About the default rounding mode

[贾迅]

Hi, everyone！

    I am using the cblas_dgemm from OpenBLAS for dense matrix multiplication. And I find the least significant bits of the computation result is different from another implementation based on fused-multiply-add instruction and rounding mode set to "Rounding to Nearest".

    So, I am wondering whether the multiply and add in clas_dgemm is mapped to fused-multiply-add and what is the default rounding mode in OpenBLAS?  

[Zhang Xianyi]

Hi,

We didn't control rounding mode. It is by default.

Which CPU do you test?

Xianyi

--
You received this message because you are subscribed to the Google Groups "OpenBLAS-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openblas-users+unsubscribe@googlegroups.com.
To post to this group, send email to openblas-users@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.