So the problem still persists.. although I thought I'd clarify some things
The sram is async.
the transfers to the ext sram are done via dma, while the reads are done directly using a pointer to the memory.
again this works for all addresses on the sram that doesn't address using the upper 2 msb's.
I have also verified the signals coming out of the psoc for address bits 16 and 17, and they seem to behave just as any of the other address bits are behaving. If I disconnect both the 16 and 17 wires, and then plug them in individually, whatever port on the psoc that was addressing the 16th or 17th bit freezes the data on my lcd.