I could be reading it wrong, but the "Quality problems" column in lines 49 through 69 appears to be shifted one row up for a few hashes: i.e. the problems noted correspond to the hash one row down. In commit https://github.com/rurban/smhasher/commit/00a4e5ab6bfb7b25bd3c7cf915f68984d4910cfd, a hash was added but the quality problems column does not appear to have been copied correctly.
The hashes affected are:
City32 vs City64: The former has no issues in the log but is noted to have 2 minor collisions, whereas the latter has no problems noted but has 2 collisions in its log (here).
Spooky128 vs xxHash32: The former has no issues in the log but is noted to have collisions with 4-bit differentials (here), whereas the latter has no problems noted but has a collision with 4-bit differentials (here).
metrohash128_2 vs metrohash64crc_1 and metrohash64crc_2: The first has no collisions in its log but is noted to have cyclic collisions with 8 byte (which is not even checked in the log, see here), whereas the second metrohash64 is noted to have no problems but also has cyclic collisions with 8 byte here.
It seems typos were introduced to the README summary table in commit https://github.com/rurban/smhasher/commit/00a4e5ab6bfb7b25bd3c7cf915f68984d4910cfd.
I could be reading it wrong, but the "Quality problems" column in lines 49 through 69 appears to be shifted one row up for a few hashes: i.e. the problems noted correspond to the hash one row down. In commit https://github.com/rurban/smhasher/commit/00a4e5ab6bfb7b25bd3c7cf915f68984d4910cfd, a hash was added but the quality problems column does not appear to have been copied correctly.
The hashes affected are:
City32
vsCity64
: The former has no issues in the log but is noted to have 2 minor collisions, whereas the latter has no problems noted but has 2 collisions in its log (here).Spooky128
vsxxHash32
: The former has no issues in the log but is noted to have collisions with 4-bit differentials (here), whereas the latter has no problems noted but has a collision with 4-bit differentials (here).metrohash128_2
vsmetrohash64crc_1
andmetrohash64crc_2
: The first has no collisions in its log but is noted to have cyclic collisions with 8 byte (which is not even checked in the log, see here), whereas the second metrohash64 is noted to have no problems but also has cyclic collisions with 8 byte here.