To produce an output stream for serial shift reg stuff, or mux applications, I did this trick:
I write the desired Pattern into a fixed location of ext RAM in the main application. Then in a timer interrupt, I do sequential READ cycles on this part of the RAM after enabling the ext. logic, to capture the 8 data bits into shift regs or using its information directly in a latch. This is very efficient, since all you have to do is looping movx a, @dptr,inc dptr (or to make it even more compact, inc dpl)..
Of course you have to delay the RD Signal which loads it in the shift regs.

Werner