The final release has been online for over half a month now, but seeing as this has been read more times than I would have expected I'm putting up a link to the file at ticalc.org.

http://www.ticalc.org/archives/files/fileinfo/436/43675.html

Here are some up-to-date screencaps:

Thank you, DJ_O, for creating this topic!

Everybody else: thank you for the compliments!

I'm loving this game. The only thing I don't like is you can't shift stuff while it's falling, but this is awesome.

Thank you! You just have to be patient I guess. You can't make any swaps in the original game either, but the animation is faster there.

By the way is that member a part of the forum here? I swear I saw his nickname somewhere.

That's right, I asked about 24-bit division, which in the end I didn't need.

Thanks, squidgetX, for starting this topic and putting up the screenshot! I didn't realize it existed before I googled 'Benumbered' to quickly go to its page on ticalc and found a link to this.

Madskillz: It's not like anyone is gonna get a score close to 9,999,999 (or 16,777,215) in this game soon, but I did not feel like capping it at 65,535, so that's why I use 24-bit scores. The highscore lists in the menu are ridiculously high and it is on my to do list to change those to reasonable highscores to beat in the finished version.

thepenguin77: For the grayscale I use a very similar method to the one used in Diederik Pieter Kingma's <a href="http://www.ticalc.org/archives/files/fileinfo/295/29520.html">GPP</a>.

Thanks everyone else for the positive comments! I'll have to have another look at it, because it is writing to some bad memory addresses on my TI-83+ sometimes and I did not locate the source of the error yet, but I'll try and do that tomorrow.

Ahh, okay, I get it. You just only keep the 4 (or 8, or X, depending more on precision than ease to work with I suppose) most significant bits of M and then shift N along that, right?

Thank you, Ztrumpet, glad to be here I will test both routines now. I doubt anyone will halve the memory footprint again, so if it works I think will use Runer's... But... Now that I think of it, maybe you've read that I managed to drop off a byte of precision so that it turned out to be only two 16-bit values I am working on. I multiply the smaller of those by 50 so I get this scale, which is going to be a progress bar of 50 pixels high that you have to fill on each level: 50 * current score(2 msbs of the real value) / score needed(also just the 2 msbs). I multiply the 2 msbs of the current score by 50 first (the result will be in 24 bits again) and then divide by the score needed. Obviously this is now only a 24/16 division.
So now that I see that you've come up with 24/24 (still don't know how both of you did that), I could leave the lsbs in, but I still have to multiply by 50 first. I know for a fact that in my case that the current score times 50 would not exceed the 16843007 (correct me if I'm wrong) max value of 24-bit. What I'm saying here is that I would need a 24by8 multiply function as well, before I can divide 24/24. Anyone up to the challenge?

EDIT: I just tested both Runer112's and thepenguin77's routines and they both seem to work fine. I don't really know what would be good test values... I tried to divide 0 by 0 and they both give 16777215 as the answer, but I guess since they both do, it's okay

Woah! How did you come up with that that fast? Thanks a lot! I actually found a way to get around the 24 bit division. Since the quotient in my specific case would always fit in one byte (I am scaling the ratio between two 24-bit values from 0 to 50), I can just omit the lsb of the dividend and the divisor and the result will still be sufficiently precise.

I will only use the two msbs of both values. If they are below 65536, I can just use the two lowest bytes, otherwise I use the two highest bytes. I multiply the lowest value with 50 (so that ends up being a 24-bit integer again) and then I divide them with a 24/16 division routine. Does the job!

Still, I am amazed at how easily you came up with this routine (I should have paid more attention at school when they were explaining long division) and I will put it in my math libary if that's okay with you. What's at $8000, though? Beause you are overwriting the 6 bytes in memory there aren't you?