We present a survey of the cosmological applications of the next generation
of weak lensing surveys, paying special attention to the computational
challenges presented by the number of galaxies, $N_{gal} ~$ 10$^{5}$. We focus
on optimal methods with no pixelization and derive a multigrid $P^3M$ algorithm
that performs the relevant computations in $O(N_{gal} \log N_{gal})$ time. We
test the algorithm by studying three applications of weak lensing surveys -
convergence map reconstruction, cluster detection and $E$ and $B$ power
spectrum estimation using realistic 1 deg^{2} simulations derived from N-body
simulations. The map reconstruction is able to reconstruct large scale features
without artifacts. Detecting clusters using only weak lensing is difficult
because of line of sight contamination and noise, with low completeness if one
desires low contamination of the sample. A power spectrum analysis of the
convergence field is more promising and we are able to reconstruct the
convergence spectrum with no loss of information down to the smallest scales.
The numerical methods used here can be applied to other data sets with same
$O(N\log N)$ scaling and can be generalised to a sphere.