Omitting pull-ups at Port0 is possible, means the SRAM gets the right signals when it's addressed by the micro, and on the other hand the micro gets the right signals from your static RAM, when being read. But, if you have a look with a scope at Port0 pins, the you will notice that afterwards potentials are floating and go into a "no-man's-land". The effect of this can be an increase of supply current of all those gates which see these undefined potentials at their inputs. So, I would recommend the use pull-ups in any case. And if you take a tiny pull-up array, then this will not cost any relevant board space.

Extra supply current due to undefined input levels is the dead of any battery buffered CMOSRAM application, for instance! In this special case I would recommend the use of pull-downs instead of pull-ups, because pull-ups increase supply current just by design, whereas pull-downs don't eat extra current, but nevertheless help to guarantee defined potentials.

It's important to note, that the standard 'C51 isn't designed to run extended circuits! Have a look for instance at the /WR and /RD lines. These show the correct signals when toggling, but afterwards only a weak internal pull-up makes logical high level. Output impedance then is far away from what's needed to effectively shunt noise and interference to supply rails. Here again 4k7...10k pull-ups can help to enhance noise rejection on /WR and /RD lines. Nevertheless, extended circuits need buffers for these lines to guarantee low output impedances (in the range of 50 Ohm).

Kai