Information Technology Reference
In-Depth Information
x ( n )
z 1
z 1
z 1
z 1
b 0
b 1
b 2
b 3
b M
y ( n )
Figure 8.3. Block diagram of an FIR filter.
nodes then broadcast a spin wave on the bus with an amplitude equal to their data.
These values, which are sent at the same frequency, are superposed and form a
wave with amplitude equal to the summation of all. The output node's receiving
frequency is tuned on the sending frequency of all the input nodes; therefore, it
receives the result of this summation. At this stage, all the previous inputs are
shifted right by one position, the new data is inputted to the leftmost node, and the
summation process is repeated. The structure of a spin-wave FIR filter module is
shown in Figure 8.4. This structure includes M+1 input node, where M is the
number of taps.
In this implementation, the shift operation is performed by using the spin-
wave switches as explained previously. (These switches are all on in the summation
step.)
As mentioned, it is possible to use another method for implementing the shift
operation in which distinct frequencies are assigned to different nodes. Note that
in implementing the FIR filter, after each shifting operation, all the nodes' sending
and receiving frequencies must be tuned on a common frequency to perform the
summation. Therefore, in each step the nodes are required to dynamically tune
their frequencies.
Having explained some of the parallel techniques for algorithm design on
spin-wave architectures, in the next section we describe some of their parallel
routing schemes.
Inputs
Output
X n
X n-1
X n-2
X i
X n-M
y
b 0
b 1
b 2
b i
b M
Shifting step
Summation step
Spin-wave switch
Figure 8.4. Structure of a nanoscale spin-wave FIR filter.
Search WWH ::




Custom Search