How does gawk allocate memory for arrays?

[Ed Morton]

A question came up recently that can be reduced to comparing the line 
numbers associated with unique key values in 2 files that are over 10G 
lines each. While unique, those values only need to be compared within 
groups of 10 contiguous lines (in reality there's other things going on 
irrelevant to this discussion).

So we had:

 * Approach 1:  while reading file1 create an array that's 10G of
   entries then while reading file2 just do a hash lookup.
 * Approach 2: while reading file1 for every 10 lines create an array
   that holds those 10 entries, then getline 10 entries from file2 into
   a second array of 10 entries, then loop through all the values in
   the file1 array to compare to the file2 array then delete both
   arrays and start over.

Obviously the 2nd approach was going to use far less memory but 
according to the OP it was also an order of magnitude faster than the 
1st approach so that got me wondering about how awk arrays are 
allocated, e.g. is there a default size that gets allocated initially 
and then new chunks of memory get allocated as needed? If so what is 
that size?  I expect I could find and read the code but I'm really 
hoping to not have to do that and the design is documented somewhere or 
someone can just tell me what it is.

    Ed.