[PATCH][RFC] Prevent implicit FPU usage

[Paul Brook]

On Monday 02 January 2006 02:23, Momchil Velikov wrote:
>   The patch works by noting for each function whether it uses
> floating point values.  For this is scans the types of the variables
> in ``cfun->unexpanded_var_list''.  If the function does not use
> floating point and the new flag ``-fno-implicit-fp'' is in effect
> (flag_implicit_fp == 0) it considers the floating point registers
> to be fixed.

It's possible to use FP without having floating point varables. For example:

void bar(float f);

void foo(void)
{
  bar(42.0f);
}

Paul