Ultrafast ultrasound imaging allows observing rapid phenomena; combined with 3-D imaging it has the potential to provide more accurate analysis on organs which leads, at the end, to better diagnosis. Coherent compounding using diverging waves is commonly used to reconstruct high quality images on large volumes while keeping frame rate high enough to allow dynamic analysis. In practice, the virtual sources that drive the diverging waves are often distributed by a deterministic way: following a regular grid, concentric rings and spirals. Even though those deterministic distributions can offer various trade-off in terms of imaging performance, other distributions can be considered to improve imaging performance. It is herein suggested to look at alternative virtual sources distributions for optimizing the lateral resolution and the secondary lobes level on several PSFs by means of a multiobjective genetic algorithm. The optimization framework leads to seven pseudo-irregular distributions of virtual source distributions that have not yet been found in the literature. An analysis on the imaging performance with a simulated phantom shows that these new distributions offer different trade-offs between lateral resolution and contrast, respectively measured on point-like reflectors and anechoic cysts. As an example, one of these optimized distributions improves the lateral resolution by 16% and gives equivalent contrast values on cysts, when compared to a concentric-rings-based distribution.