Fzero vs. Solve

Hello, I am trying to solve for the grid covergence index of a simulation using the following equation. All values are known in the equation except for P.

I am solving for P in this equation, and I was wondering if fzero or using solve would be more accurate?

Here is my code for both of my versions

Function = @(P) ((abs(log(abs(F32/F21)))+(log(((r12^P)-(1*sign(F32/F21)))/((r23^P)-(1*sign(F32/F21))))))/(log(r12)))-P;

solP=fzero(Function,2);

syms P
eqn = (abs(log(abs(F32/F21))+(log(((r12.^P)-(1*sign(F32/F21)))/((r23.^P)-(1*sign(F32/F21)))))))/(log(r12))-P == 0;
solP=solve(eqn,P);

Using fzero is significantly quicker by a few minutes as I have to solve this equation for 100 cases. How to these two work, and which would be better for my application?

Thank you

[dpb]

Well, what's to argue against speedier? When you don't have need for symbols anyway?

You didn't provide data with which to compute the result but what does

Function(solP)

return? Is it not close enough to zero for your application? Is it significantly (or at least enough different than the solution form solve that your subsequent results will be markedly different using either?

[Star Strider]

If you have an idea of what the range of values of ‘P’ should be, first plot it to see how it behaves, specifically looking for multiple roots. That will help guide your choice of solvers.