Make gcc ignore the distinction between +0.0 and -0.0, this allows it to

change x+0.0 to x. Though it alone does not allow gcc to remove x*0.0,
that would also require -ffinite-math-only.

Originally committed as revision 12252 to svn://svn.ffmpeg.org/ffmpeg/trunk

configure

@@ -1824,6 +1824,7 @@ elif enabled optimizations; then
     fi
 fi
 check_cflags -fno-math-errno
+check_cflags -fno-signed-zeros
 
 # PIC flags for shared library objects where they are needed
 if enabled shared; then