Whilst I have no direct experience with the Keil compiler, I think you may be doing something fundamentaly wrong - Keil is probably (arguably??) the #1 compiler for 8051 and a great many other people have got even the demo version to work (within it's limitations). I've been using a number of different C compilers for different cpu's and each has it's little idiosyncrasies that you need to understand and work with. Whenever things don't make sense - stop and read the manual. It's sounds like the same old song, but boy have I been caught out sooooooo many times! I can tell you a story about string conversion macros in VC++ for windows CE! I went full circle only to find, once I'd fixed the problem - it was spelt out in the help file - only I didn't care to read the fine print.


Anyway, reverting back to your lcd problem, in your code you would set E, then output the data - bad move.
Output the data THEN assert E. Delay a little, then clear E THEN change the data for the next cycle. This is for your data setup & hold time - see data sheet.

If you really don't like Keil, try Raisonance - it gave me some weird errors when trying to compile code that was written for IAR compilers - until I removed the offending line (far removed from what the errors were telling me) and then it was happy. Not to say it is a bad compiler - just saying it has some quirks that confuse the unwary. We could always talk about the gcc compiler!!!!!!