If you just wanted to write your own first compiler, I'm not sure that the 8052 would be a good target to start with.
The 8052 imposes particular challenges to the 'C' compiler writer - especially the very limited hardware stack, which renders the "normal" approach to parameter passing and automatic local variables impractical.
Might be better to start on something more "mainstream" before you start worrying about extensions for bit-addressing, SFRs, DATA/IDATA/PDATA/XDATA, etc, etc.