I have a code that performs heavy floating point calculations.
Whith gcc 2.95.3 under Linux, I have to specify -ffloat-store with -O2
to get the same result as in the normal case. I get the same behavior
with solaris x86.
With gcc 2.95.2 mingwin I get the same results with -O2 as in the normal
case (i don't need to put --float-store to prevent storing floating
point in registers). Does that mean that mingwin doesn't use floating
point registers the same way as gcc on other x86 platforms?
Pascal Quesseveur, info@...
ABAK Systèmes, ZI Le Rachat - BP 5 - 35890 LAILLE
Tel: +33(0) 2126.96.36.199, Fax: +33(0) 2188.8.131.52