Print

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

[Nadaka]

Wouldn't you be much better off using the time as a seed?

Time as a seed presents a few challenges.

For close calls, the time may not have changed, producing identical numbers.
For close calls, the change in the seed may be insignificant, and depending on the algorithm may produce a random number "near" the previous one more often than not.
If you want to be able to reproduce the sequence of random numbers at a later time, such as in procedural terrain generation, then time does not work at all.

[MorleyDev]

Wouldn't you be much better off using the time as a seed?

Time as a seed presents a few challenges.

For close calls, the time may not have changed, producing identical numbers.
For close calls, the change in the seed may be insignificant, and depending on the algorithm may produce a random number "near" the previous one more often than not.

You don't manually seed it after every call. An LCG works by feeding the last value back into it to generate the next one. time is just used to get the starting seed that determines the pattern.

X[0] is the seed.
X[n+1] = (A * X[n] + C) mod M

So for the first value (X[1]), the seed of time (X[0]) is used. For the second (X[2]), the X[1] is used. X[3] uses X[2], X[4] uses X[3], X[5] uses X[4] and so-on ad infinitum.

[GlyphGryph]

For close calls, the time may not have changed, producing identical numbers.

This... isn't how seeds work. The number still changes each time. That's why it's called a seed, not a whole tree. (or whatever) You only need to seed once.

For close calls, the change in the seed may be insignificant, and depending on the algorithm may produce a random number "near" the previous one more often than not.

Again, this is not, from my understanding, how a seed works. Why would it?

If you want to be able to reproduce the sequence of random numbers at a later time, such as in procedural terrain generation, then time does not work at all.

Save your seed. Attach the seed to the generated data. Done.

[Nadaka]

1: that isn't the way i read his question.

2: there are random number generators that take the current time as the seed for each random number in order to remove the pattern inherent in conventional seeds. For instance (IIRC), the default random number generators in javascript and java. For sufficiently low resolution timers combined with sufficiently high speed processors, this presents the problem I mentioned.

EDIT: yep. that is how javascript works, but not java.

[GlyphGryph]

Since the seed is, by definition, the initial value used for a pattern of random results from an RNG, I do not see how it could reasonably be called a seed if you only generate one number with each seed. That's just... that's not anything, really.

If you have a decent algorithm there shouldn't BE a pattern inherent in a conventional seed.

Now, where it MIGHT cause problems is, I suppose, with forks, if the RNG in both forks continues along the same path...

[Virex]

That's why you shouldn't fork an RNG in the first place, or at least be very careful about it.

Since the seed is, by definition, the initial value used for a pattern of random results from an RNG, I do not see how it could reasonably be called a seed if you only generate one number with each seed. That's just... that's not anything, really.

You're hashing the time then. Still, for that to work your hash function needs to be good, which is not necessarily the case for linear hashing (if the modulus is chosen wrongly you're not getting the whole space from 0 to m for example).

[MorleyDev]

Yeah, using a new time for each generation is more hashing than seeding, it's called a "seed" because all the other numbers grow from it ^^

If you need multiple generators to exist (multithreading is the main reason I can think of where this would be important), each would be seeded separately. That's one of the complaints people have with rand(), but that is a side-effect of it being part of the C library that simply gets dragged kicking and screaming into C++ xD

Of course mersenne twister has its faults too. It's can be...unfriendly on the cache, to say the least. Largely due to the large array of values it has to store.

[Valid_Dark]

Just finished the first beta of my first game in SDL

http://www.filefactory.com/file/c2d3f9f/n/RoboJuggle.zip

you can download from there,

there's only 1 ball right now, and a few bugs, such as it getting launched from the right hand too fast,
although thats purely cosmetic as it doesn't effect the time it takes to reach the left hand.

press space to drop a ball, currently only 1 ball can be on screen at a time.

[Siquo]

Awesome! Works quite well. The ball animation is flickering, though, that's kinda annoying.

[Valid_Dark]

I noticed that too, I think the sprite might be cycling too fast,

It doesn't flicker if the ball isn't moving.

And I should have mentioned, arrow keys move right hand, wads move left hand

[Valid_Dark]

I can't figure this out, I slowed down the cycle rate of the sprite and it still flickers while moving, but at least now I have a variable to control spin speed, and I can't figure out whats causing the jump from the right hand. it doesn't make sense, and when I speed up the ball speed it doesn't happen.  So i'm going to find the slowest speed it can go and not skip and have that be the starting speed.
I have a variable that controls ball speed, so the ball can speed up later if you get so many in a row.
I think think if I try switching to triple buffering instead of double it'll stop the flickering, but that means I have to look up how to do that.  going to put the resources in a folder instead of being in the same folder as the .exe . 
then going to work on the score system, once that's in place going to add second ball.
any ideas as to what to work on next ?

[RulerOfNothing]

Have you tried using double buffering?

[kaijyuu]

I'm not sure how you do you graphics, but.... Are you freeing and loading the ball surfaces every single time it animates? If so, bad mojo. Pre-load them and instead. Animation can be done via deciding which surface to blit to the screen, rather than always blitting the same surface with different graphics applied to it.

[Valid_Dark]

I'm not loading the surface every time it animates, it's a pre loaded sprite sheet, I feel stupid because I had it set to there being 15 sprites in the sheet when there were only 14. I feel dumb for not noticing it, because I thought it didn't flicker when standing still but it did, I just didn't notice it.  So got that bug fixed.  also found out it's stupid hard to triple buffer with SDL, they should make a triple buffer flag like the double buffer one.  my distances were all intervals of 80 and I was dividing by 50, and for some reason it was making a +/- 50 or so pixel difference which really doesn't make sense. so I changed the 50 to 40 and now it runs a lot smoother.
that 50/40 variable controls the speed of the ball, by deciding how many ticks between min and max distance.
I really wish I could go to school for programming.  the only computer class I've ever taken was typing in high school,  I wanted to but was too busy with science classes. (During high school I was dual enrolled in my city Highschool and a Biotechnology acadamy) My code has a very bad structure to it and in every sense of the word is a mess.
but it works!

now to add features, i'm going to pile more crap on the already giant pile of crap.  can I have someone who's good at c++ and sdl take a look at my code and give me some structure pointers?

[Mego]

Well, I'm good at C++, but have very minimal experience with SDL. I can probably figure it out if I have a reference open.