Spell of Mastery Progress

Ok. The new voxel rendering algorithm performs 1.4 times faster. Not much, but voxel models can also be split into 16-bit chunks, which locally use only 16-bit addressing. That should speedup the algorithm further, if the bottleneck is with the memory bandwidth.

I've also discovered that GCC's __builtin_popcount is slower than a lookup table, since it branches into subroutine __popcountdi2. Why GCC doesnt use the x86 popcnt opcode? No idea. No idea. But it probably would be easier to answer why Stallman molests little children.

Surprisingly the hardest part of working with lex/yacc was location tracking, which is a bit obscure. Even more surprising was discovering that simple SEXP lists from Lisp can be used to hold C/C++ parsed code, although one will have to add a bit of metainfo for holding the locations. At least for my purposes, since I want to modify them. But during writing the actual compiler, one will still need locations to report errors. Like, well, syntax errors. While C++ syntax is notorious for its monstrosity (surpassing even Haskell in difficulty), even the plain C is not context free, since the parser needs access to the context to properly create the aforementioned SEXP.


Anyway, now I can parse structs, typedefs and functions, while skipping the rest of the syntax. That should be just enough to mod the C for my purposes.

$ cat example.c
int foo(int n, int m) {
return n*m+1;
}

enum russia {govna,poesh,suka};

int bar(int n) {
return n+1;
}


$ cat example.c | ../lib/cext
function:
row,col: 1,0
type: int
name: foo
arg0: int n
arg1: int m
function:
row,col: 7,0
type: int
name: bar
arg0: int n


No. I don't need enums.