I'm not sure about the C8051F120 specifically, but I'd be surprised if its low-going open-drain mode wouldn't sink the 5 mA that a 1K resistor would source. You can probably get by just fine with 10K pullups on D[7..0] and RS, but R/w and E should probably be pulled up with 1k or at most 3.3 k-ohms, else their risetime may be to slow to work properly.

Keep in mind, too, that the E pulse has a minimum width, as do other signals. The E width on the standard HD44780, IIRC, is 400 ns. You'll have to see that that width is never reduced. All the other control signals should have a longer width, as they're supposed to be valid at both ends of E. Data can be allowed to be valid only at the falling edge of E, though there is a setup time and a hold time. You should know what those are before you proceed.

RE