Also GPGPUs may be less accurate with fractional numbers. They do not support decimal types and double precission floating point operations are not as fast as s

Also GPGPUs may be less accurate with fractional numbers. They do not support decimal types and double precission floating point operations are not as fast as single.