Re: [newbie] Floating point

["Jason L. Wright" <jason@thought.net>]

On Fri, May 16, 2003 at 02:10:21PM -0400, Will Backman wrote:
> How does this address the goal of "correctness"?
> 
What's "correct" about floating point in the kernel?  It's difficult
to do correctly and gives very close to zero additional functionality
over scaled arithmetic.  The correct thing to do is keep the heavy
duty math stuff in -user- space and let the kernel perform integer
arithmetic for feeding your user program pages of data =)

Think about this.  If a floating point instruction is not implemented
by the hardware, but the compiler generates the instruction, it will
trap to the kernel to perform the operation.  Now you get to deal with
yet another kernel traps into itself set of poo.