Mail Archives: pgcc/1999/06/22/03:42:28
Marc Lehmann wrote:
> - you haven't changed anything: hardware error
But the old binary is still faster! So can't be that.
> - you have changed the compiler settings
Possible, but not probable as I became very familiar with the settings
and in one particular dir had a totally old-good just-compiled-before
version(i.e. make didn't make anything new, etc.). I rm *.o and remade
the bin and walla, different and slow.
> - you have replaced the library or compiler in between
ldd reports same exact libraries(hex code agrees).
Did you look at my other message about the contents of the binaries? It
seems odd. I KNOW that fast one was compiled with pgcc, but while the
new-slow binary has references internally about pgcc+egcs the old-fast
one only refers to egcs. I definitely used the -mcpu=pentiumpro
-march=pentiumpro options with the fast and slow one, so it was always
pgcc. No doubt.
>
> > I get:
> >
> > cmp: EOF on ./twod
> > 44387 133161 665805
>
> _MANY_ differences.
Ya, as I mentioned in the other email.
> > But as I mentioned in a new mail, they do have odd differences.
This comment is a reference to the human readable portions of the binary
I mentioned in my other email.
Can you take a look at that?
> Most probably a look at the binaries won't help.
Well, it might. One obvious difference is the one I mentioned in the
other email(and here shortly).
-Jon
- Raw text -