updated with Munair's quicksort

cool to finally see a mergesort here

nice! with the help of a video by inigo quilez "iq", and my own skills, i made an optimized sphere hit. but to use it, you'll need normalize declare property norm as d3d '' or change to function .. property d3d.norm() as d3d var s = 1 / sqr( x * x + y * y + z * z ) return d3d( s * x , s * ...

the range thing. with a 64 bit state i guess the bias is ike a needle in a haystack. for a speedy range, you've found a highly unexpected (since mod relates with division) optimization [update] haven't debugged yet with yours (have 2 tests running) Dim Shared As Ulongint x = 12, w = 0 Dim As String ...

finally got practRand working. Did a Windows restore and installed vc redist 2013. my new generator works well.

I put hi first in a lot of range-based what-have-you's b/c more often than not, lo is 0

your test code, i gather from brief glances at various threads, is a good short-term indicator of randomness.

these silly RNGs keep distracting me from my new ray tracer :-)

i'm testing the mod method on a ubyte LCG. just need "a few hours" debugging my test. i kind of already know the result but .. gotta code it ... Boom. function LCG as ubyte: const mul = 23, add = 3 static as ushort state: state = mul * state + add return state shr 8 End Function function r...

also, you could get rid of the hi (+ 1) if you change params to double

deltarho[1859] wrote:Perhaps we need to write our own Mod function if it is fast enough.

:-)

i'm having a blast with this RNG stuff

theory vs. practice. are they frozen or are they moving? (dont say yes :P)

range using modulus may fit unevenly into 2^64. the remainder goes to the least significant digits

imo you should just show them something.

acorns

always looking for ways to speed up my non-critical applications :-)

' Complex Sequence Generator 3 by dafhi - 2018 Nov 13 ' high period, high configurability PRNG ' breezes PractRand to 32TB type CSG3_LITERAL as ubyte 'ulongint dim shared as CSG3_LITERAL csg3_mul = 255 ' non-ideal (32TB: 1645) dim shared as CSG3_LITERAL csg3_add = 1 ' 2^bits co-prime? (&H5851F4...

Int() is notoriously slow. even gcc -O 3 doesn't optimize it out

use this instead

those clouds.

[update] totally forgot. happy birthday :-)