adaptive block size for QR used in MAGMA's version 2.0?

[nahla]

Hello =)

I have a short question to the blocking of MAGMA's dgeqrf method:
Is the adaptive block size algorithm proposed in "Adaptive Block Size for Dense QR Factorization in Hybrid CPU-GPU Systems via Statistical Modeling" (https://www.researchgate.net/publicatio ... l_Modeling) implemented in the current MAGMA version 2.0 or does MAGMA use a fixed block size?

Thanks a lot for your help,
nahla

[mgates3]

MAGMA QR uses a single block size.
-mark

[nahla]

Thanks a lot! =)

nahla