next »

[8BitPimp]

I ported Notch's Minecraft4K javascript demo to C++ and SDL 

The source is there for anyone who is interested in 3D raycasting.

http://the8bitpimp.wordpress.com/2014/07/30/minecraft-4k-c-port/

[RandyGaul]

Threaded the island solving in my physics engine

It was pretty rough as I ended up finding some ginormous (and old) bugs from uninitialized memory. Implementing the threading shed light onto the crazy memory bugs... But it's all working now!

[Sik]

I do want to pose a question:

I only have 8 bits to work with for 3 variables (gloss, roughness, specular amount) in the BRDF.  If I can find a way to mess deeper with Unity's deferred lighting system (perhaps impossible), then I can get another 8 bits (by cramming the normal into 16 bits instead of 24).  With 16 bits, I'd feel comfortable with ~5 bits per variable, but with 8 bits, ~2 bits per variable leaves a bit wanting.  
Now, the question comes, should I just use a look up table?  

I can only have 256 distinct triplets (barring interpolating between things), but I imagine that for my own personal uses I can easily get away with 256 distinct reflectance parameters (pretty much all metals can use the same one, so too with high gloss things like car paint, so I can cram a lot of different materials into that range).

If you have the X and Y axes of the normal you can compute the Z axis from them. I don't remember the exact calculation out there but it's a common trick to reduce space usage, so it shouldn't be hard to search. This means that with 16 bits you'd get two 8-bit values instead of three 5-bit values (and with 8 bits you'd get two 4-bit values instead of three 2-bit values).

[standardcombo]

I'm a grumpy programmer and I came here to see if I can become happy.

[manabreak]

Managed to create a tiny C++ project from scratch and run it on Windows, Android and Windows Phone 8 successfully, with platform-specific code and common interface. Cross-platform games, here I come!

[8BitPimp]

I do want to pose a question:

I only have 8 bits to work with for 3 variables (gloss, roughness, specular amount) in the BRDF.  If I can find a way to mess deeper with Unity's deferred lighting system (perhaps impossible), then I can get another 8 bits (by cramming the normal into 16 bits instead of 24).  With 16 bits, I'd feel comfortable with ~5 bits per variable, but with 8 bits, ~2 bits per variable leaves a bit wanting.  
Now, the question comes, should I just use a look up table?  

I can only have 256 distinct triplets (barring interpolating between things), but I imagine that for my own personal uses I can easily get away with 256 distinct reflectance parameters (pretty much all metals can use the same one, so too with high gloss things like car paint, so I can cram a lot of different materials into that range).

If you have the X and Y axes of the normal you can compute the Z axis from them. I don't remember the exact calculation out there but it's a common trick to reduce space usage, so it shouldn't be hard to search. This means that with 16 bits you'd get two 8-bit values instead of three 5-bit values (and with 8 bits you'd get two 4-bit values instead of three 2-bit values).

Yes you should be able to take the cross product of X and Y to find Z.  If X and Y are unit vectors then the derived Z vector should be unit also IIRC so you dont even have to normalize.  I think thats quite a common optimization to store a normal map using only two channels.

Perhaps you could take it even farther and have a 8bit lookup table of normals which would also avoiding the calculation of the cross product.  Perhaps the quantization would look too ugly though.

[MorleyDev]

Finally clawed my well out of callback hell and refactored some node.js code to use promises. The end result is code that is easier to read, easier to test, and more closely follows SOLID principles.

Let us never speak of the horror of the 5-level deep nested callbacks again.

[standardcombo]

Yikes, I can remember having a similar callback nightmare back in a facebook project. Debugging was not a strength of AS3.

[Fallsburg]

If you have the X and Y axes of the normal you can compute the Z axis from them. I don't remember the exact calculation out there but it's a common trick to reduce space usage, so it shouldn't be hard to search. This means that with 16 bits you'd get two 8-bit values instead of three 5-bit values (and with 8 bits you'd get two 4-bit values instead of three 2-bit values).

Well, the problem is Unity. I know of that trick, it's just whether I have enough control over Unity's rendering system to control how the depth/normals are written. I haven't done the digging to find that out yet.

[MorleyDev]

Yikes, I can remember having a similar callback nightmare back in a facebook project. Debugging was not a strength of AS3.

Yeah, I've found the issue with callbacks is you lack a stack, so it's very difficult to avoid either repeating code or having to call the same function in multiple paths, so the exact same error handling code gets sprinkled around like pixie-dust. Then you have to make sure you only call the callback once (some prominent js libraries have had bugs where that didn't happen) and it's not as immediately obvious where that 'exit point' is compared to a standard function call.

There are ways around this that still use callbacks, but promises let you get around it nicely and in a way that I find a lot easier to follow, since it produces a code-flow that is a pipeline of sequential steps that return and process data. It appeals to me

[Slader16]

Just learned about indexes, for example:
"Tigsource"[2]

Would print "g".

I had no idea I could do that, even though I'm a pretty decent programmer

[framk]

That's a pretty integral part of programming. How have you gone so long without it?

[Sik]

Not all languages let you do that (some require you to use a function instead).

[Geti]

Slader do you mean indexes in general or just that you could apply them to strings in c/++ due to strings being char pointers? If you mean indexes in general you really need to like, look into basic data structures and algorithms. I would not call anyone oblivious to arrays a decent programmer.

[Sik]

Considering how he explicitly showed them being used on a string to retrieve an individual character, I assume he meant using indexes in strings (which is probably rather obvious to a C or C++ programmer, but not that obvious on higher level languages where strings are not treated as arrays).

[InfiniteStateMachine]

That would be my assumption too. It's been a little while since I've used java but you can't use the index operator on a java string right?

[Geti]

treated as arrays

They're not just "treated as" arrays, they are arrays.

[Sik]

Not in some languages, that was my point =P (I'm aware that they're literally arrays of char in C and C++)

[Agumander]

Once you've been made to peer into the abyssal home of the old gods implement a memory allocator, it's all char arrays and a few helper functions anyway.

[Garthy]

Conceivably, if you started out with any language lacking both operator overloading and providing specific string index get/set calls, perhaps even did most of your development on existing large codebases, then you could find yourself going a long time without ever needing string indexing via an array-style operator. You could still be quite skilled, just never had a need to use them.

It would be an unusual path to travel for sure, but it looks like at least one person has done just that!