Michael J. Mahon wrote:
In the case of the Apple II, the (somewhat unusual) layout of display memory was a consequence of using the video refresh to accomplish the refresh of the DRAMs.
I don't agree. No matter what the screen memory layout is, the video circuit will refresh all the DRAM as long as 128 consecutive addresses are accessed (or all 128 combinations of low 7 address bits are used).
The screen layout is what it is for a different reason: the number of characters in one line is not a power of 2.
Suppose you have 64 chars per line. Then the addresses can be consecutive. Same for 32 chars like in Sinclair Spectrum for example. Woz chose 40 chars per line. Memory was quite expensive. To avoid big holes - 24 of every 64 bytes - he used an adder to scramble the addresses so that there are only small 8 byte "holes".
The result is that no special refresh circuitry is used, and, since all video accesses occur on the alternate half-cycles not used by the processor, there is never any interference with the processor for video or DRAM refresh.
This is all true. But will also be true in case of say a 64 char per line display and linear memory.
Of course it is possible to have 40 (or say 80) chars per line and still have linear addressing, like in a VGA card, but that requires a lot more hardware.
-Alex.