jaydfox Wrote:Using gp, I computed Andrew's solution for base e, using a 50x50 matrix. (Side question: Andrew, how much faster are other libraries at solving these large matrices?)

Jay,

could you provide me your gp-code for comparision? You may put it here or send me an email.

Also I would like to check, whether your last computations can be smoothed by accelerating convergence using Euler-summation of terms. But I've only Pari/GP and maxima(the latter with extremely little experience) and I don't know how complex your computations were and how much work it would be to get it implemented.

What do you think?

Gottfried

@Jay

Wow, this is indeed interesting.

I mean theoretically it is clear that

must have period 1 for two (strictly increasing continuous) solutions

and

of

.

Because

.

But to see it as graphs is another thing

.

And I would think that it is in no way a sinus wave. However the periodic analytic functions can built by Fourier series.

Several things. I have never used pari/gp but I have 3 implementations of my super-log code and Carleman-matrix stuff in languages other than Maple/Mathematica:

- C with GMP (super-logarithm)

- Perl with BigInt (super-logarithm)

- Maxima (for Carleman-matrix)

However, I'm not sure I know where the first two are, but I just found my maxima code, so I can post that soon.

On another note, I like these graphs, although I still don't know how to use Jay's method.

I can understand the oscillation, since the further you get away from integers the "less defined" tetration is... so it makes sense to me.

I would like to post some graphs of Daniel's and my extensions, and I think I'll stick with the (critical) interval -1<x<0 since its the most well-behaved, and use the close-to-zero plots for

and

so that any oscillation is immediately obvious. My next post should have them, but I need more time to make the graphs.

Andrew Robbins

PS. That was a beautiful derivation Henryk

Gottfried Wrote:jaydfox Wrote:Using gp, I computed Andrew's solution for base e, using a 50x50 matrix. (Side question: Andrew, how much faster are other libraries at solving these large matrices?)

Jay,

could you provide me your gp-code for comparision? You may put it here or send me an email.

Also I would like to check, whether your last computations can be smoothed by accelerating convergence using Euler-summation of terms. But I've only Pari/GP and maxima(the latter with extremely little experience) and I don't know how complex your computations were and how much work it would be to get it implemented.

What do you think?

Gottfried

I posted my code here:

http://math.eretrandre.org/tetrationforu...213#pid213
I gave up on PARI/gp, the solve_right() function in SAGE (using maxima) was about 50-100 times faster. The matsolve function in gp must not be a low-level function, and I didn't bother looking for one once I got it working with maxima.