Giving green a value somehow clears red

Question

1.00/5 (1 vote)

See more:

I have a pixel shader in which I have the following part

imgcolor.b = 0;
imgcolor.r = 1 - dist1 / 5;
imgcolor.g = 0;

if I make the following change

imgcolor.b = 0;
imgcolor.r = 1 - dist1 / 5;
imgcolor.g = 1 - min(1,max(0,dist2));

where both dist1 and dist2 are float the red value becomes 0 in parts where it shouldn't be. Why does this happen? How do I prevent it?

First case: https://www.dropbox.com/s/bq2irg4nsjarne2/good.jpg?dl=0[^]
Second case: https://www.dropbox.com/s/jy0z2zk56h9jl96/wrong.jpg?dl=0[^]
Link to shader's source(with comments): http://pastebin.com/3128bpkC[^]

UPDATE: Thank you al for your suggestions. What I ended up doing was to rewrite the code that calculates dist1 & dist2 and the issue somehow dissapeared.

Posted 31-Dec-14 3:31am

mostwanted4

Updated 1-Jan-15 22:38pm

v2

Add a Solution

Comments

Richard MacCutchan 31-Dec-14 10:13am

What are the values of dist1 and dist2?

mostwanted4 31-Dec-14 10:36am

I don't know exactly but in the test from the pictures they're between 0 and sqrt(2) * 40. Although, I'm not sure it matters because I've tested with multiple values from that range and it doesn't produce the same results.

[no name] 31-Dec-14 11:05am

That is great if you don't know it exactly. Does it also mean you do not know what you are doing?

Richard MacCutchan 31-Dec-14 11:39am

:)

[no name] 31-Dec-14 11:57am

That is from self made experience, I know this from myself....sometimes trying and playing with things I don't know/understand
:-)

mostwanted4 31-Dec-14 11:41am

Well, I can't know because they depend on the RANDOM noise from my texture and apart from setting each pixel's color to their values compressed to [0,255] I am not aware of any debugging method.

And I would like to assure you that even though I'm not an expert in shaders, I have quite some programming experience to give me a decent understanding of what's happening in the shader.(Unless any bug slipped by)

Richard MacCutchan 31-Dec-14 11:40am

With information that vague it is unlikely that anyone can help you. You need to be specific about where the error happens and what values are used to produce the error.

mostwanted4 31-Dec-14 11:44am

I'm trying to create a custom voronoi shader, dist1 is the distance to the closest point and dist2 the second closest distance to another point. And now I'm trying to visualise dist1 and dist2.

[no name] 31-Dec-14 11:49am

Ok that is some real Input. I don't know "voronoi shader" have first to read about it. For a first overview (but that is only a vague guess) you are fighting "only" with scales 0...1 vs. 0....255.

What type is "imgcolor"?

mostwanted4 31-Dec-14 11:54am

I didn't mention what kind of shader I was creating because I though it's irrelevant to the question, as I am struggling with the shader itself, but with this little part. I know that the colors in glsl are in 0..1 range, but(out of curiosity) I tried to put other values and it worked, it didn't throw any error and the red color didn't disappear as in the case above where I use a variable.

imgcolor is the pixel's final color stored in a float4(the one that I return in the pixelshader)

[no name] 31-Dec-14 12:00pm

Still only a guess (did ot have time to read for "voronoi shader": Pay Attention on any calculation that it will be double. Means in if something should be double dont use 1/xyz use 1.0 / xyz.

Anyway for first test I suggest to calculate anything in double (and make sure that no integer artefacts will affect). Aftwerwards you can optimize all the stuff.

mostwanted4 31-Dec-14 12:10pm

OK, so converting every number to a x.x or x.xf form doesn't make any difference. Also replacing every float variable with its double counterpart doesn't make any difference(float2 -> double2, float -> double etc.)

[no name] 31-Dec-14 14:03pm

float/double I think is not the question. My guess was more float/double vs. int.
Bruno

mostwanted4 31-Dec-14 12:21pm

I also tried manually setting dist2 to random values from -1000 to 1000 with both float and double and no value reproduced the "error".

Graham Breach 31-Dec-14 13:10pm

I'll add my guesswork: It looks like dist2 might not always be set to a value, so your code might be failing at the imgcolor.g assignment. Since shader code doesn't necessarily execute in the order that it is written, that might be preventing the imgcolor.r assignment too.

There are a lot of "might"s in that, but I hope it helps. You could also try using clamp() instead of min(max()) too.

mostwanted4 31-Dec-14 13:53pm

Initializing it(and "dist" from my code) makes no difference.

Could you elaborate(or give me a link) on "doesn't necessarily execute in the order that it is written"?

Graham Breach 31-Dec-14 15:41pm

The compiler can reorder instructions for optimization. I can't seem to find a decent reference that says that, but the /O0 and /O1 flags to FXC disable instruction reordering: http://msdn.microsoft.com/en-us/library/windows/desktop/bb509709.aspx

1 solution

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Sergey Alexandrovich Kryukov · Answer 1 · 2014-12-31T15:14:00

Solution 1

This is pure and very simple arithmetic. If, for example dist1 becomes 5, your r will be zero. Note that actually it could be non-zero value close to zero. Use the debugger and find the reason. This is all to be said based on your limited information.

Good luck,

—SA

Posted 31-Dec-14 15:14pm

Sergey Alexandrovich Kryukov

Comments

mostwanted4 2-Jan-15 4:41am

Maybe I wasn't clear enough. Let me give you an example: The first code draws entire red circles, while the second only draws part of those red circles, and the only change is the line where I give green a value.

Sergey Alexandrovich Kryukov 2-Jan-15 10:17am

Do you think that the miracle happened. I don't know R value you assign, because I have no information on that. I just tell you that your shown part of code does not show that you are doing it right or wrong. Just use the debugger. Have you done that?
—SA