That equation is a custom hash function that someone came up with to ensure that codons hash to unique sequential values for memory efficiency. In a modern language like python there are built in hash functions, so one could instead just do:

someSequence = 'atgatg'
d = dict()
for idx in xrange(0, len(someSequence), 3):
    codon = someSequence[idx:idx+3].upper()
    if codon not in d:
        d[codon] = 0
    d[codon] += 1

for codon, cnt in d.items():
    print('{}: {}'.format(codon, cnt))

If for some reason you absolutely HAD to use this custom hashing function, then you have to use a vector of values:

valueConversion = {'T': 1, 'U': 1, 'C': 2, 'A': 3, 'G': 4}

def customHash(codon):
    P1 = valueConversion[codon[0].upper()]
    P2 = valueConversion[codon[1].upper()]
    P3 = valueConversion[codon[2].upper()]
    return ((P1 - 1)*16) + P2 + ((P3-1)*4) - 1  # Note the conversion to 0-based indexing!

someSequence = 'atgatg'
v = [0] * 64  # N.B., python uses 0-based indexing
for idx in xrange(0, len(someSequence), 3):
    codon = someSequence[idx:idx+3].upper()
    v[customHash] += 1

Now v contains the counts starting from AAA to GGG. There are probably some typos in there. Finding and correcting these errors can be an exercise for you.

As Devon says, you don't need that equation to count exons. If the fasta sequence is serialized, you can just use this perl oneliner:

echo "atgatggtagtacatcatcat" | perl -lne '$cs{$1}++ while /(...)/g;
END { foreach $c (sort keys %cs) { print uc($c).": $cs{$c}"  } }'