The generation of granitic magmas begins with melting in the lower crust, under active participation of the underlying mantle. Thermally driven, melting is a pervasive and continuous process that develops over a wide region. In contrast, the building of a granitic pluton is highly discontinuous in time and space. Several inputs of magma, sometimes with a different chemical compositions, are focused toward a region where they accumulate, forming a large pluton, often separated by some 50 km from an adjacent one. The switch from a continuous to a discontinuous process represents a fundamental point of magma generation. It gives place to the modified model m(M-SAE), in which the mantle (m) and Melting (M) are separated from the Segregation (S), Ascent (A) and Emplacement (E) modes. Discontinuities result from non-linear processes that develop during segregation and ascent of the magma. They rely on the non-linear rheology of partially molten rocks. Thresholds control the change from a solid-like to liquid-like behaviour of the magma. In between, the rheology exhibits sudden jumps between states. Because two phases continuously coexist (matrix and melt), strain is highly partitioned between them. This may induce highly discontinuous melt segregation, which needs both pure and simple shear to develop. Melt focusing is controlled by the viscosity contrast between the two phases. It gives rise to different compaction lengths depending on the region, a partially melting source or a nearly brittle crust, where it develops. Because ascent and emplacement are discontinuous in time, this allows the crust to relax, avoiding the room problem for a pluton intruding the upper crust. Intermediate magma chambers could develop with different temperature and magma composition. They could be the place of enhanced magma mixing. Finally, the stress conditions, which differ for each tectonic setting, influence the shape of the granitic body.