Re: Good hash for pointers

Am 26.05.2024 um 19:39 schrieb Bonita Montero:

Am 26.05.2024 um 19:20 schrieb Tim Rentsch:
 

I say the output quality is poor because I have run tests that
show the poor output quality.

 If you chose a prime whose double is beyond 64 bit there's an
equal distribution among the 64 bit modulos.
 

I've done that with a prime of my own choosing and also with
18446744073709551557, the value you suggested.

 Show me your code.
  

This is the bucket's loads of my hash-algorithm with a number which
is a multiple of eight, i.e. uint64_t-aligned data:
#include <iostream>
#include <vector>
using namespace std;
int main()
{
vector<size_t> buckets( (1ull << 30) / sizeof(size_t) );
for( size_t b = buckets.size(); b; )
++buckets[(size_t)(--b * 8u * 0xFFFFFFFFFFFFFFc5u) % buckets.size()];
vector<size_t> loads;
for( size_t f : buckets )
{
if( loads.size() <= f )
loads.resize( f + 1 );
++loads[f];
}
for( size_t l = 0; size_t ld : loads )
cout << ++l << ": " << ld << endl;
}
1: 117440512
2: 0
3: 0
4: 0
5: 0
6: 0
7: 0
8: 0
9: 16777216