Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Skanky]

And perhaps compare and contrast to the one I get using this seed.
8 rejects

[Flame11235]

There's always the possibility that the he posted the wrong seed.

[Skanky]

Not really, due to:

quote:Originally posted by 0x517A5D:
<STRONG>Couldn't say.  It works for me.</STRONG>

[Sinnyil]

No no, I mean I can guarantee it.  I have a laptop running XP and my home computer running Vista.  It has to do with how operating systems handle RNG.  That's really all I know about it, perhaps someone with more knowledge of computers can explain it more thoroughly...I know I've seen it explained in detail before.

Essentially, when I see a seed on the forums or on the wiki I really want to try out, I wind up generating it on the XP machine and sending it to myself.

[Squeegy]

No, I'm on Vista and it generates fine for me.

[Daniel Charms]

I just generated a new world using this seed on Vista, using 33g, and I got the exact same spot at exactly the same location. My map is almost identical to his, except mine doesn't have the large blue omega area in the upper half.

[0x517A5D]

If it is true that the same seed yields different worlds on different operating systems, then it is a bug, and should be reported as such.

"has to do with how operating systems handle RNG" is, essentially, nonsense.

Windows does not have a random-number API call.  (Attention, nitpickers: yes, I know about the crypto API.  That's beyond the bounds of this discussion.)

The C++ compiler Toady uses has, in its library, a LCG PRNG.  It's named rand(), and is of good but not excellent quality.  +Finely-crafted+, in DF terms.  The corresponding routine which initializes or "seeds" it is named srand().

srand() is called once, using the current time as the source of randomness.  That is rand()'s only dependency.

The vast majority of the random numbers that DF uses are generated by Toady's own code, using a better (*superior quality*, in DF terms) PRNG called Mersenne Twister.  There are no operating-system dependencies in Mersenne Twister.

There is a routine which seeds MT using, again, the current time as the source of randomness.  That, and a call to rand() are MT's only dependencies.

However, Toady's implementation of MT can also be seeded with a specific number.  When doing this, neither the current time nor the call to rand() are used.  Only a single user-supplied number is used.  When this is done, there are no dependencies left in MT.

So if the same seed generates different worlds on different operating systems, report it as a bug.

[0x517A5D]

Daniel: the blue omega area (oooh, Blue Omega, what a name for an album) represents your chosen civilization.  You have three choices, selectable from the Your Civilization tab of the Choose Fortress Location screen.  Press Tab twice, then press + or - to choose.  I think you'll find that the first choice, "Persuasive-Wall of Cur-something" matches the screenshot.

[Jifodus]

0x517A5D, last I checked, Toady staticly linked s/rand() into the executable, did he not this time? And if it is staticly linked, shouldn't it always be the same where it's run.

Yes, yes, I do know that with the soon to be Mac port, there will (most likely) be a difference in the seeding of the MT. But for now, it shouldn't matter what OS is running, the results should always be the same.

Maybe if Toady hasn't done something similar yet, we should convince him to use LGC RNG off Wikipedia for cross-platform MT seeding.

It has previously been determined that differences in worlds aside from version changes, are caused by modded raws.

[Skanky]

Reported.

[polpoint]

I noticed the same thing with one of the maps in dwarffortresswiki. It came out completely different from the image that was posted.

[Skanky]

And then promptly discovered it was my error after all. It seems that reaction_standard does affect world generation. With the original version of the file, the game creates the same world as the original seed.

[0x517A5D]

Hmm, wouldn't have expected that.  I would have thought it would mess with the stone layers, maybe, but not with the shape of continents.

Well.  Good to know.

[Daniel Charms]

0x517A5D: Ahh, that clears it.

[Sinnyil]

Odd, did this used to be the case a few builds ago?  I know (I think I upgraded around 33d) when I tried to generate Paul's seed that it came up different and I wound up having to generate it on the laptop and transferring it to my desktop.  And this was before I ever started touching any type of raw or playing with any of the utilities.  

If this was fixed with the fix to mineral dispersion, I profusely apologize for spreading untruths.