representation of fp16. |
My opinion on the topic: #899 (comment) |
Closing. Please create a proposal if you want to see this feature in future Wasm. |
As fp16 are broadly used in nowadays, and i16 is something corresponding to i16. So I suggest add these two type as native support in webassembly. Anyway, i8 seems not necessary, but i16 somewhat are necessary for fp16
The text was updated successfully, but these errors were encountered: