Do you mean a 40x40 i.e. 1600 keys matrix? First expand your I/O to 80 lines!!
Not necessary to have so many I/O. Not even for a 40X40 Keypad. I can do this with 8+10=18 lines. How?

On the rows I will use 5 74HC573 latches. with data input from port 0 and each having latch enable from 5 port pins of some other port say P2.0 to P2.4.

On the columns I will use 5 74HC573 buffers. With data inputs from port 0 and each having Gate enable from 5 port pins of some other port say P2.5,P2.6,P2.7,P1.0,P1.1.

That makes 40X40 keypad using only 18 I/O lines and 10 Logic chips. If you use memory mapped I/O than it is fairly possible to interface this keypad using just 8 I/O lines of Port 0. However number of logic chips will increase in that case.

Regards,
Prahlad Purohit