-
Notifications
You must be signed in to change notification settings - Fork 1.5k
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
xNRM2 failures with INCX=0 #5110
Comments
Mixing these is currently not expected to work, as Reference-BLAS has changed the implementation of NRM2 and OpenBLAS' kernels are still using the old one. (#4188 (comment)) |
The main difference in the Ref-BLAS is the scaling in case of numerically critical situations. the mini example does not touch this, so this is not covered by the different implementation ideas. |
Thought OpenBLAS already had an utest for that, but apparently not... |
As far as I can tell from the calculated "true" values in cblat1.f/zblat1.f, the reported failures for INCX != 0 appear to be due to some mishandling of "infinity" results in the test itself - which also explains why their printed test ratio (TRAT) is given as NaN |
I am running the BLAS test from the reference implementation but linked against OpenBLAS, then
SNRM2
andDNRM2
fail withINCX=0
andSCNRM2
andDZNRM2
even fail in cases whereINCX!=0
.Even a minimal example:
returns 0 instead 2.
The text was updated successfully, but these errors were encountered: