But what’s the constant overhead? This hurts us with SipHash; even though it claims to be speed competitive, it has high constant overhead…
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Off-topic -- I thought modern processors just use an opcode that reads quantum effects within the die these days. Am I way off?
-
RDRAND has reciprocal throughput of like 320 cycles per Agner’s tables. xorshift easily beats it.
- 1 more reply
New conversation -
-
-
Similar: web framework benchmarks that penalize rails for: - signed cookies by default - verifying signed cookies by default - bcrypt by default (constant time hashing) It takes a bit of time to do those things, but I don't mind paying 1ms for security by default.
-
Oh and btw, 1ms of overhead means you can only ever get 1000 requests per second if everything else takes 0 time. Which makes frameworks like rails look "ridiculously slow" because of the fallacy of measuring "requests per second" rather than "acceptable overhead"
- 2 more replies
New conversation -
-
-
In some (many?) languages, rand() is explicitly not a csPRNG and can't be. For example requirement of reproducible sequences from a seed.
-
csPRNGs are seedable...
- 5 more replies
New conversation -
-
-
This has come up a number of times on the
#golang mailing list. People also forget that Go's default random source (which is slow to seed) has a mutex which hurts all the threaded benchmarks. Not only does it benchmark rand(), also lock contention :/ -
Almost without exclusion “why does this code not scale with extra cores” bug reports are because the reporter is benchmarking the RNG.
- 1 more reply
New conversation -
-
-
Good languages provide "fastrand()" and "goodrand()" and gives neither a shorter/more convenient name.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.