Use hash_hasher in hashmaps/sets where the key is a Cid/Multihash #258

[CHr15F0x]

This PR implements a fix for #258, ie. using hash_hasher to improve performance in HashMaps/Sets what utilize Cid as key. hash_hasher provides approx. 40% increase in performance in this case.

There are two things that still need to be done:

• code is not linted yet, since this will be done in a separate PR by Mirko,
• a simple benchmark is included for the sake of comparing HashedMap<Cid,> vs HashMap<Cid,>; Imho this bench should not be merged.

[koivunej]

Looking good!

[CHr15F0x]

Comparison using fetch_and_cat on a 10GiB file shows some slight improvement:

Benchmark #1: ./opt-fetch_and_cat QmTR3tCt1SRd96t7rSZpvMPbj48GijPz8crEuuujSygEC1 /ip4/127.0.0.1/tcp/33109/p2p/QmTjjZBvf9fZxUcSqqdXPy135zWwCisGpswAF38jvQ8Y9G > /dev/null
  Time (mean ± σ):     58.706 s ±  0.407 s    [User: 46.545 s, System: 24.116 s]
  Range (min … max):   57.808 s … 59.624 s    100 runs

Benchmark #1: ./fetch_and_cat     QmTR3tCt1SRd96t7rSZpvMPbj48GijPz8crEuuujSygEC1 /ip4/127.0.0.1/tcp/33109/p2p/QmTjjZBvf9fZxUcSqqdXPy135zWwCisGpswAF38jvQ8Y9G > /dev/null
  Time (mean ± σ):     58.838 s ±  0.425 s    [User: 46.751 s, System: 24.169 s]
  Range (min … max):   57.927 s … 59.728 s    100 runs

[koivunej]

This needs to have the Cargo.lock removed, sorry for not seeing that. After that, it's good to go. Thanks for the benchmarking.

bors d+

[bors]

✌️ CHr15F0x can now approve this pull request. To approve and merge a pull request, simply reply with bors r+. More detailed instructions are available here.

[koivunej]

bors d-

Could you add a CHANGELOG.md entry?

[CHr15F0x]

Changelog is there, let's try with Mr b

bors r+

[bors]

Merge conflict.

[CHr15F0x]

bors r+

[koivunej]

bors r-

[bors]

Canceled.

[koivunej]

@CHr15F0x could you avoid the merge commit, rebase on master instead?

[CHr15F0x]

bors r+

[bors]

Build succeeded:

• ci