We develop algorithms for nonlinear problems with multicommodity transportation constraints. The algorithms are of the row-action type and, when properly applied,decompose the underlying graph alternatingly by nodes and edges. Hence, a fine-grain decomposition scheme is developed that is suitable for massively parallel computer architectures of the SIMD (i.e., single instruction stream, multiple data stream) class. Implementations on the Connection Machine CM-2 are discussed for both dense and sparse transportation problems. The dense implementation achieves computing rate of 1.6-3 GFLOPS. Several aspects of the algorithm are investigated empirically. Computational results are reported for the solution of quadratic programs with approximately 10 million columns and 100 thousand rows.