This work is a review of our main results obtained from general relativistic numerical computations of primordial black-hole formation during the radiation-dominated era of the universe. As initial conditions we consider supra-horizon-scale perturbations of a type which could have come from inflation, with only a growing component and no decaying component. We use a spherically symmetric Lagrangian code and study both super-critical perturbations, which go on to produce black holes, and sub-critical perturbations, for which the overdensity eventually disperses into the background medium. For super-critical perturbations, we confirm the results of previous work but noticing that the threshold amplitude for a perturbation to lead to black-hole formation is substantially reduced when one considers initial perturbations with a length scale sufficiently larger than the cosmological horizon, according to initial cosmological perturbations coming from inflation. For sub-critical cases, where an initial collapse is followed by a subsequent re-expansion, strong compressions and rarefactions are seen for perturbation amplitudes near to the threshold. In order to study perturbations with amplitudes extremely close to the supposed critical limit, we have introduced into the code an adaptive mesh refinement scheme. This allows us to see that scaling-law behaviour continues down to the smallest black hole masses that we are able to follow and we see no evidence of shock production such as has been reported in some previous studies and which led there to a breaking of the scaling-law behaviour at small black-hole masses. We attribute this difference to the different initial conditions used.