A refined grid of a conformational potential energy surface ( PES) and a conformational entropy surface for glycine diamide was generated by ab initio molecular computations. The possible network of reaction paths was recognized in terms of the linear combinations of internal coordinates corresponding to conrotatory and disrotatory modes of motions. Such a Woodward-Hoffmann-like path selection principle was detected for the folding of this peptide from extended to some virtually cyclic structure. It seemed reasonable to assume that this principle ( or its generalized form) might be applicable to protein folding. A reaction path network was projected on the potential energy, and a continuous entropy surface was constructed under the condition of reduced dimensionality. The low entropy of the folded conformation indicated an information accumulation between 326% and 1414% with respect to the fully extended or unfolded structure. It is found that the location of existing and 'latent' critical points on the surface is revealed by the extrema and inflection points of the entropy curve.