The different regions are connected to the bus matrix with separate connections. This way, all can be used at the same time. If the only user is CPU, this does not matter, but as you start using DMA channels, the cpu has to wait on dma and dma has to wait for cpu, unless you are not sharing an interface.
Keep in mind that the SRAM1-3 segments are after each other in address space. You can just define one segment in your linker script and use that as one large chunk if the separation is not needed for you.