The code generated by:
for (i = 0; i < 10; i++) __asm__ __volatile__ ("; loop");
Seems to have grown again. Sure its slow and bulky now, but I feel so
much happier with it :-)
Thanks Dimitry.
Regards,
Steve
Steve Underwood wrote:
Hi Dmitry,
I love to see good optimisation more than most, but I think that when
you optimise
for (i = 0; i < 10; i++) __asm__ __volatile__ ("; loop");
to nothing you have gone a little too far. Especially when your own
documentation says it works. On the other hand, the speed up was much
better than the 10% to 20% you claimed. :-)
Regards,
Steve
Dmitry wrote:
Fellows,
I just committed a new gcc version.
This should be better, than is was before.
At least my tests are about 10 to 20 % faster.
Would be nice if someone can try it.
Another thing - can anybody check sin() function in libm?
cheers,
~d
P.S. no warranty of course. please report bugs as soon as possible.
