Concerning the subject of my message, I have a first question.
Is the ELF file loaded in flash or in RAM? Or in other words, is RLP code executed from RAM or from flash?
Neither, it is in SDRAM and there is no way to run in internal memories if this is what you will ask next 
I know that would make RLP much faster but we use those fast memories for important CLR functions.
The CPU uses a memory acceleration module. But because RLP is running from SDRAM, the CPU can’t accelerate memory access when you are busy updating the screen (also in SDRAM).
When the CPU reads an instruction and modifies the screenbuffer, the CPU is constantly switching between 2 totally different memory locations, which can’t be optimised.
This could possibly be improved if the RLP code could run direclty from FLASH.