Currently, all the count*driver functions dispatch to the function compiled with the highest instruction set (e.g., when AVX and SSE are both available, only the AVX function will be called). By using varargs, this should be implementable reasonably easily.
Currently, all the
count*driver
functions dispatch to the function compiled with the highest instruction set (e.g., whenAVX
andSSE
are both available, only theAVX
function will be called). By usingvarargs
, this should be implementable reasonably easily.