Models for Power-Aware Testing - Models in Hardware Testing - page 195

Fig. 7.9 The Search-A-Seed algorithm

Function Return_#patterns (seed, target_FC)
begin
FC = 0 ; #patterns = 0 ;
while ( FC < target_FC )

{

FC = generate_one_pattern (seed) ;

#patterns ++ ;

}
return #patterns ;
end

Fig. 7.10 The Return #patterns function

and the targeted fault coverage. The output is a seed of the LFSR that leads to a
pseudo-random test sequence with the shortest length required to achieve this fault
coverage.
We denote BestCost the global cost function to be minimized, which repre-
sents the smallest number of patterns in a test sequence required to achieve the
targeted fault coverage, and which corresponds to the minimum energy solution.
The function new seed() is a trivial function that determine a new seed randomly.
The function Return #patterns() (Fig. 7.10 ) generates a pseudo-random sequence
of input test patterns from a given seed, and returns the number of patterns in the
test sequence that is required to achieve the targeted fault coverage (this is done by
fault simulation). We denote the variation of the objective function. This variation
is used to determine whether a current solution is advantageous or not. If advanta-
geous, the solution is definitively accepted as local minimum. Otherwise, it may
only be accepted if the acceptance probability is greater than a given threshold.