Print

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

[itisnotlogical]

So in Unity, I want to create a shader that behaves like an alpha mask. Suppose the scene with regular materials is rendered as one image. However, one mesh has my hypothetical shader on it. Instead of rendering the object, a hole is cut in the rendered scene, revealing a flat, unlit texture underneath the "rendered scene" layer. If you've ever seen how characters in plaid clothes were animated in old cartoons, that's the effect I'm going for.

I have the feeling I'm not going to accomplish this with any of Unity's built-in stuff or standard assets, and I don't want to go with an Asset Store solution. So, where do I start reading to figure out how to accomplish this?

[Rose]

Render a transparent texture that writes to the depth buffer.

Easier may be having a texture applied to the shader that's mapped to screen coordinates.

[Reelya]

Is there any sort of decent documentation anywhere about writing shaders in Unity? I looked around but can't find anything useful. Unity's own documentation is pathetic, you can find some function names and primitives mentioned some places, but no clue as to how you're meant to use them or connected them together.

 So far I'm just cutting and pasting stuff from other sources hoping it works without know what the allowed syntax even is, or how the sections work. Even the guy at the game studio where I'm helping out, who writes the Unity shaders didn't understand half the stuff in the shader I cobbled together for my terrain renderer, so I get the feeling it's hard to come by good documentation.

I couldn't understand any of the stuff in his shader, and I had a simple request: how to make the output from my shader brighter? Any sort of brightness / gamma thing I can play with, and he couldn't get it to work. It seems like a simple thing, just scaling up the colour values.

[Rose]

I just did a shitload of google searching to find out what I needed.

[Moghjubar]

I couldn't understand any of the stuff in his shader, and I had a simple request: how to make the output from my shader brighter? Any sort of brightness / gamma thing I can play with, and he couldn't get it to work. It seems like a simple thing, just scaling up the colour values.

I can't believe it would be that hard... I mean, that shit is basically trivial in GLSL.

I'll help look around, be back if I find anything.

Edit: So, I don't have Unity installed to mess around with things... but looking at the docs, apparently Unity uses either CG or HLSL so those are the docs you should probably look at.  In any case, from http://docs.unity3d.com/Manual/ShaderTut2.html  I see "texture modulated by color" so that seems simple enough.

[itisnotlogical]

Oh wow. I've skimmed the ShaderLab stuff before so I knew it was pretty skimpy, but damn. It's just uncommented source and a token "Take our word for it, this is what it does" line for every function. At least I found a snippet that (apparently) maps a texture to screen space. The might as well have just said "Go read a book on OpenGL, dickhead" under every heading.

[Scoops Novel]

Can anyone code a parser (config style) for this forum so i can view it in Siilihai? It's a forum client (like thunderbird for email or pidgin for messaging), and i'm itching to use it with this. There's a parser maker in the client itself, and Simplemachine forums are supported. I've appealed to Toady to take 20 minutes and install a TapaTalk plugin for the site. He's verra busy though so i hope u can take it off his hands.

He's unfamiliar with how TapaTalk works and fears security issues, so it's up to us to code a parser for Siilihai.

Wish i'd thought of this before!

[Orange Wizard]

Is that a config file or do you actually need a script to pull all the data off the forum and display it in a slightly different format?

[Puzzlemaker]

Damn it took me way to long to get it to select and move a sphere around the screen.

Working on Imperium, still.

http://www.bay12forums.com/smf/index.php?topic=151256.0

Hopefully I can have a working prototype sometime this year 

[Scoops Novel]

Is that a config file or do you actually need a script to pull all the data off the forum and display it in a slightly different format?

It looks a lot like a config file, but I could be wrong.

Edit: Toady's unfamiliar with how TapaTalk works and fears security issues, so it's up to us to code a parser for Siilihai. The client itself has a parser maker.

[Bumber]

Okay, I'm at wit's end here trying to multiply double precision floats in x86 Assembly. Our first assignment (to be due on the first day of class before we've even had a single lecture for some screwed up reason) is to write a program that calculates the surface area and volume of a sphere of a given radius. The relevant section looks a bit like this so far:

Code: (Assembly) [Select]

PI equ 0x400921fb54442d18                                   ;Represents 8-byte float Pi (NASM Manual)
;rbx = radius

mov qword   rax, PI                                         ;Set rax to Pi so we can calculate surface area 4*Pi*r*r
add         rax, rax                                        ;Pi + Pi = 2*Pi
add         rax, rax                                        ;2*Pi + 2*Pi = 4*Pi
mul         rbx                                             ;Multiply 4*Pi by r, result in rdx:rax (most significant in rdx, least significant in rax?)
;Something about storing rax to memory here?
mov qword   rcx, rdx                                        ;Save significant part into rcx
;More stuffOnce that's done, I'm going to need to multiply by r again, store that result, then multiply by r and 1/3 to get the volume.

Can I just ditch the least significant digits between multiplications or will that incur noticeable rounding errors?

Best guide I could find was this, but I'm not really sure what's going on with the saving to and retrieving from [bignum]. Maybe I'd be better off working with the SSE vector registers?
(Edit: It looks like that's for multiplying two 32-bits to make a 64-bit. I only need to multiply two 64-bits to make another 64-bit. Maybe just chopping off the end is fine? It's limited to 15 digits of accuracy with my value for Pi anyways.)

[LoSboccacc]

rax will have the result, rdx the overflow (as you noted yourself)

so mov rax in rcx

[Bumber]

rax will have the result, rdx the overflow (as you noted yourself)

so mov rax in rcx

So can you help me clarify this? Suppose we have two 32-bit floats (for simplicity's sake.) In memory they are stored as 1 bit for sign, 8 bits for exponent, and the remaining bits for the mantissa:

SEEEEEEE EMMMMMMM MMMMMMMM MMMMMMMM
01000000 01001001 00001111 11011010 = 3.1415925
00111110 10101010 10101010 10101010 = 0.33333331

Suppose they are multiplied through arcane magic and you get a 64-bit float like so:

/---------------EDX---------------\ /---------------EAX---------------\
SEEEEEEE EEEEMMMM MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM MMMMMMMM
00111111 11110000 11000001 01010010 00010110 11000001 11011000 11001100 = 1.047197426696175    (I got the binary from the proper decimal value.)

Assuming I can just drop EAX, I'd be left with:
/---------------EDX---------------\
SEEEEEEE EEEEMMMM MMMMMMMM MMMMMMMM
00111111 11110000 11000001 01010010 = 1.8808997    (When read as a 64-bit float.)

Which seems like unacceptable rounding error, not to mention it's not a readable 32-bit float anymore. Maybe it's better with 64-bit floats? Maybe I shouldn't assume it would switch to 64-bit exponents from the MUL instruction?

For reference, I used these to come up with the conversions:
http://babbage.cs.qc.edu/courses/cs341/IEEE-754hex64.html
http://babbage.cs.qc.edu/courses/cs341/IEEE-754.html

[LoSboccacc]

no that'd be your life now, there are ops to keep working with doubles until you divide them and then you get back a normal single precision float


you could probably divide by 1 to return to single precision if you aren't allowed to touch mmx stuff

otherwise cvtsd2ss

[Bumber]

no that'd be your life now, there are ops to keep working with doubles until you divide them and then you get back a normal single precision float


you could probably divide by 1 to return to single precision if you aren't allowed to touch mmx stuff

otherwise cvtsd2ss

We're allowed to use it, and that's what I'll probably end up doing. It's just another set of operations I have to figure out on my own (including the 16 byte alignment it requires), with only 3 days left to finish.