I'm learning shader program and need some general direction.
I want to add noise to my laser beam.
Which is the best way to handle it?
I could pre-compute an image and pass it to the shader. I could then use the image to change the opacity and easily animate the smoke by changing the offset of the texture lookup.
I could also generate noise in the shader and do the same thing the texture was used for.
Is it generally better to avoid I/O to the graphics card or the opposite?
I'd go with the first approach. It's not only easier to pull of but also more elegant. While this kind of overhead on the GPU mustn't necessarily be avoided (it's not fatal to go with your second approach), I'd rather save it up for something more worthwhile.
Now if the question was about whether to use a high-poly mesh or tessellate/displace a low-poly one in realtime, I'd definitely say go for the later. Such would be a case in which utilizing a shader rather than a pre-baked asset would be most advisable, as you'd otherwise waste a lot of good bandwidth.
depending if you are fillrate bound or memory bound, you can chose from that. but other than that, but i would go with the texture. that way you can let artist control the noise aswell.
