Information Technology Reference
In-Depth Information
given by
h
sₒc
+(
l
+1)+1+
h
cₒk
+(
l
−
1). The minimum inter-arrival time
{
between consecutive flits is given by
Max
time required to shift-in a flit into
core,timerequiredtoshift-outaresponsefromcore+unitcycletounpackthe
flit + time taken to generate the response
. Therefore, entire test time for for a
particular core is obtain by
h
sₒc
+(
l
−
1)+1+[1+
Max
{
h
sₒc
+(
l
−
1)
,h
cₒk
+(
l
−
1)
}
]
×
(
p
−
1)+
h
cₒk
+(
l
−
1)
⃒
[1 +
Max
{
h
sₒc
,h
cₒk
}
+(
l
−
1)]
×
p
+[
Min
{
h
sₒc
,h
cₒk
}
+(
l
−
1)]
Now, the test schedule problem can be formulated as
Given an NoC with the test information such as core type (preemptive/non-
preemptive), number of wrapper scan-chain, number of test patterns etc. and m
number of test port (I/O pair), find an assignment of cores to test ports and its
duration in such a way that the overall test application time is minimized.
3 Related Works
An
Integer Liner Programming
(ILP) based non-preemptive test strategy has
been developed in [6]. It is computationally expensive when system has more
number of cores. A
Genetic Algorithm
(GA) based meta-search technique has
been developed for non-preemptive test schedule [7]. In that work, cores and I/O
pairs are embedded separately into chromosome structure. A non-preemptive
scheduling strategy has been developed based on another meta-search technique
named
Simulated Annealing
(SA) [8]. Different types of wrappers have been as-
signed to each core for reducing the test application time. Another SA based
technique have been presented in [9]. In this approach, dynamic routing path
has been selected to reduce the test application time. Flit size has been com-
puted using Pareto-optimal strategy. The core location has been decided using a
mapping algorithm. An ant colony based meta-search technique has been used to
minimize the test application time in NoC based SoC environment [10]. A non-
preemptive test scheduling based on
Particle Swarm Optimization
(PSO) based
meta-searchhaverepottedin[3].Inthiswork, each core tested by a clock rate, all
cores are assumed to be nonpreemptive and each particle consists of two parts—
core part and IO pair part. Many other strategies have been proposed in the
literature for multi-frequency testing [11], power and thermal-aware strategies
[12-16] and so on. However, these works mostly consider non-preemptive policy.
This paper presents a basic mixed test scheduling policy having both preemptive
and non-preemptive cores. The policy can be extended easily to-words power and
thermal-aware tests.
4 Continuous and Discrete Particle Swarm Optimization
Particle Swarm Optimization
(PSO) based evolutionary strategy has been de-
veloped in 1995 [17]. It is a population based stochastic search and optimization
technique, each particle represents a potential solution. Evolution of particles
over generation is guided by both the self and the swarm intelligence.
