[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#290438: produces wrong code on i386 compiling with -fno-automatic -O1



To answer upstream's question:  yes, compiling that one particular input
file (radmul.f) with -ffloat-store causes all the problems (with both
g77-3.3 and g77-3.4) to go away.

However I don't see this as a real solution, only a workaround:

> `-ffloat-store' tries to remove the extra precision by spilling
>      data from floating-point registers into memory and this typically
>      involves a big performance hit.

(from the g77 Info manual).  I'll still try to come up with a closed
test case, but it looks like this has been fixed in upstream's latest
branch?

regards,

-- 
Kevin B. McCarty <kmccarty@princeton.edu>   Physics Department
WWW: http://www.princeton.edu/~kmccarty/    Princeton University
GPG public key ID: 4F83C751                 Princeton, NJ 08544



Reply to: