TABLe 8.2
Example of Design_wrapper Algorithm
1
Wrapper Scan Chains
2
3
4
Internal scan chains
Wrapper input cells
Wrapper output cells
Scan-in length
Scan-out length
12 + 6
2
3
20
21
12 + 6
2
3
20
21
8 + 6 + 6
0
0
20
20
8 + 8
4
5
20
21
Testing of Network-on-Chip Architectures
249
Next, the time needed to test a core is computed. For this, it is assumed that
the core has p test patterns, the maximum length of scan chain (corresponding to the number of flits per packet) is l, and the numbers of hops from
source to core and core to sink are h src→core and h core→sink , respectively. It is
further assumed that the cores are scheduled for testing using a nonpreemptive scheduling strategy. The scheduler assigns each and every core a fixed
path during the scheduled test time. All resources of this path are reserved
throughout the test time of the core. Test data are moved along the reserved
path in a pipelined fashion. Since each flit is unpacked in one cycle, the time
required for a test vector to be sent to the core plus the test response to be
received at the sink is be given by
h s rc →core + (l − 1) + 1 + h c ore →sink + (l − 1)
That is, first the l flits corresponding to a test pattern are sent in a pipelined
fashion from the source to the core. In the next cycle, the pattern is applied
to the core. Next, the response bits are transferred to the test sink in a pipelined fashion. During testing, test stimulus and responses move in a pipelined
fashion with an overlap in the shifting-in of the current test pattern and shifting-out of the previous response. For a set of p patterns, the test time needed
consists of the following:
1. Time to shift in the first pattern, h src→core + (l − 1)
2. Time to apply the first pattern, 1
3. Time for overlapped transmission of the test bits of the current pattern and the response bits of the previous pattern plus the test application time for the current pattern, for next (p – 1) packets, which is
given by
+
h
[1 Max { src →core + (l −1), h core →sink + (l −1)}] × (p −1)
4. Time required to shift out the response of pattern p, h core→sink +
(l − 1).
Example of Design_wrapper Algorithm
1
Wrapper Scan Chains
2
3
4
Internal scan chains
Wrapper input cells
Wrapper output cells
Scan-in length
Scan-out length
12 + 6
2
3
20
21
12 + 6
2
3
20
21
8 + 6 + 6
0
0
20
20
8 + 8
4
5
20
21
Testing of Network-on-Chip Architectures
249
Next, the time needed to test a core is computed. For this, it is assumed that
the core has p test patterns, the maximum length of scan chain (corresponding to the number of flits per packet) is l, and the numbers of hops from
source to core and core to sink are h src→core and h core→sink , respectively. It is
further assumed that the cores are scheduled for testing using a nonpreemptive scheduling strategy. The scheduler assigns each and every core a fixed
path during the scheduled test time. All resources of this path are reserved
throughout the test time of the core. Test data are moved along the reserved
path in a pipelined fashion. Since each flit is unpacked in one cycle, the time
required for a test vector to be sent to the core plus the test response to be
received at the sink is be given by
h s rc →core + (l − 1) + 1 + h c ore →sink + (l − 1)
That is, first the l flits corresponding to a test pattern are sent in a pipelined
fashion from the source to the core. In the next cycle, the pattern is applied
to the core. Next, the response bits are transferred to the test sink in a pipelined fashion. During testing, test stimulus and responses move in a pipelined
fashion with an overlap in the shifting-in of the current test pattern and shifting-out of the previous response. For a set of p patterns, the test time needed
consists of the following:
1. Time to shift in the first pattern, h src→core + (l − 1)
2. Time to apply the first pattern, 1
3. Time for overlapped transmission of the test bits of the current pattern and the response bits of the previous pattern plus the test application time for the current pattern, for next (p – 1) packets, which is
given by
+
h
[1 Max { src →core + (l −1), h core →sink + (l −1)}] × (p −1)
4. Time required to shift out the response of pattern p, h core→sink +
(l − 1).