[Rohans]

high posted:

float precision issue

Math.Floor() shouldn't do anything to a whole number, that's no precision issue.

[esalaka]

.95f * 100 is not a whole number.

Edited:

In fact, I believe in single precision it is just below 95

[Richy19]

Shanethe13 posted:

Would anyone be interested in working on a community project akin to "A Whiff of Steam" from the Love2d forums? There's a lot of talented/hard-working people here, and with some organization, I think we could work together to make a really cool game. I for one, think it would be a blast to make some attempt at this sort of game: http://www.terraria.org/

Edit: By community, I just mean some sort of Github project that we could fork off of and merge as wanted.

I love the idea but before we just jumpeed in, i think we should think of what language to use and what IDE, and library

I vote for SFML, and VS if C# or CodeBlocks if C++

[BlkDucky]

Why restrict IDE choice?

[layla]

Night-Eagle posted:

I have a separate VBO for every 8x8x128 voxels. I build them every time a change is made (dig block, place block). I use a triangle strip and render with glMultiDrawArrays(GL_TRIANGLE_STRIP, vaoFirst, vaoCount, vaoPrims);
Oh, I also only build faces that are not adjacent to other faces (0 chance of being viewed). Instead of translating to X, Y, Z coordinates when checking for this visibility, I simplify the formula and keep it in terms of the index.
I'm also doing some lazy culling and just rendering chunks within a 50.0 meter radius (will be larger when I get frustum culling and start running without the debugger) (each cube is 1.0 meter wide).

You can also just reuse the same buffer by adding faces where other faces got removed.

[ThePuska]

high posted:

Just tested, it is indeed a float precision issue.

I don't think so.

[DrLuke]

BlkDucky posted:

Why restrict IDE choice?

Maybe so the project files are interchangeable?

[danharibo]

Richy19 posted:

I love the idea but before we just jumpeed in, i think we should think of what language to use and what IDE, and library

I vote for SFML, and VS if C# or CodeBlocks if C++

More like CMake if C# or CMake if C++

[Dlaor-guy]

ThePuska posted:

I don't think so.

I think it's weird enough that those two lines output a different number.

[TVC]

Why real developers go for C/C++. :smug:

Put cast in there just to make sure.

danharibo posted:

More like CMake if C# or CMake if C++

I like Premake 4, it's Lua. :biggrin:

[high]

ThePuska posted:

I don't think so.

[danharibo]

high posted:

what.

Seriously, what.

[high]

danharibo posted:

what.

Seriously, what.

That is what it is compiled into.

[Rohans]

high posted:

So uh, a float CAN contain 95 as a whole but it doesn't when it's supposed to?

[Dlaor-guy]

I think we might just have found the first bug ever in C#.

[danharibo]

high posted:

That is what it is compiled into.

I was 'what'ing the compiler's choice to turn 0.95 into.. that.

Edited:

Dlaor-guy posted:

I think we might just have found the first bug ever in C#.

I doubt it.

[ThePuska]

It works if you cast the float you calculate inside the floor function into a float manually

Math.Floor((float)(.95f * 100))

for non-believers

[Rohans]

Dlaor-guy posted:

I think we might just have found the first bug ever in C#.

Prints 94 and 95, remove the casts and it prints 95 and 95. wat

[high]

Okay it has to do with going from a float to a double.

0.95 float turns into a 0.949999988079071 double.

ThePuska posted:

It works if you cast the float you calculate inside the floor function into a float manually

Math.Floor((float)(.95f * 100))

for non-believers

What .net version because the C#4 compiler is just optimizing it out for me, even in debug.

[thomasfn]

What about:

Console.WriteLine(Math.Floor(.95f * 100f));

Will that have any effect?

[Chris220]

DrLuke posted:

Maybe so the project files are interchangeable?

As long as everyone sets up their IDEs to use the same source files, you can have as many project files as you want. It really doesn't matter.

[ThePuska]

high posted:

Okay it has to do with going from a float to a double.

0.95 float turns into a 0.949999988079071 double.



What .net version because the C#4 compiler is just optimizing it out for me, even in debug.

3.5

[high]

ThePuska posted:

3.5

Maybe resharper is getting rid of it.

Anyways the compiler most likely is storing the result in a double when it does the compile time evaluation. Resulting in .95 becoming .949 and then multiplying by 100.

[TheCloak]

thomasfn posted:

What about:

Console.WriteLine(Math.Floor(.95f * 100f));

Will that have any effect?

[danharibo]

Chris220 posted:

As long as everyone sets up their IDEs to use the same source files, you can have as many project files as you want. It really doesn't matter.

But that's a stupid idea, Use CMake or your obscure build system/Project generator of choice.

[TheCloak]

And..doubles work

[Chris220]

I think we're all in agreement that floats store a value slightly under 0.95, and doubles do not

Then again, I don't have any content either so... I guess I'm not really helping.

[esalaka]

WAYWO v17 - Rounding to 95

Edited:

Constant floating point numbers are (In (many of) the C family languages) doubles by default, hence it working without the 'd'.

Conclusion: fuck floats they can't do 95 right

[Shanethe13]

Richy19 posted:

I love the idea but before we just jumpeed in, i think we should think of what language to use and what IDE, and library

I vote for SFML, and VS if C# or CodeBlocks if C++

Oh, of course. I was just throwing the idea out there. Without any planning there'd be a near zero-chance of getting anything done. Restricting the IDE sounds a bit much though. I'd personally be fine with any language: I don't have as much experience in the C family of languages, but how hard can it be?

[Icedshot]

Shanethe13 posted:

Oh, of course. I was just throwing the idea out there. Without any planning there'd be a near zero-chance of getting anything done. Restricting the IDE sounds a bit much though. I'd personally be fine with any language: I don't have as much experience in the C family of languages, but how hard can it be?

Considering that 95 rounds to 94, probably quite hard

[ColdFusion]

Frugle posted:

You can vote only if you submit a game.

It wasn't always like this.

[Dlaor-guy]

esalaka posted:

WAYWO v17 - Rounding to 95

Edited:

Constant floating point numbers are (In (many of) the C family languages) doubles by default, hence it working without the 'd'.

Conclusion: fuck floats they can't do 95 right

Maybe it has something to do with Windows 95?

[q3k]

I'd like to see how the CIL bytecode looks like...

[Chris220]

Dlaor-guy posted:

Maybe it has something to do with Windows 95?

Next thing we know, it'll turn out that C# is just some massive ARG leading up to the release of Microsoft's next product...

[Irtimid]

Just added zlib compression to my package format, so far it's doing very well reducing the file size one kilobyte.

[Eudoxia]

I'm writing a little Interactive Fiction parser. I [mostly] don't care about the Interactive Fiction part, but disassembling natural language always intersted me

[Chris220]

Natural Language systems are very interesting indeed. I still look forward to the day when I can command my computer to do things, in common English instead of specific, pre-recorded commands...

Edit:
I really hope they don't get clever enough to talk back at you though

"Computer, open Steam"
...
"no"

[esalaka]

Sorta like this?

$ vim
Fuck you, use emacs
$

[high]

Looking at .NETs System.ServiceModel (mono supports it <3). It looks very nice. Kinda weird though.

host = new ServiceHost(typeof(WebHost), new Uri("net.tcp://localhost:5123/"));
host.AddServiceEndpoint(typeof(IWebHost), new NetTcpBinding(SecurityMode.None), "servicename");

It creates an instance of WebHost and calls the method that the client called. Was kinda expecting to give it a handler instance. I am assuming you can still do that, this is just the default behavior. Still a hell of a lot better than writing the server end in php and translating between the 2.

[Overv]

I decided to make an add function for 8 bit numbers that only uses logic operators .

// Headers
#include <iostream>

// Binary addition
void add( bool a[], bool b[], bool r[] )
{
	bool c;
	r[0] = a[0] ^ b[0];
	c = a[0] & b[0];
	r[1] = ( a[1] ^ b[1] ) ^ c;
	c = ( c & ( a[1] | b[1] ) ) | ( a[1] & b[1] );
	r[2] = ( a[2] ^ b[2] ) ^ c;
	c = ( c & ( a[2] | b[2] ) ) | ( a[2] & b[2] );
	r[3] = ( a[3] ^ b[3] ) ^ c;
	c = ( c & ( a[3] | b[3] ) ) | ( a[3] & b[3] );
	r[4] = ( a[4] ^ b[4] ) ^ c;
	c = ( c & ( a[4] | b[4] ) ) | ( a[4] & b[4] );
	r[5] = ( a[5] ^ b[5] ) ^ c;
	c = ( c & ( a[5] | b[5] ) ) | ( a[5] & b[5] );
	r[6] = ( a[6] ^ b[6] ) ^ c;
	c = ( c & ( a[6] | b[6] ) ) | ( a[6] & b[6] );
	r[7] = ( a[7] ^ b[7] ) ^ c;
}

// Helpers
void dec2bin( unsigned char in, bool out[] )
{
	out[0] = in & 1; out[1] = in & 2; out[2] = in & 4; out[3] = in & 8; out[4] = in & 16; out[5] = in & 32; out[6] = in & 64; out[7] = in & 128;
}

unsigned char bin2dec( bool in[] )
{
	return in[0] + in[1]*2 + in[2]*4 + in[3]*8 + in[4]*16 + in[5]*32 + in[6]*64 + in[7]*128;
}

// Example
int main()
{
	bool res[8];
	bool a[8];
	bool b[8];

	dec2bin( 117, a );
	dec2bin( 25, b );

	add( a, b, res );

	std::cout << (int)bin2dec( res ) << std::endl;

	std::cin.get();
	return 0;
}

Eventually I want to make my own ALU by implementing the logic with electronics. (That is why I'm using arrays)