Print

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

[Skyrunner]

Did you try it lots of times?

I agree with Ruler; probably floating point is the norm, and 8-bit integers have to (1) be cut from 64->8 bit (2) throw away everything after the float.

You're a 64-bit system, right?

[dreadmullet]

Hmmm. Maybe downcasting to a 8-bit integer takes some time, or possibly the second scenario is better positioned to take advantage of certain features of modern CPUs?

You misunderstood me. But I misunderstood myself: boost gives me a 32 bit integer between 0 and 255 which is casted to an 8-bit integer. But that 0.5 second difference is still strange. I've been trying to improve the performance of generating random numbers just for fun, but when something like this happens it makes my head hurt. Maybe I could extract four 8-bit numbers from a 32-bit number using bit shifting or something?

[Mego]

If you are using int's, that is more than an 8-bit integer. chars are 8 bits. int's are 4 bytes, so 32 bits. Also, the compiler is smarter than you think. It's probably optimizing out the 64-bit float part and just doing it straight to 32-bit floats. So, really, you're generating a bunch of strings of 32 bits, and then generating a bunch of strings of 32 bits. The slight discrepancy in time can be attributed to other processes borrowing a bit of the CPU.

[kaenneth]

I would guess that the random shuffler takes the bulk of the time; everything else is insignificant.

Yeah, taking 64 random bits and doing

a[0] = r || 255;
a[1] = r << 8 || 255;
a[2] = r << 16 || 255;
a[3] = r << 24 || 255;
a[4] = r << 32 || 255;
a[5] = r << 40 || 255;
a[6] = r << 48 || 255;
a[7] = r << 56 || 255;

would be quicker (I might have the << reversed)

[dreadmullet]

Your code is completely wrong. However, it got me going in the right direction and I came up with this:

a[0] = r & 255;
a[1] = r >> 8 & 255;
a[2] = r >> 16 & 255;
a[3] = r >> 24 & 255;

It splits a 32-bit integer into four 8-bit integers. It reduced the time from 8.0 seconds down to 7.5 seconds. Woohoo!

[kaijyuu]

If you're concerned about speed, you might want to find an alternative to generating 16 million random numbers.

[dreadmullet]

Apparently I was using a slow floating point distribution. (boost reference) I switched from uniform_real_distribution to uniform_01, and now I got the floating point benchmark down to 4.7 seconds. Using 8-bit integers is still 7.5 seconds. :S


EDIT: I got the 8-bit part down to 1.25 seconds! Apparently, passing a byte array through a pointed function 17 million times is slower than using four static byte variables. I have no idea what I'm doing.

So yeah, I can now generate a 4096x4096 random noise image in 1.25 seconds. Awesome. (It takes a few seconds to save the compressed .png to disk, though.)

[Siquo]

The challenge lies within concepts, abstract stuff, and connecting all that stuff in your brain, before you write it down.

Quite similar to natural languages, if you ask me. The basic grammar takes only a few months for me to get--but it's the deeper things like idioms, slang, informal grammar, and the like, that are the truly hard part to learn.

Ah, but you can still write a great book without them, just as you can program great things without the deeper stuff of most languages (even if they will make it easier to express yourself, once you know them).


Other question: Anyone know of a really good Database Modeling tool? If it spits out sql, it'd be great but it's no prerequisite. Must work on browser or mac, I just need the graphics.

Edit: Nvm, found a great one for my purposes: wwwsqldesigner.

[Akura]

Not sure if it counts for this thread, but...

A while back I programmed a thing for RPGToolkit that types out each letter of a message as opposed to the normal behavior of writing the whole line using the built-in functions. I went back and looked at it again over the weekend, and noticed it was doing something odd to the first letters of a given string, probably because it was redrawing the same text over itself over and over. Specifically, it redraws the entire line, adding the next letter to the string.

Then I figured I could just write one letter at a time. That didn't work, all the text just jumbled together. I figure I could fix the original problem by clearing the image(why I didn't do that to begin with I have no idea, it seems so obvious) before redrawing it.

I haven't tried it yet, so odds are it's going to screw up horribly like it usually does.

[Skyrunner]

Two simple questions (bear in mind I'm a C++ newb >_<)

(i) Can a structure hold a structure? As in:

 

Code: [Select]

(There is another structure called 'eexample')
  struct xample
    {
       eexample oranges;
       int apples;
       float cherries;
     };
(ii) How do I create variables, then name them sequentially?

The only way I've thought of is making nodes that hold (a) the 'variable' name, (b) the actual content of the 'variable' and (c) the next node's address.


If the above two are doable, I shall start on my largish console-based D&D combat simulator xD

Another optional, but not necessary, thing to know would be how to access big files with lots of info in them, which would let me both (a) reduce .exe file size and (b) allow users to edit the files.

[Sowelu]

Yes, a struct can hold a struct.  Or a pointer to a struct, if that's your thing.

As for naming variables sequentially, bad idea--how would you access them?  Consider using a list vector, or...whatever passes for a dictionary in C++.  I've been so totally spoiled by C#.

Two simple questions (bear in mind I'm a C++ newb >_<)

(i) Can a structure hold a structure? As in:

 

Code: [Select]

(There is another structure called 'eexample')
  struct xample
    {
       eexample oranges;
       int apples;
       float cherries;
     };
(ii) How do I create variables, then name them sequentially?

The only way I've thought of is making nodes that hold (a) the 'variable' name, (b) the actual content of the 'variable' and (c) the next node's address.


If the above two are doable, I shall start on my largish console-based D&D combat simulator xD

Another optional, but not necessary, thing to know would be how to access big files with lots of info in them, which would let me both (a) reduce .exe file size and (b) allow users to edit the files.

[Skyrunner]

At naming variables sequentially : I don't know those things is the problem xD What's a vector? Is it those line things in math where planes fly and stuff? No? D:

Uh ...

As for the node list thing method, I was planning on accessing every single node in that list every single time it is called for, cout'ing the names of the list, and then when the user chooses one, going to that variable and accessing the memory~ xD


Random question :

  What happens if I do new int = 8 and then don't assign it anywhere? o_O Does it get lost in memory?

(Again, C++.)

[Sowelu]

At naming variables sequentially : I don't know those things is the problem xD What's a vector? Is it those line things in math where planes fly and stuff? No? D:

I believe that lists in C++ are called vectors in the standard library.  It's just, like...well, it's a list of things.

As for the node list thing method, I was planning on accessing every single node in that list every single time it is called for, cout'ing the names of the list, and then when the user chooses one, going to that variable and accessing the memory~ xD

Yeah, that's a job for a dictionary or two lists.  In C# this would be beyond trivial.  In C++, you'd iterate through a 'name' list and print them all out.  Then, when the user selects one by name, iterate through the 'name' list again until you find the index of that string.  Then index into the 'value' list and return the value at that position.  So, if the name they typed was the third element in the name list, you bring up the third element in the value list.  You have to keep things in order; the first element of the name list is always associated with the first element of the value list, etc.

If your list will always be a constant size, just use an array instead.

  What happens if I do new int = 8 and then don't assign it anywhere? o_O Does it get lost in memory?

Yes, AND you get a memory leak.  You should always delete what you free.  Do not call up what you cannot put down.

[Skyrunner]

In C# this would be beyond trivial.  In C++, you'd iterate through a 'name' list and print them all out.  Then, when the user selects one by name, iterate through the 'name' list again until you find the index of that string.  Then index through the 'value' list and return the value they wanted.  You have to keep things in order; the first element of the name list is always associated with the first element of the value list, etc.

I guess I'll have to do it that way, with nodes. Good thing nodes are basically chained structures >.>

Any words on file I/O as a database?

[Mego]

In C# this would be beyond trivial.  In C++, you'd iterate through a 'name' list and print them all out.  Then, when the user selects one by name, iterate through the 'name' list again until you find the index of that string.  Then index through the 'value' list and return the value they wanted.  You have to keep things in order; the first element of the name list is always associated with the first element of the value list, etc.

I guess I'll have to do it that way, with nodes. Good thing nodes are basically chained structures >.>

Any words on file I/O as a database?

There isn't a really easy way to generate sequential variable names without doing some funky things with preprocessor directives that you probably don't need to be doing anyway.

The std::vector class is like an array, only its capacity is not fixed. So, you can continue to add objects to it, so long as there is memory left on your computer. Nowadays, that shouldn't be a problem.

As for dictionaries in C++, look at std::map.