How to optimize?

[filia&sofia]

Hello, I am using bitwise operations in my data compression algorithm.
Basically, I have a buffer of n bytes from which the algorithm reads m
bits (m<=n). If m doesn't divide n, there will be some bits that need
to be transferred from end to start, because all bits must be used.
Now, the problem is that my algorithm uses many bit operations to do
this. I wish to do this as fast as possible.

Example.
n=2, m=3 and the data to the buffer from a file:

01001101 00101001 01111011 01110010 01110010

1. Fill the buffer B. B={01001101 00101001}
2. Read m bits for x times. After x=5, buffer will contain B={1} (of
course, the length of the buffer is still two bytes).
3. Now, we have to fill the buffer again. The actual problem arises
when the last bit needs to be moved to the start. If the buffer is
large, the algorithm will sequentially put one bit into a new byte
from previous byte and seven bits from the next byte. That is, B={1
0111101} and again one bit has to be moved to the next byte...

I would like to avoid this. Any suggestions? Design patterns?