Purpose
    Estimate the dimension of the panels that will fit into a given level of the cache hierarchy following the
    principles in the paper "Analytical modeling is enough for enough for high performance BLIS" by 
    T. M. Low et al, 2016

    NL:   Number of sets
    CL:   Bytes per line
    WL:   Associativity degree
    SL:   Cache size

    Note: CLn = SLn / (WLn * NLn)

  Rule of thumb: subtract 1 line from WL (associativity), which is dedicated to the
  operand which does not reside in the cache, and distribute the rest between the two 
  other operands proportionaly to the ratio n/m   
  For example, with the conventional algorithm B3A2B1C0 and the L1 cache, 
  1 line is dedicated to Cr (non-resident in cache) while the remaining lines are distributed
  between Ar (mr x kc) and Br (kc x nr) proportionally to the ratio nr/mr to estimate kc