This is defined by the LUT entries for each core. Each core can access any memory controller over the mesh network; the LUTs define which parts of a core's address space are served by which memory controller. See Section 8.1 and 8.2, on page 26/27 of the External Architecture Specification. The default mappings are also specified in the appendix, from page 32 onwards.
Thank you very much for your answer this is very helpful!