Hi Joe,

the table method, was the fastest, but most code size eaten.

You can be a little slower, but many times smaller, if you use a table for every nibble (only 16 byte table size):

;input: a = byte
;output: a = bit count

count_bits_fast:
mov r0, a
anl a, #0Fh
add a, #cbf3 - cbf1
movc a, @a+pc
cbf1:
xch a, r0
swap a
anl a, #0Fh
add a, #cbf3 - cbf2
movc a, @a+pc
cbf2:
add a, r0
ret
cbf3:
db 0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4


It's four times faster, than your and Eriks code


Peter