Systolic Arrays for Matrix Transpose and Other Reorderings
Dianne Prost O’Leary · IEEE Transactions on Computers · 1987
In this correspondence, a systolic array is described for computing the transpose of an n × n matrix in time 3n - 1 using n2switching processors and n2bit buffers. A one-dimensional implementation is also described. Arrays are also given to take a matrix in by rows and put it out by diagonals, and vice versa.