This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: slow complex<double>'s with g++
Hmmmm, yeah... I guess it shouldn't make an 8.5 time difference.
You're right, it should be a smaller difference. The squaring is 2
muls and an add vs 4 muls and 2 adds, so it should be less than twice
as fast when you include the other portion.
The instruction ordering might be more optimal also.
8.5 times though... weird.
Brian
On 3/3/06, Greg Buchholz <greg@sleepingsquirrel.org> wrote:
> Brian Budge wrote:
> > It may be that I'm missing something, but you appear to be computing
> > different things...
> >
> > Zi = 2.0*Zr*Zi + Ci;
> > Zr = Tr - Ti + Cr;
> >
> > vs
> >
> > Z = Z*Z + C;
> >
> > No doubt the 2.0 makes a difference...
>
> Hmm. I don't get what you are saying. Squaring a complex number
> like a+ib you get...
>
> (a+ib) * (a+ib)
> a*a + iab + iab - b*b
> (a*a - b*b) + i2ab
>
> ...Are you saying a multiplication instead of an addition is the cause?
> Yes, the C version is a little more optimized, but I was surprised at
> the factor of 8.5 time difference for gcc-4.1pre021006. I guess I
> expecting something small, say a 20-50% slowdown. And you get all of
> the speed back if you just use "-ffast-math". Maybe I'm just surprised
> that the optimizations that "-ffast-math" turns on do such a good job.
>
> Greg Buchholz
>
>