* test with debug config option
* rename CamelCase symbols to linux_style (?)
* test on 32bit x86 qemu
* test on non-x86 hardware
* consider hash functions with better performance on other arches.
* consider hash table with say, 8, possible matches.
* check what compression ratio, speed, memory use looks like if instead of
 hash table with 16K entries for 32K values and no chaining we use a hash map
 that stores all values (establish upper limits).