Integer can be hashed rapidly as well
#58440
Conversation
| @@ -209,7 +275,7 @@ end | |||
| else | |||
| pos = 1 | |||
| i = buflen | |||
| while i ≥ 48 | |||
| if i > 48 | |||
There was a problem hiding this comment.
this was just an unrelated existing typo
| @@ -182,7 +182,7 @@ isinteger(::AbstractIrrational) = false | |||
| iszero(::AbstractIrrational) = false | |||
| isone(::AbstractIrrational) = false | |||
|
|
|||
| hash(x::Irrational, h::UInt) = 3*objectid(x) - h | |||
| hash(x::Irrational, h::UInt) = 3h - objectid(x) | |||
There was a problem hiding this comment.
also unrelated to the meat of this PR, but makes hash(::Irrational) more consistent with hash(::Any
|
@nanosoldier |
|
The package evaluation job you requested has completed - possible new issues were detected. Report summary❗ Packages that crashed1 packages crashed only on the current version.
5 packages crashed on the previous version too. ✖ Packages that failed24 packages failed only on the current version.
1444 packages failed on the previous version too. ✔ Packages that passed tests21 packages passed tests only on the current version.
5221 packages passed tests on the previous version too. ~ Packages that at least loaded4 packages successfully loaded only on the current version.
2925 packages successfully loaded on the previous version too. ➖ Packages that were skipped altogether921 packages were skipped on the previous version too. |
|
the |
as noted in #58388 (comment) , using
hash_bytesforBigIntlimbs in isolation will break some hashing invariants. accordingly this PR updates also thehash_integerfallback to use rapidhashsome surgery was needed to make
BigFloatandRationaland such still match, but I think (hope?) I got it allbelow are some benchmarks. the
yaxis is nanoseconds and thexaxis is an input of size1234^x, solength = 10means inputhash(big(1234^10))hashing
BigFloatgot a small bit slower, but this already allocates and already seems less common than hashingBigInt(whose hashing remains non-allocating)