-
Notifications
You must be signed in to change notification settings - Fork 137
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
fmaf is sometimes incorrect #160
Comments
FWIW, glibc |
For reference, this is what glibc seems to be doing, if I translated the code right (in works in my tests):
|
One idea I had is that the
|
We can't really use glibc code for licensing reasons. musl code shouldn't be a problem. |
@simonbyrne Right, sorry, I wasn't thinking about licenses when I posted that, I was just curious what the right approach was. They both seem to be based on fma emulation through rounding-to-odd, like this: Emulation of a FMA and correctly-rounded sums: |
In Julia, JuliaLang/julia#42783 fixes the software |
|
Is the openlibm function fixed? |
Huh - I thought this issue was about the Julia fma implementation, which is fixed. I see that the openlibm one is still broken. |
This example was pointed out by njuffa on scicomp.stackexchange.com. I checked it also with a recent version of Z3 and exact rational arithmetic.
The text was updated successfully, but these errors were encountered: