Good quality Random number generator implementation

[Tommi Prami 157]

Could someone point into nice implementation?

Would be nice (but not 100% necessary) that would have similar functionality than RTL version have. More than less drop in replacement (Easy to port)

 

-Tee-

[David Heffernan 2456]

What properties are you looking for. The reason that there are many different algorithms in use is that different applications have different requirements. 

[Tommi Prami 157]

For now, I am open to all suggestions 🙂 Just need to get random numbers and have some Randomize-functionality. 

As long it is way better tna build in RTL version. If can use it to replace RTL version in code, way or other, I am happy 🙂

 

-Tee-

[David Heffernan 2456]

I don't get this. If you don't have any particular needs, then I don't see how anybody can offer suggestions. And if all you want is a list of libraries, then websearch will get you that.

 

Do you know what you are going to do with these random numbers? 

[aehimself 404]

Both XOrShift and ISAAC has nice Delphi implementations. They are both PRNGs but it you initialize them with the Random function, they will work quite well.

[Tommi Prami 157]

3 hours ago, David Heffernan said:

I don't get this. If you don't have any particular needs, then I don't see how anybody can offer suggestions. And if all you want is a list of libraries, then websearch will get you that.

 

Do you know what you are going to do with these random numbers? 

I just need random numbers that have better quality than stock Random() function in the RTL.

I think that is valid requirement.

 

-Tee-

[Lars Fosdal 1866]

Don't know about good PRNGs, but get your seeds here. 
https://www.random.org/ 

 

[Fr0sT.Brutal 904]

Any crypto library should have strong PRNG

[Anders Melander 2051]

35 minutes ago, Tommi Prami said:

I just need random numbers that have better quality than stock Random() function in the RTL.
I think that is valid requirement.

Yes it is, but if you can't define what "better quality" is then your request is meaningless.

 

In any case, like David said, a web search should give you plenty of candidates. Did you try that first?

[KodeZwerg 54]

How about BCryptGenRandom from Windows Api? Is that random/quality enough for your needs @Tommi Prami ?

[Stefan Glienke 2148]

https://xkcd.com/221/

[Lars Fosdal 1866]

@KodeZwerg
That's not portable, though - and the quality of the PRNG would depend a lot on the parameters to BCryptOpenAlgorithmProvider.

[Wagner Landgraf 46]

7 hours ago, Tommi Prami said:

I just need random numbers that have better quality than stock Random() function in the RTL.

I think that is valid requirement.

 

-Tee-

What is the problem about the current one? What makes it "low quality"?

[aehimself 404]

program Project2;

{$APPTYPE CONSOLE}

{$R *.res}

uses
  System.SysUtils;

Const
 MAX_NUMBER = 1000;
 CYCLECOUNT = Integer.MaxValue;

Var
 numbers: Array[0.. MAX_NUMBER - 1] Of Integer;
 a, min, max: Integer;

begin
 Randomize;

 For a := Low(numbers) to High(numbers) Do
  numbers[a] := 0;

 For a := 1 To CYCLECOUNT Do
  Inc(numbers[Random(MAX_NUMBER)]);

 min := CYCLECOUNT;
 max := 0;

 For a := Low(numbers) To High(numbers) Do
  Begin
   If numbers[a] < min Then min := numbers[a]
     Else
   If numbers[a] > max Then max := numbers[a];
  End;

 WriteLn('After ' + CYCLECOUNT.ToString + ' cycle(s), minimum amount of occurrences of the same number is ' + min.ToString + ', maximum is ' + max.ToString);
 ReadLn;
end.

After 2147483647 cycle(s), minimum amount of occurrences of the same number is 2144157, maximum is 2150811

That means, the luckiest random number was hit 6654 times more than the least lucky one (from more than 2 billion attempts!).

 

Delphi's internal random number generator has a healthy dispersion. It's not perfect, but it is more than enough for average use.

Edited April 28, 2021 by aehimself

[julkas 12]

On 4/28/2021 at 10:28 AM, Tommi Prami said:

Could someone point into nice implementation?

Would be nice (but not 100% necessary) that would have similar functionality than RTL version have. More than less drop in replacement (Easy to port)

 

-Tee-

OK, it depends. Anyway, I suggest - xoshiro / xoroshiro generators

Or 🙂 start with simple SplitMix64

Efficiently Generating a Number in a Range

Edited May 1, 2021 by julkas

[Guest]

If you have €1800/yr you can go /n software (IpWorks + SecureBlackbox).

No sources 😞

[Tommi Prami 157]

OK,

 

Let me try again, any reasonably fast, Free and OpenSource implementation of good Random number generator algorithm, that is tested statistically to give better randomness than Stock RTL one.. 

I have no knowledge on which algorithm is best for what, that is why I am asking. I have an impression that Mersenne twisters are not suggested anymore. As mentioned above xoshiro / xoroshiro generators I have heard are good. But again, have no idea how to choose better one. I would not ask if I had all needed info 🙂

 

-Tee-

Edited May 3, 2021 by Tommi Prami

[Mike James 1]

There's always...

 

https://software.intel.com/content/www/us/en/develop/articles/intel-digital-random-number-generator-drng-software-implementation-guide.html

 

Regards, Mike.

[Anders Melander 2051]

Since we're now in full bike-shed mode here's my contribution: https://hackaday.io/project/4628-nuclear-random-number-generator

A random number generator based on radioactive decay.

[Stefan Glienke 2148]

pcg32 should be easy enough to implement from the c code.

[aehimself 404]

1 hour ago, Anders Melander said:

Since we're now in full bike-shed mode here's my contribution: https://hackaday.io/project/4628-nuclear-random-number-generator

A random number generator based on radioactive decay.

Holy crap this is awesome! Imagine shrinking this to a hardware key 😄

[Anders Melander 2051]

19 minutes ago, aehimself said:

Holy crap this is awesome! Imagine shrinking this to a hardware key 😄

Just as random, almost as cool, and has a slightly smaller risk of invoking a visit from Homeland Security: https://www.orau.org/ptp/collection/consumer products/dudice.htm

A dice made of depleted uranium 🙂

 

Anyway, it is possible to get USB-sized true randomness, just without the coolness factor. For example: https://www.amazon.com/TrueRNG-V3-Hardware-Random-Generator/dp/B01KR2JHTA/ref=pd_sbs_1

[Bob Devine 13]

Some that look interesting - I haven't investigated these in detail so can't give any opinions on them. The CodeProject ones would need to be ported.

 

https://github.com/LUXOPHIA/Random

 

https://www.codeproject.com/Articles/1083372/Random-Number-Generator-Recommendations-for-Applic

 

https://www.codeproject.com/Articles/1190459/Randomization-and-Sampling-Methods

 

https://www.codeproject.com/Articles/5272482/Partially-Sampled-Random-Numbers-for-Accurate-Samp

 

[David Heffernan 2456]

14 hours ago, Tommi Prami said:

OK,

 

Let me try again, any reasonably fast, Free and OpenSource implementation of good Random number generator algorithm, that is tested statistically to give better randomness than Stock RTL one.. 

I have no knowledge on which algorithm is best for what, that is why I am asking. I have an impression that Mersenne twisters are not suggested anymore. As mentioned above xoshiro / xoroshiro generators I have heard are good. But again, have no idea how to choose better one. I would not ask if I had all needed info 🙂

 

-Tee-

You have already had numerous suggestions that meet all your requirements. 

[Tommi Prami 157]

On 5/3/2021 at 5:46 PM, Anders Melander said:

Since we're now in full bike-shed mode here's my contribution: https://hackaday.io/project/4628-nuclear-random-number-generator

A random number generator based on radioactive decay.

You got good Delphi implementation of that 😄

Edited May 4, 2021 by Tommi Prami