v2.1

fix compiler warnings

v2.0

updated with hackathon improvements
-- less syncthreads
-- unrolling k
-- decided to leave m/n blocking in

v1.4.6a

try 0 indexing inside kgemm loops to reduce integer unit usage

v1.4.6

fix cpu version

v1.4.5

fix error

v1.4.4

remove unused asserts

v1.4.3

trim locals, remove unusued variables

v1.4.2

trim unused variables

v1.4.1

trim unused variables

v1.4

remove blocking over m and n to reduce integer instructions. performance increases by 20% without blocking on 6d -l 4 -d 4.