Does instructions in L1P cache run faster than that in shared RAM?

Hi All,

I am writing some time-sensitive codes (in C) on OMAPL138 DSP core. I created a new project by default settings with a main function. The .cmd file put all the codes and data onto shared RAM. But it turned out that this code runs slower than I wanted. 

I am thinking instead of putting all the codes and data on shared RAM, if I put them on L1P and L1D respectively, will it run faster? If it does, why?

Thanks,

Shallen