Thread: int into bits Bitwise

Originally Posted by phantomotap

If I've done the little mathematics needed and accounted for mirroring correctly, you only 15 bits.

That's terrific but so far it has not been done in any number of bits. I think the OP should focus on the basic procedures here and then if s/he has time left can consider some potential optimizations.

Originally Posted by jeffcobb

I have an IPC server and an autoptr class to get going by tomorrow...

Not sure what kind of smart pointer you need, but Boost has a pretty decent set of them, IMO. They don't have a non-shared, non-slicing version, though...but if you do need something like that, you're welcome to use mine (100% unrestricted license, of course).

That's terrific but so far it has not been done in any number of bits. I think the OP should focus on the basic procedures here and then if s/he has time left can consider some potential optimizations.

O_o

I don't know how my comments may have confused you; I'll try to be clear this time.

This, the above quote, is exactly my point. The suggestion offered so far present a clear and obvious solution to this problem. Those suggestions are what the OP should be working with.

At this level of educational development it is misleading at best to try and force an obscure representation when it clearly has nothing to do with the stated goal of understanding bit manipulation and integer packing.

[Edit]non-slicing version[/Edit]

O_o

What?

Soma

Originally Posted by Sebastiani

Not sure what kind of smart pointer you need, but Boost has a pretty decent set of them, IMO. They don't have a non-shared, non-slicing version, though...but if you do need something like that, you're welcome to use mine (100% unrestricted license, of course).

Actually I have that. It was my thought to include into it the possibility of overloading the [] operator and leveraging internal information to catch buffer under and over runs. Seems kinda obvious but as I mentioned in another thread I am trying to squeeze all the extra debugging tricks/info out of standard and semi-standard tools that I can. Probably a fools errand but I thought to try...

Originally Posted by phantomotap

O_o

I don't know how my comments may have confused you; I'll try to be clear this time.

This, the above quote, is exactly my point. The suggestion offered so far present a clear and obvious solution to this problem. Those suggestions are what the OP should be working with.

At this level of educational development it is misleading at best to try and force an obscure representation when it clearly has nothing to do with the stated goal of understanding bit manipulation and integer packing.

[Edit]non-slicing version[/Edit]

O_o

What?

Soma

Interesting. The smallest number of bits I was able to get it done in was 18 (9 cells with three possible states each). I will be interested to see this...

Originally Posted by phantomotap

Nope. I'll be damned if I'm going to be responsible for leading the OP to such a confusing path.

Soma

The OP does seem to be scoring about a 12 on the confusion meter. Part of his problem is stress, part of it is lack of experience and part of it is several of us all submitting working yet different approaches to his problem. That is another reason I backed out of the conversation.

It did bring something to mind though...has this forum ever done friendly contests to see who can submit the best solution to various problems like this? I think it would be a hoot for us more experienced folk and educational for the more junior members...and it would also do away with this sort of confusion...just a thought.

Anyhow back to my IPC system. On a roll and want to have something working before my head hits the pillow tonight...

Originally Posted by phantomotap

[Edit]non-slicing version[/Edit]

O_o

What?

It just allows you to clone an object (that you only have a pointer to the base class to) without it getting "sliced". All of the ones that I had seen required a 'clone' method - kind of a kludge, IMO. So I decided to work on the problem myself, and finally came up with a way to do it that only requires the class to be copy-constructible. Plus, as a bonus, each object internally uses just a single pointer as data, so it's as light-weight as, say, an std::auto_ptr (well, not quite - each class that uses the template creates an addition pointer of overhead, but hey).

Originally Posted by phantomotap

Let your lecturer know that he is an idiot. If I've done the little mathematics needed and accounted for mirroring correctly, you only 15 bits.* Are you expected to save that last bit as well?



O_o

Yep. I read this stupid thread thinking I might find something interesting. I found a solved problem and a pedantic fool causing confusion. Lovely.

Soma

*Edit: I just realized that a much easier option exists. ;_;

Just skimming over this thread, and if the goal is to represent all possible tic-tac-toe boards in the least space, then 15 bits is the typical answer I would expect. There are exactly 3^9 = 19683 possible board combinartions if you assume that the number of X's and O's can differ by more than one.

You can get it down to 14 bits if you don't represent all the possible boards that contain a difference between the X's and O's of more than one. Not worth doing unless one is writing a compression algorithm of course.

Is there anything left to answer here?

Originally Posted by iMalc

Is there anything left to answer here?

Yeah -- who on earth would consider saving 3 bits worth any kind of thought at all? IMO that whole discussion constitutes sabotaging the thread, since you are clearly doing the OP no kind of favour. Real nice. Are we supposed to be impressed?

Originally Posted by phantomotap

I see. I'm an idiot. When I read your post my thoughts were "What business does a smart pointer have being polymorphic?".

I'm just guessing, you store the underlying object by indirection through a proxy typed from a templated constructor of the smart pointer class and clone the proxy instance when you need a copy of the underlying object?

Basically, yes - each class generates a global allocator object that handles construction/destruction, and a pointer to this object is stored in the smart pointer (so two pointers are actually required, not one, as I claimed earlier). Another cool thing about it is that the base class doesn't even have to have a virtual destructor.*

Oh, and sorry for the OT discussion, BTW...

*I do realize that no sane programmer would define a base class having a non-virtual destructor. It's just that I've always had this fascination with writing code that overcomes limitations of the language, and one aspect of C++ that always irked me was the fact that destructors aren't virtual, by default. So when I realized that the mechanism for that could be added with little overhead, I went for it, naturally.

So...what did the OP eventually do?

Originally Posted by MK27

Yeah -- who on earth would consider saving 3 bits worth any kind of thought at all? IMO that whole discussion constitutes sabotaging the thread, since you are clearly doing the OP no kind of favour. Real nice. Are we supposed to be impressed?

Huh?!
That was a genuine question. What question are we currently supposed to be answering? I haven't read the whole thread and I'm not going to go back and read all five pages now. I'm genuinely trying to work out if there's reason the OP hasn't featured in any of the posts on this page. I don't know what you're on about.

At the top of this page I'm seeing a question about whether a ttt board can be represented in only two bytes. I'm simply confirming that it can. As for why anyone would want to, I dunno.