Original commit message from CVS:
adds functions gst_cast_float and gst_cast_double. Will use lrint and lrintf if available and a pure c rounding macro if not.
This is where optimised asm alternatives can be placed - omega, this means you!
However you have to prove that your asm version is faster than lrintf ;)
Original commit message from CVS:
currently this creates an empty shared library, just so the header can be installed. I would appreciate help on how to do this properly with automake.
