I wanted to clip pixels for which the depth is below the previously sampled depth to avoid complex light calculations. But I wasn't able to measure a significant performance boost. I guess this is due to the fact that I sort most of my geometry into the solid pass already, I guess.