For certain match input, hobbes might generate really large IR functions and it takes forever for llvm to parse
be conservative with "inlining" when translating DFA to IR if current DFA might have too many states
stop FunctionInliningPass() if there is one function in this module is already too large
There is a flag buildColumnwiseMatch() designed for preventing large DFA generation. However, the code path introduced by this flag has never been properly tested, and it has upper row count limit which is not that hard to exceed
For certain
match
input, hobbes might generate really large IR functions and it takes forever for llvm to parseFunctionInliningPass()
if there is one function in this module is already too largeThere is a flag
buildColumnwiseMatch()
designed for preventing large DFA generation. However, the code path introduced by this flag has never been properly tested, and it has upper row count limit which is not that hard to exceed