> I'd have to look at the two asm outputs more closely to see why the
> main loop (RC5_KEY_CHECK) is so lame with newer gcc's but I am not
> too much of an an asm guru and so I doubt I'd see the problem even
> if it was staring me in the face. For what its worth, the number
> of asm lines in the fast one is 1067 and 1269 in the slow one.
> That in itself is an increase of 19% more instructions. Hrrm,
> I wonder if gcc-2.7.2 is slower for other things as well, or if it
> is just something this app triggers...
I wouldn't count the result of this particular benchmark too high:
The RC5 code exercises just a very small part of the instruction
set. It is even very Intel and 68k friendly because these CPUs have
rotate instructions which the RISCs I know don't have. That's why
these machines, especially the new Intel ones look very good in
comparison with normally lightning fast Alphas.
Similar with GCC. The RC5 code hits may well hit a particularly weak
point in the GCC/GAS/ld complex. I wouldn't put too much weigth on
the above results though they are obviously a proof that something is
really wrong with GCC.
I did btw, some benchmarks with the RC5 code for the 48 challenge using
GCC 2.7.2 and a snapshot of GCC on different architectures - the results
have changed, sometimes even extremly.
Ralf