pdf review
rurban opened this issue · comments
There are small typos in the paper pdf "Modern Non-Cryptographic Hash Function
and Pseudorandom Number Generator"
- Abstract:
compliers
- page 5:
compliers
- page 10:
which losses entropy
=>which loses entropy
,
not loss entropy
=>not lose entropy
- page 11: precessed => processed
- page 11:
_wyr# functions reads # byte from the key using memcpy.
=>_wyr# functions reads # bytes from the key using memcpy.
Questionable:
- page 11:
predefined 64-bit integer with 32 1bits
=>predefined 64-bit integer with 32 bits
a bit is by definition only 1 bit long. nobodys calls them 1bits. (2x) - page 11:
(Discussion)
misses a link. - page 12:
It is obvious that its cycle length is 2^64 as p0 is a large prime
. => why? maybe improve it by sayinglarge prime > 2^63
(0xa0761d6478bd642f = 11562461410679940143)
wyrand is also tested in dieharder, and turns out to be the best of all available prng functions (for HW rng's I only tested the broken AMD one, and linux /dev/random/ which is HW assisted)
Thank you very much. updated.
The "obvious" period discussion is at:
lemire/SwiftWyhash#10
#15
I also found few minor typos and stylistic imperfections. Check attached diff, you will figure it out.