Explore Numba Function Inlining
seanlaw opened this issue · comments
In functions like _stump
, the sub-functions _compute_diagonal
and core._shift_insert_at_index
may be called a lot. Therefore, we may gain some (small) performance improvement by adding @njit(inline='always')
to those (and other) functions and possibly reduce the overhead cost of calling a function repeatedly. See more here.