CODEWITHSUNDEEP
opengl-estextures
Albus Dumbledore
Albus Dumbledore

Reputation: 12616

Issues when processing an image using OpenGL ES 2.0

I made a simple app that takes an input image and outputs a processed one using a fragment shader. When using a 2^n image it's OK. But if I use a rectangular not-power-of-2 image I get a black stitch running from top to bottom.

This is the original:

enter image description here

This is after being processed:

enter image description here

Here's my fragment shader:

precision mediump float;
uniform vec2 uSize;
uniform sampler2D sTexture;
void main()
{
    gl_FragColor = texture2D(sTexture, vec2(gl_FragCoord) / uSize)
}

Where, uSize is a vec2 having the size of the image

Generally I can work with a power-of-2 textures, but as OGLES2 supports rectangular textures I was thinking of sparing myself some work.

Thanks!

UPDATE

Here go my vertices:

static GLfloat vVertices[] =
    {
     -1.0f,  1.0f, 0.0f,
     -1.0f, -1.0f, 0.0f,
      1.0f, -1.0f, 0.0f,

      1.0f, -1.0f, 0.0f,
      1.0f,  1.0f, 0.0f,
     -1.0f,  1.0f, 0.0f,
    };

And here is the vertex shader:

attribute vec4 vPosition;
void main()
{
   gl_Position = vPosition;
}

UPDATE

Now, that's strange. Here are the two triangles rendered separately (first using the first 3 vertices, and then using the next (last) three).

Here go for the square image (as expected):

enter image description here enter image description here

And here are how the same triangles look for the rectangle image/texture. They look bizarre:

enter image description here enter image description here

They don't look like triangles at all. Does anyone know what's happening?

Upvotes: 1

Views: 529

Answers (2)

Albus Dumbledore
Albus Dumbledore

Reputation: 12616

Seems the drivers were bad. Updating them fixed the problem.

Upvotes: 0

Bahbar
Bahbar

Reputation: 18015

Are we looking at a chunk of the image or the full thing ? Asking, because the triangle boundary does not align with the diagonal.

That said, every single non-artifact pixel is a perfect match between the 2 images, which I think rules out off-by-one errors that I thought of. Not only that, the artifact is actually based on a single of the color channels being completely wiped. (For each series of 3 pixels, the first one has R=0, the second has G=0, and the third has B=0).

All in all, this looks like an error that is completely unrelated to the shader or the drawing. Are you sure it's not something that stomps on the memory post reading ? It could also be a platform bug (h/w, driver ...). Is it reproducible on multiple platforms ?

Upvotes: 1

Related Questions