Do you think I could just leave this part blank and it'd be okay? We're just going to replace the whole thing with a header image anyway, right?
You are not logged in.
Here is the summary:
• Mersenne Twister is very old
• xorshift is the outdated version of xoshiro/xoroshiro
• xoshiro/xoroshiro is the competitor of PCG
• PCG is the competitor of xoshiro/xoroshiro
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
Offline
Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
even if it takes 1000 years for each random number it's fine?
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
Gosha wrote:Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
even if it takes 1000 years for each random number it's fine?
What lol
Just use random.org
Offline
PiotrGrochowski wrote:Gosha wrote:Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
even if it takes 1000 years for each random number it's fine?
What lol
Just use random.org
The problem is, it isn't FAST like prng are. xoshiro256** generates random numbers in sub-ns speed.
But, most importantly, PCG is better than xoshiro256**, while xoshiro256** is better than PCG.
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
TaskManager wrote:PiotrGrochowski wrote:Gosha wrote:Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
even if it takes 1000 years for each random number it's fine?
What lol
Just use random.orgThe problem is, it isn't FAST like prng are. xoshiro256** generates random numbers in sub-ns speed.
But, most importantly, PCG is better than xoshiro256**, while xoshiro256** is better than PCG.
So speed is your criteria for an RNG that could be considered best?
Offline
Just use whatever's easiest in whatever language/environment. If it ever becomes a problem, then look into other options.
"Sometimes failing a leap of faith is better than inching forward"
- ShinsukeIto
Offline
Bruh, random.org has unnoticeable speed, you could just use their RNG for near-instant speed.
Offline
Bruh, random.org has unnoticeable speed, you could just use their RNG for near-instant speed.
Quota that maxes out at 1000000 bits and regenerates 200000 bits a day, FYI. xoshiro256** is an algorithm so it doesn't have this limitation.
But, most importantly, PCG is better than xoshiro256**, while xoshiro256** is better than PCG.
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
Moved to Bots and Programming.
One bot to rule them all, one bot to find them. One bot to bring them all... and with this cliché blind them.
Offline
Moved to Bots and Programming.
Is this a joke or what? The PCG vs xoshiro256** competition has nothing to do with EE.
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
Why are you so hung up on PCG vs xoshiro256** specifically? According to the articles you linked, they both suffer from many of the same flaws. If those flaws are acceptable, xoshiro256** is 3x faster. Although 3 * 0.95 nanoseconds is still tiny. And it seems like all they seem to be saying in those articles is that PCG and xoshiro256** aren't CSPRNGs, which wasn't a goal of theirs anyway. I don't see any complaints about either being statistically not-random so they're probably fine for most use cases?
If you don't care, use your language/environment default for ease of use.
If you need speed, either honestly seems fine. xoshiro256** is faster.
If you need quality, find a cryptographically secure PRNG.
"Sometimes failing a leap of faith is better than inching forward"
- ShinsukeIto
Offline
Also, moving on back to Off Topic. This is programming-related, but not EE Programming which is what Bots and Programming is for.
"Sometimes failing a leap of faith is better than inching forward"
- ShinsukeIto
Offline
Why are you so hung up on PCG vs xoshiro256** specifically? According to the articles you linked, they both suffer from many of the same flaws. If those flaws are acceptable, xoshiro256** is 3x faster. Although 3 * 0.95 nanoseconds is still tiny. And it seems like all they seem to be saying in those articles is that PCG and xoshiro256** aren't CSPRNGs, which wasn't a goal of theirs anyway. I don't see any complaints about either being statistically not-random so they're probably fine for most use cases?
If you don't care, use your language/environment default for ease of use.
If you need speed, either honestly seems fine. xoshiro256** is faster.
If you need quality, find a cryptographically secure PRNG.
Of course neither is a CSPRNG, this topic is for people to discuss on which PRNG is the best for general purpose.
I'm known as "haslo" in EE. Also, I refuse to play EEU.
Offline
They seem very similar practically speaking. Both past statistical tests, both fail if you don't take care in seeding them. xoshiro is 3x faster but, again, both are very speedy. I would probably lean towards xoshiro myself.
"Sometimes failing a leap of faith is better than inching forward"
- ShinsukeIto
Offline
PiotrGrochowski wrote:Gosha wrote:Why do you need the best pseudo-random number generator?
As long as they give pseudo-random numbers and don't start repeating numbers too soon, you are fine
even if it takes 1000 years for each random number it's fine?
What lol
Just use random.org
I am sorry for gravedig, but to just let others know, random.org has a HIDDEN currency system. When you run out of currency or get in the negatives, the whole site will lock up for you. You won't even know about it until you get the message that you do not have enough bits to just generate a number.
The currency works as follows:
- Generating ANYTHING will consume "bits" - the name of the currency.
- Each day, 200000 bits is refilled for free, up to a cap of 1 million.
- To exceed the cap or to skip the waiting time, you have to pay. Bits are not that cheap, you have to buy $10 worth of them at least with each dollar being a million bits.
- Using the "number list" tool where it generates a number of random strings will eat up a TON of your bits, about 600k per use.
- You can go into negative balances. It only doesn't let you generate if you have 0 or negative bits, so if you use a particularly demanding generator, you can end up with negative millions which will either take up $150 or 2 years to refill.
- There is no link on the main page to the bit checking page. The link is only shown when you get locked from generating due to a negative balance.
- It lets you search for other IP's balances and even gift them bits.
I don't see why a number generator needs to run on a complicated currency system.
Offline
It makes sense, random.org's main feature is that they use real noise (which they collect using radios or something from the real world), rather than generating it whenever people ask for it, so they only have a certain amount of supply they need to share between everyone. When you have a limited resource like that you have to have a way of limiting how much people can use so that you don't run out.
Offline
[ Started around 1732960224.6052 - Generated in 0.088 seconds, 12 queries executed - Memory usage: 1.69 MiB (Peak: 1.92 MiB) ]