The way these things work is that there are several chips that control a single row. My panel has 64 pixels in a row. The chips have 16 outputs, so we need 4 of them for 64 pixels. We have 3 colors so we need 3*4 = 15 of these chips for a single row.
These chips are chained one to another (4 of them are) to form a 64 bit shift register. A clock tells them when to shift in a new bit. A rising edge on clock, shifts in the new pixel data. The clock can be running all the time, you just have to decide when to start feeding in data. You count to 64, then you pulse the latch signal and that transfers all 64 bits from the shift register into a latch that actually drives the LEDs. The shift register keeps on shifting away (assuming the clock is continually running), but the latched data is "captured".
There are another set of chips that select which row is actually lit up. These chips are driven by the 5 address bits. Note that if you have a pattern latched by the row drivers (as described in the paragraph above), you can use the address bits to control which row it gets displayed on. If you wanted to, you could then loop through all the addresses and it would show up on all the rows.
This pretty much explains everything, except the OE (output enable) signal. This can be used to turn off the display entirely. The usual thing is to shift in a rows worth of data, then use this signal to blank the display. Then latch the data (which would be displayed to the wrong row if you didn't blank the display) and set the new row address. Then unblank the display to show the new row. This can be done fast enough to not be noticed. Remember that each row on the display is only actually lit for 1/32 of the time, so this has essentially no impact on the display.
Tom's Electronics pages / tom@mmto.org