Image Processing Reference
In-Depth Information
where Π is a (1× N )- D vector (composed of the first row of T ) which (in the segmenting
terms) determines the time scheduling, and the ( N - 1)× N sub-matrix Σ in (24) is composed
of the rest rows of T that determine the space processor specified by the so-called
projection vector d (Kung, 1988).Next, such segmentation (24) yields the regular PA of ( N -
1)- D specified by the mapping
TΦΚ
,
(25)
where K is composed of the new revised vector schedule (represented by the first row of
the PA) and the inter-processor communications (represented by the rest rows of the PA),
and the matrix Φ specifies the data dependencies of the parallel representation algorithm.
Hyper-planes
For n=m=4
Matrix-Vector
Processor Array
(PA)
y
y
y
y
 
Π
11
2
y
D
x
Data-Skewed
03
aaaa
000
P
a
a
a
a
D
03
23
13
33
33
23
13
03
x
D
02
T
d
10
a
a
a
a
aaaa
00
02
22
12
32
P
D
33
23
13
03
x
Mapping
transformation
01
D
a
a
a
a
01
21
11
31
aaaa
03 0
x
P
D
00
33
23
13
a
a
a
a
00
30
10
20
D
0
0
0
0
aaaa
P
D
Matrix-vector DG
33
23
13
03
For m=4
P
P
P
P
0
0
0
0
Bit-level
Array of PEs
for Processor
4
5
6
 
7
Π
12
x
0
4
P
P
0
3
d
10
x
0
2D
2D
2D
3
aaa
x
x
x
P
P
0
m
00
00 00
21
0
0
0
m
1
2
2
Mapping
transformation
x
D
D
D
0
2
P
0
1
x
0
1
a
a
a
a
00
1
00
2
00
3
00
4
Bit-level Multiply-Acumulate
DG
Fig. 2. High-Speed MPPA approach for the reconstructive matrix-vector SP operation
For a more detailed explanation of this theory, see (Kung, 1988), (CastilloAtoche et al.,
2010b). In this study, the following specifications for the matrix-vector algorithm onto PAs
Search WWH ::




Custom Search