"kuratkull" <spamtrap@[EMAIL PROTECTED]
> wrote in message
news:38432b4f-1f55-45ed-ba66-deedac70d272@[EMAIL PROTECTED]
> Thanks for the long post, I really appreciate it Phil. Although the
> point was not about the algorithm itself, but about any means to have
> the app run faster with "simple" ASM tricks. But It doesn't matter any
> more, since I have implemented a Sieve of E. in 55 lines which finds
> 10 million primes in 0.54 seconds, so I guess I managed to write fast
> code anyway(because it's possible to write a rather slow S. of E.
> too).
> The 3DNow! problem is fixed by simply not using it, so I guess this
> topic has served its purpose.
It's funny how everyone who writes a Sieve program thinks it's fast.
Fot timing results on my old 1.4 GHz Athlon, see
http://home.comcast.net/~kmbtib/Sieve30h/time1000000000.dat
and similar files. Writing a fast sieve takes a lot of effort
because there are so many features that can be added, each of
which contributes significantly to performance.
--
write(*,*) transfer((/17.392111325966148d0,6.5794487871554595D-85, &
6.0134700243160014d-154/),(/'x'/)); end
|
