Note: Chuma states, "If it matters – the original lists are sorted, they don't all contain the same words, all numbers are positive integers." For clarity, the lists are sorted by (value descending, word).

Links to file generators and llil4 variants:

gen-llil.pl, gen-long-llil.pl, shuffle.pl
llil4map, llil4map2, llil4hmap, llil4emh, llil4umap
llil4vec, llil4vec-tbb
llil4judy

April 2024 updates:

llil4map2 is a memory efficient version, using a vector of pointers, to the phmap key-value pairs. This results in sorting taking longer, possibly due to CPU cache misses. However, memory consumption is significantly less.

200mil unique keys   llil4map   llil4map2
------------------   --------   --------
fixed-length=12      ~ 8.0 GB     6.0 GB
long-strings          17.8 GB    12.3 GB
[download]

See also, a more memory efficient version, by Gregory Popovitch, author of the C++ parallel-hashmap library.