> AFAIK , C compilers do all floating point operations with doubles,
> even when adding two floats they are converted to doubles and then
> the result is converted back to float, so using doubles should be faster,
> but modern compilers may behave differently.
As for SPARC chip, there are two sets of FP instruction - for single and
double precision. Division and sqrt are twice faster for single and others
have the same execution time. Of course it's bad when single and double
operands are mixed - this requres some time to convert to the same format.
Single precision is enough for decoder - I've tested it and I'm sure.
And SPARC version works faster with single prec. also allowing some
optimization tricks.
It seems to be enough for encoder too, but I'm not sure about overflowing.
-Leonid
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
