In SMhasher, it seems that data (keys) are read one by one sequentialy, and not splited and:
MiB/sec: The average of the Bulk key speed test for alignments 0-7 with 262144-byte keys. The higher the better.
/\ This is written in its README.
The author said that are the output of hash functions that is measured.
Take a fast hash function tested by SMhasher, for example Blake3, it has 1288.84 MiB/S is a modern processor.
As the output is measured, 1288.84 MiB/S equals ~1351446691-bytes that means ~42232709 outputs of Blake3, each output has 32-byte (256-bits).
But if we multiply the number of Blake3 invocations per 262144-byte keys hashed sequentially as input we have this: 42232709 * 262144 = 11071051268096-bytes read as input (11.07-terabytes).
How can it be possible that the input of a hash function tested can read data at 11.07-terabytes per second?