If you load the program at address zero as you normally would, because it is only half the size of the new IC, it will only load into the bottom half of the storage space. If you have an option to set the load address, just leave it at zero as you would with the 2532.
Each time the size of a memory doubles, as it does between the 'x32' and 'xx64' it needs another address line to allow all the capacity to be reached. With A12 tied to ground, A0 to A11 will access addresses 0000 to 0FFF (as in the 2532) and with A12 tied to VCC, A0 to A11 will access addresses 1000 to 1FFF, in other words the next 4K block above the first one.
In very early days of EPROMs when they were expensive and in short supply, it wasn't uncommon to see a 2764 made by soldering two 2732 ICs, one on top of the other. All pins were soldered together except the -CE pins which were wired to an inverter and fed from the A12 pin on the socket. The idea was that A12 controlled which of the two smaller ICs was active at a time. As far as the socket was concerned it had a 2764 plugged into it.
Brian.