Cascaded shiftregisters VS separate slave-slelect

[suku]

I'm designing a circuit with simple I/O extension using 74 series shiftregisters connected to the micro via SPI. Is it better to cascade them and pull all the data in at once, or to give them all their separate slave select (GATE) signals and only shift in the particular data that I'm interested in?
(I know both of them have advantages and disadvantages, I guess my question is which technique is more common?)

[nctnico]

I always cascade them because slave select lines tend to be long and thus inconvenient to route on a board.

[David Hess]

Which way is better just depends on layout and design considerations.  If you have a lot of shift registers in series, then eventually you can exceed the fanout of the clock, chip select, and latch signals.  You may want to read separate shift register strings in parallel for extra speed.  Or maybe you want a separate set of shift registers for galvanically isolation, low noise, or time critical I/O.

[Ian.M]

Daisy-chain (cascade) them in meaningful groups.  I am not aware of any 74xx series PISO shift registers with tristatable or open collector/drain shift out pins so you will need extra logic to parallel them.