SM75 (Turing) support for FP6 kernel (#942)

* SM75 support for FP6 kernel

* More consistent argument ordering in benchmark function

* Add a note about SM75 support in the floatx README

* Handle FP6 + SM75 + N>=64 edge case

* Document changes made for FP6 SM75 support

Tobias van der Werff committed 1y ago

96e8fee0f83245689cdd0b341dfbc6adaad57bb0

Parent: 39ce823

Committed by GitHub <noreply@github.com> on 9/29/2024, 12:00:02 AM