On Apr 3, 12:47 pm, kuratkull <spamt...@[EMAIL PROTECTED]
> wrote:
> Hello all you happy people :)
>
> I have experience with C, Python and CLisp, but ASM is very new to me.
> So I am doing the very-fast-prime-search(which is almost a must when
> learning a new language), and I succeeded with a rather slow version
> using idivl, at first. I profiled my code and decided that the idivl
> could be replaced with the AMD 3DNow! instructions.
>
> The first failure is with %ecx = 3 and %edx=9.
> The rounding to int, just before the debug, should round the 1/3*9 to
> 3, but because of the binary representation of the float, it rounds it
> down to 2.
> Is there a way around it?
>
> MOVD %ecx, %MM0 /* divisor */
> PI2FD %MM0, %MM0 /* divisor to float */
> MOVQ %MM0, %MM2 /* copy of divisor */
> PFRCP %MM0, %MM0 /* 1/divisor */
> MOVD %edx, %MM1 /* prime candidate */
> PI2FD %MM1, %MM1 /* prime. to float */
> MOVQ %MM1, %MM3 /* copy of prime. */
> PFMUL %MM0, %MM1 /* 1/divisor * prime. */
> PF2ID %MM1, %MM1 /* answer to int ...*/
> MOVQ %MM1, %MM4 /* DEBUG for viewing value */
> PI2FD %MM1, %MM1 /* ... and back to float */
> PFMUL %MM1, %MM2 /* int(1/divisor*prime) * divisor */
> PFCMPEQ %MM2, %MM3 /*if the %MM2 and the prime candidate are
> equal, then != prime */
> MOVD %MM3, %eax
>
> I know the code is rather ugly, but remember I'm still new, and I am
> currently trying to make it work first :)
>
> Every answer(even if it's just for saying "You're fckd!") is
> appreciated.
>
> Thanks,
> kuratkull
PLEASE DONT ANSWER TO THIS THREAD, I DOUBLED BY MISTAKE, the correct
topic is:
http://groups.google.com/group/comp.lang.asm.x86/browse_thread/thread/6ca6fd909eb8195f#
Sorry again!
|
