Suppose we have an array **x** with *m* rows and *n* columns and a vector s with *m* elements. I would like to sum the elements of each row of array **x** and save these results in vector **s**.

This can be easily implemented in Julia using `v = sum(x, dims = 2)`

. However, this procedure allocates extra memory for the operation. Is it any way to do this operation without extra allocation (and not using cycles)? I was trying function `sum!`

, but it does not accept keyword argument `dims`

.