Double precision vs Single precision "Performance vs Compatibility"

I know that most GPU projects still use single precision instead of double precision.

But this is not the most productive way to go about. Double precision is better than single precision because the results are better and performance on scale bases is more.

In my humble opinion we should be given a choice to use either SP or DP and depending of the performance of our GPU in each case we can choose the best alternative.

I have been running milkyway@home which is double precision in nature its performance is quite satisfactory. I hope other GPU projects also give DP as a choice.

In any case i feel double precision is the future, more meaning for the time spent :up: