for input, I just have it read them all at once and stick the results in an array. in NESASM:

JOY_A = $00
JOY_B = $01
JOY_SEL = $02
JOY_STA = $03
JOY_UP = $04
JOY_DOWN = $05
JOY_LEFT = $06
JOY_RIGHT = $07
DELTA_A = $08
DELTA_B = $09
DELTA_SEL = $0A
DELTA_STA = $0B
DELTA_UP = $0C
DELTA_DOWN = $0D
DELTA_LEFT = $0E
DELTA_RIGHT = $0F
; or wherever you prefer to put them.
; JOY_* = whether the button is pressed or not ($01 = pressed)
; DELTA_* = whether the button's status has changed since the last read ($01 = changed)

read_joy:
ldx #$01
stx $4016
dex
stx $4016
readjoy_2:
lda JOY_A,x
sta DELTA_A,x
lda $4017
sta JOY_A,x
eor DELTA_A,x
sta DELTA_A,x
inx
cpx #$08
bne readjoy_2
rts

----
This takes advantage of the fact that the status variables are arranged in the same order as $4017 returns them. The 'delta' part can be used to respond to a button press only once, for example, if you had a game menu, pressing a button would result in instantaneous, repeated selections while the button was pressed (by simply checking to see if it's pressed or not). With 'delta', you would only respond to a button if it is pressed AND its status has changed.
lda JOY_A
and DELTA_A
bne respond_to_input
do_not_respond:
...

Anyhow, just thought I'd throw that out there.