This PR refactors the implementation of the nrm2 operator.
These changes are based on the work done for asum operator which was merged with PR #458. Now nrm2 implementation shares the same structure of asum adjusting the execution call to its needs.
Update for nrm2 each architecture backend file to manage the different kernel sizes and these numbers are set empirically using available hardware.
This PR refactors the implementation of the
nrm2operator.These changes are based on the work done for
asumoperator which was merged with PR #458. Nownrm2implementation shares the same structure ofasumadjusting the execution call to its needs. Update fornrm2each architecture backend file to manage the different kernel sizes and these numbers are set empirically using available hardware.