How can I optimize my frag shader containing an if-statement within a for-loop?

The following code needs to be optimized as it takes way too much memory, but I simply can't figure out how to do it. I'm fairly new to shaders
Any help would be appriciated

Code (CSharp):

1. fixed4 frag (v2f i) : SV_Target
2.             {
3.                 fixed4 col = float4(1,1,1,1);
4.  
5.                 //How far equals one step
6.                 float step = 1 / _OcclusionMapHeight;
7.  
8.                 //Checks all pixels below the current in the occlusion Map (_MainTex)
9.                 [loop]          
10.                 for(float y = 0; y < i.uv.y; y+=step)
11.                 {
12.                     float4 colTemp = tex2D(_MainTex, float2(i.uv.x, y));
13.                                    
14.                     if(colTemp.a > 0) //checks if the pixels isn't completely transparent
15.                     {
16.                         col = float4(0,0,0,0);
17.                         return col;
18.                     }
19.                 }
20.  
21.                 return col;
22.             }

Update: Here is what the code does: I have a texture with transparent parts and none transparent parts. The shader colors every pixel from the bottom up to the first none transparent pixel white, while every pixel above those, black and transparent.
The code does exactly what I want, except it slows the fps tremendously.

Update 2: I changed the [unroll(1024)] to [loop], which improved the code significantly, although it is still rather heavy/slow. Thanks for the tip bgolus.
I also changed it in the code above, for an easier read of the current code. Should I update a code block in such a manner or not? I don't really know, if that is the normal manner.

 

I'm not a shader guru like some folks over here, but I bet that if-statement isn't going to entirely fix the issue if you think branching is the culprit here, as you unroll your loop to 1024... that's quite a lot.

You could also use the built-in TexelSize variable, just make a float4 with name (TextureName)_TexelSize, and you then have it, x and y have that 1.0 / width and 1.0 / height if I remember correctly.

I would also definitely avoid overriding the step intrinsic function with your step variable... As step is very useful, among other reasons for not doing that.

EDIT: By the way, what are you exactly trying to do? Knowing that might help others to help you.

 

Yeah ... that's going to explode the size of your shader. Try just removing that line?

 

If I remove that line, then I get the following error:
unable to unroll loop, loop does not appear to terminate in a timely manner (1024 iterations) at line 54 (on d3d11)

 

Replace it with

[loop]

?

Honestly though, what you're doing is really, really bad and is either going to be super duper slow and make a giant shader, or even more super super duper slow but a small shader.

 

Thanks for the tip, changing the line to [loop], made a signifigant difference. It changed the fps from an average of 30 to and average of 180. But it's still rather heavy, as the program without the shader has a fps of around 1000.