.fx shaders and HLSL - GameCreators Forum

Author

Message

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 7th Sep 2006 05:32 Edited at: 7th Sep 2006 06:00

Link

I want to learn HLSL, so I've been looking at some tutorials on shaders online. I was trying this tutorial:

http://www.neatware.com/lbstudio/web/hlsl.html

but I couldn't seem to get it working - the object would just act as if hidden when I applied the shader.

Then, I took a look at some of the shaders in the free shader pack (in particular, I looked at one of Evolved's shaders) and it seemed like the structure of the .fx is nothing like what is described here. So, are there different forms of HLSL? And, what's the difference? Also, does DBPro suppport all of them?

And, while I'm at it, I would like to ask all of the people here who can write shaders how they learned - can you suggest a book or web page?

UPDATE:
I just tried this:
http://www.cfxweb.net/modules.php?name=News&file=article&sid=1305
tutorial, and again, it just hid my object. I've got an nvidia geforce 5200, brand new, so I don't think it's my hardware... also, all of the examples from the free shader pack work for me, so shouldn't these work too?

EDIT:
Oh, yea, here's the code I'm using (press space to apply test.fx):

+ Code Snippet

global shader$="test.fx"
global tex_in_shader=1

set window on
set window size 800,600
sync on
sync rate 60
autocam off
make object cube 1,10
make object plain 2,50,50
xrotate object 2,90
position object 2,0,-15,0
fix object pivot 2
zrotate object 1,45
position object 1,0,0,0
fix object pivot 1
position camera 0,10,30
point camera 0,-5,0
set camera range .1,300
color_image(1,255,0,0)
texture object 1,1
color_image(2,0,255,0)
texture object 2,2
hide light 0
make light 1
set directional light 1,-1.0,-1.0,-1.0
`set effect on 1,shader$,tex_in_shader

d=0
repeat
   if spacekey() then set effect on 1,shader$,tex_in_shader
   d=wrapvalue(d+1)
   yrotate object 1,d
   sync
until escapekey()

end

function color_image(n,r,g,b)
   create bitmap 1,1,1
   dot 0,0,rgb(r,g,b)
   get image n,0,0,1,1
   set current bitmap 0
   delete bitmap 1
endfunction

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 7th Sep 2006 13:05 Edited at: 7th Sep 2006 13:07

Link

Yes, I've found the available material on shaders very confusing. Some are written in HLSL and some (like Evolved's) are written in asm. There are also several ways of passing parameters to the functions within shaders which can be confusing at first. Best to examine several shaders which work and adopt whichever style suits you. Practice with very simple ones first.

A very good way to start is to download FX Composer from nVidia (it's free) and try the shaders that come with it. Most of them work with DBP quite nicely once you've set up the constants correctly in DBP. You can learn a lot just by experimenting with the examples and seeing what effects various code changes have. After that try RenderMonkey from ATI (but I've found the shaders there need alteration before they work properly in DBP - AND the output FX files look awful with very long variable names like My_First_Effect_Technique_Two_Pass_One_Pixel_Shader ...).

However, I found I couldn't make much progress till I got a book. I've tried:

"Shaders for Game Programmers and Artists" by Sebastien St-Laurent

and

"Programming Vertex and Pixel Shaders" by Wolfgang Engel.

Neither book is completely satisfactory - but both were very helpful. Engel's book gives more accurate explanations, but is rather heavy reading. Both books, especially St-Laurent's book, are littered with typos which make the tutorials hard to follow in places - and you need to "read between the lines" to correct the typos in the list of intrinsic functions and their arguments!

There may be better books around but one of those should help a lot - but be prepared to be left figuring some things out for yourself.

Have fun - and post questions here. Someone can probably help.

Edit: corrected a typo!

Back to top

Profile PM Email

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 7th Sep 2006 13:13 Edited at: 7th Sep 2006 13:15

Link

Quote: "Oh, yea, here's the code I'm using (press space to apply test.fx):"

Could you give us the "test.fx" file as well? The problem could be there rather than your DBP file. Will try it for you if you post it.

Edit: And, of course, any texture (image) files you are using...

Back to top

Profile PM Email

Pulsar Coder

20

Years of Service

User Offline

Joined: 3rd Jan 2004

Location:

Posted: 7th Sep 2006 20:41

Link

Engel's a very good book.

There's another one that can be useful: "The COMPLETE Effect and HLSL Guide"

~Pulsar Coder

Back to top

Profile PM Email Website

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 8th Sep 2006 00:39 Edited at: 8th Sep 2006 00:53

Link

Quote: "Could you give us the "test.fx" file as well? The problem could be there rather than your DBP file. Will try it for you if you post it.

Edit: And, of course, any texture (image) files you are using..."

I keep deleting the test.fx file and creating a new one with each tutorial, and none of them seem to work.

I'll post a couple of pieces of test code I tried in the .fx files, though...

And, the textures are 1-pixel images generated at runtime using bitmaps...

UPDATE:

Here are a couple of the test.fx files I used:

+ Code Snippet

ps_1_1
def c0,1.0f,0.0f,0.0f,1.0f
mov r0,c0

That's supposed to fill the screen with red, c0 stores red, and r0 is the current pixel...

+ Code Snippet

vs_1_1
dcl_position v0
m4x4 oPos, v0, c0
mov oD0, c4

Very simple - supposed to just render the object normally, I think...
This is the HLSL version of that same code:

+ Code Snippet

struct a2v { 
	float4 Position : POSITION; 
};
struct v2p {
	float4 Position : POSITION;
};
void main(in a2v IN, out v2p OUT, uniform float4x4 ModelViewMatrix) 
{
	OUT.Position = mul(IN.Position, ModelViewMatrix); 
}

All of these just hid the object... also, if you can point out an error in any of these (they're copied almost directly from their respective tutorials) I would appreciate it, some of the tutorials look terribly old and probably out of date...

EDIT: Oh, btw, run the program, then press the spacebar to apply the shader... escape to exit...

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 8th Sep 2006 01:20

Link

The snippets you've just given are only fragments of a complete shader. If you just want a shader to output red (purely as an example, of course) I'll write one now and post it in a few minutes.

Back to top

Profile PM Email

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 8th Sep 2006 01:40

Link

Here you go. First the DBP program:

+ Code Snippet

sync on: sync rate 60: sync

autocam off
position camera 0,0,-250

make object cube 1, 100
load effect "GGSimplyRed.fx",1,1
set object effect 1,1

repeat
  inc angle#, 0.3
  rotate object 1, angle#,angle#,0
  sync
until spacekey()

end

And now the FX file (which you should call "GGSimplyRed.fx"):

+ Code Snippet

//
// Green Gandalf's Simply Red Shader
//
// Created 7 September 2006
//
// Colours a DBP object red regardless of any lights in the scene.
// 

matrix wvp : WorldViewProjection;

struct VS_INPUT
{   float4 Pos          : POSITION; };

struct VS_OUTPUT
{   float4 Pos          : POSITION; };

VS_OUTPUT GGRedVShader(VS_INPUT In, VS_OUTPUT Out)
{   Out.Pos = mul(In.Pos, wvp);
    return Out;
}

struct PS_INPUT  // not really needed but you would usually have some input
{ };
struct PS_OUTPUT
{  float4 col : COLOR; };

PS_OUTPUT GGRedPShader(PS_INPUT In, PS_OUTPUT Out)
{  Out.col = float4 (1.0,0.0,0.0,1.0);
   return Out;
}

technique t0
{   pass p0
    {  VertexShader = compile vs_2_0 GGRedVShader();
       PixelShader = compile ps_2_0 GGRedPShader();
    }
}

I used Vs and PS2.0.0 but it should work with most earlier versions as well.

Any questions, just let me know.

Back to top

Profile PM Email

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 8th Sep 2006 02:09 Edited at: 8th Sep 2006 02:10

Link

Just looking at your code fragments more closely.

The first one should output something. I think it needs something like the following as the FX file:

+ Code Snippet

technique t0
{
  pass p0 
  {  //  no vertex shader
     PixelShader = asm
     {  // Initialise Shader
        ps_1_1
        def c0,1.0f,0.0f,0.0f,1.0f
        // output the colour
        mov r0,c0
     };
  }
}

[Edit

The above works when called by the following DBP program:

+ Code Snippet

sync on: sync rate 60: sync

autocam off
position camera 0,0,-250

make object cube 1, 100
load effect "Media/Bmad6 query.fx",1,1
set object effect 1,1

repeat
  inc angle#, 0.3
  rotate object 1, angle#,angle#,0
  sync
until spacekey()

end

No more time to look at your other fragments now. Will look again tomorrow. Good luck.

Back to top

Profile PM Email

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 8th Sep 2006 06:56

Link

Wow... that's really simple, but about 1000 times more helpful than any of the tutorials I saw... thanks!!

Now, a quick question about vertex shaders versus pixel shaders:

In your example, the vertex shader takes a position and returns a position and the pixel shader just returns a color (i understand that it could also take various values); my question is: what are all the possible values you can have as input and output from these 2 different types of shaders?

And, I've already got NVidia's FX Composer (but don't know how to use it

) and ps and vs 2.0 work well for my gpu...

And, another thing: I see a lot of places in these tutorials that talk about 4x4 matrices, which aren't that complicated by themselves, but what are they used for and what values do they hold? For example, in your example, you include the WorldViewProjection matrix and call it wvp. I understand that this is the world matrix multiplied by the view and projection matrices (right?), but what are the world, view, and projection matrices?

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

coolgames

18

Years of Service

User Offline

Joined: 26th Sep 2005

Location: Oregon, USA

Posted: 8th Sep 2006 07:31

Link

I believe when you multiply the objects position by the matrix it returns the objects position on the screen. But I don't really know for sure.

Play my game "Hang Glider"! http://forum.thegamecreators.com/?m=forum_view&t=70854&b=8

Back to top

Profile PM Email

VR2

19

Years of Service

User Offline

Joined: 14th Mar 2005

Location:

Posted: 8th Sep 2006 11:32

Link

The world matrix is the matrix that translates the 3D object into World space.

The view matrix is the matrix that describes the camera position (as an inverse of a world matrix)

The projection matrix describes how the vertices are transformed from 3D world coordinates to 2D screen coordinates.

Bearing in mind that the view matrix is the inverse of the world matrix you can see that

World * View = How an object would appear to the viewer (camera).

thus World * View * Projection = How an object would appear to the viewer (camera), projected onto a 2d screen.

Back to top

Profile PM Website

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 8th Sep 2006 14:25

Link

I'm not sure I understand it entirely, but that gives me a good enough idea of what's going on that I think I can work with it...

Now, how do you add textures to a shader and how do you use pixels or coordinates from the in your output?

In one of Evolved's shaders I see the input TEXCOORD and the outputs TEXCOORD0 and TEXCOORD1, but I assume these store uv data only, not rgba data? And, I assume that these different outputs are different textures? Also, why do the outputs have a number after them, but not the input? Could you change the input to TEXCOORD0 with the same result?

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

Green Gandalf

VIP Member

19

Years of Service

User Offline

Joined: 3rd Jan 2005

Playing: Malevolence:Sword of Ahkranox, Skyrim, Civ6.

Posted: 8th Sep 2006 17:42 Edited at: 8th Sep 2006 17:48

Link

Have a look at the project and fx file in the attached zip file. The FX file uses two textures, one is used as a "noise" file ("magic clouds.png") and the other ("lava colours.png") is used as a simple colour lookup table depending on the value found in the noise lookup. Look at the code and see if you can work out what is going on. You can also lookup 3D textures to get interesting effects with a bit more work.

The output from the vertex shader has two variables labelled as TEXCOORDs, i.e. TEXCOORD0 and TEXCOORD1 - and there are two input variables similarly labelled for the pixel shader. These labels tell the compiler two things: first, that the variables are unrestricted in value (i.e. not COLORs which are restricted to the range 0 to 1 for each component), and second, how the output variables from the vertex shader link to the input variables of the pixel shader (i.e. the variable labelled as TEXCOORD1 from the vertex shader will be matched to the variable labelled TEXCOORD1 in the pixel shader). You can see what goes wrong if you swap the 1 and 0 in the names in the pixel shader and re-run the program.

[Edited silly error in zip file.]

Attachments

Login to view attachments

Back to top

Profile PM Email

Pulsar Coder

20

Years of Service

User Offline

Joined: 3rd Jan 2004

Location:

Posted: 8th Sep 2006 23:18

Link

Good example. Easy to understand ...

~Pulsar Coder

Back to top

Profile PM Email Website

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 9th Sep 2006 06:34 Edited at: 9th Sep 2006 06:41

Link

The example looks great, Green Gandalf. It's very easy to understand, but I have a couple of questions about the code:

1 - What do the UIMin and UIMax values for scrollspeed x and y do? Do they somehow limit the possible values?

2 - At what rate does Time increase? By milliseconds?

3 - I assume tex1D returns the position (1,n) from a texture given a sampler and n, and tex2d returns the position (x,y) from a texture given a sampler and a float2 with the tex coords?

4 - When you define a sampler, you define a number of values. ADDRESSU=WRAP and ADDRESSV=WRAP are self-explanatory, but what about the rest (ADDRESSW=WRAP, MAGFILTER=LINEAR, MINFILTER=LINEAR, MIPFILTER=LINEAR)?

EDIT:
Also, why does it have all those little lines when it renders? It looks like some sort of wierd noise pattern.

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

Bmad6

18

Years of Service

User Offline

Joined: 25th Aug 2005

Location: Virginia, United States

Posted: 9th Sep 2006 17:10

Link

And, one more thing - if I change the pixel shader so that it says "Out.color = In.color;" it just renders the cube all black. Is this because of the way darkbasic professional's texture object command works?

Harvard's Law, as Applied to Computers:
Under the most rigorously controlled conditions of pressure, temperature, volume, humidity and other variables, the computer will do as it d*mn well pleases.

Back to top

Profile PM Email Website

Sorry your browser is not supported!

DarkBASIC Professional Discussion / .fx shaders and HLSL

Attachments