machine:workbench Mike$ sh tools/count_loc.sh
c (antlr parser): 1484033 loc 6 files
cpp: 418494 loc 704 files
cxx: 28484 loc 2 files
mm: 31926 loc 97 files
m: 9795 loc 37 files
py: 87652 loc 170 files
cs: 43149 loc 150 files
h: 143743 loc 928 files
Total: 2247276 (763243 without ANTLR parser)
Total Files: 2094 (1166 without headers)
The reason for the big size is the support of the full Unicode BMP for identfiers, which requires some really big state tables in the lexer.