Sorting also forms the bottleneck in several memory hard Proof-of-Work schemes. These differ from the general problem in two important ways:
1) the input data is generated from cryptographic hash functions, and can thus be considered random, leading to e.g. very even distribution of elements over buckets.
2) each round of sorting serves to identify groups of items that match on a subrange of their bits, and somehow transform these items.
There is a lot of cryptocurrency money to be made by developing the most optimized mining software and charging 1% or so "developer-fee" on the mining proceeds.
1) the input data is generated from cryptographic hash functions, and can thus be considered random, leading to e.g. very even distribution of elements over buckets.
2) each round of sorting serves to identify groups of items that match on a subrange of their bits, and somehow transform these items.
There is a lot of cryptocurrency money to be made by developing the most optimized mining software and charging 1% or so "developer-fee" on the mining proceeds.