Hi, I can show you slide from some marketing presentation:
Up to 2x performance is only theoretical value (both cores executes from cache accessing only CPU registers).
DPM is basically common dual-core MCU whilst LSM is basically common single-core MCU from performance point of view. In practice there is access to shared resources by both cores thus overall performance will be much lower and it is application dependent.
I don't know document comparing the performance in detail. You may see following white paper discussing DPM/LSM pros and cons:
http://www.nxp.com/docs/en/white-paper/MPC564XLWP.pdf
"The decision between lockstep mode and decoupled parallel mode depends on the application, intended safety concept and the required performance, as well as the acceptable software effort."
