It does indeed look like the only change is to CFLAGS, turning off the
use of more floating point registers.
The changes did not seem to cause GCC to produce any different code,
however. I used objdump -d to disassemble the object modules before
and after the CFLAGS adjustments and the code was exactly the same.
FICL seems to use floating point in only one place, and it's
#if'd with a FICL_WANT_FLOAT test and doesn't get compiled.
-Matt
Matthew Dillon
<[EMAIL PROTECTED]>
