-
Notifications
You must be signed in to change notification settings - Fork 630
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
Draft of Zfh extension for IEEE 754 binary16 support #496
Conversation
Any plan for merge this? I think it's just 0.1 that mean we still have chance to change anything? otherwise it's like a hidden/secret extension in branch. When I said there is already have a spec draft for zfh/fp16 and then people told me they don't seem the spec, that happened several times, and then I need to point out this is not exist in master branch yet, you need to find it out in |
I would love to merge it, but IIRC the plan was to wait until @kasanovic wrote an email proposing it. |
Oh, OK, thanks for clarification :) |
I really hope this PR could merge and freeze soon, this is a very useful extension. |
src/zfh.tex
Outdated
half-precision operands to single-precision, performing the operation | ||
using single-precision arithmetic, then converting back to half-precision. | ||
Performing half-precision fused multiply-addition using this method incurs | ||
a 1-ulp error on some inputs. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Suggest clarifying that the "error" is only on RNE and RMM (double-rounding is innocuous for directed rounding).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@knightsifive Yeah, good point.
Signed-off-by: Chih-Min Chao <chihmin.chao@sifive.com> Co-authored-by: Chih-Min Chao <cmchao@gmail.com>
Signed-off-by: Chih-Min Chao <cmchao@gmail.com> Co-authored-by: Chih-Min Chao <cmchao@gmail.com>
This proposal follows the template of the existing FP extensions. Future extensions might also provide widening ops, but I think that's out of scope for the baseline.