[LLVMdev] Is LLVM expressive enough to represent asynchronous exceptions?

Andrew Trick atrick at apple.com
Mon Jun 13 14:43:37 PDT 2011


On Jun 13, 2011, at 11:37 AM, Jakob Stoklund Olesen wrote:

> The explicit CFG forces the ordering of trapping instructions. Something that may not be a language requirement?

That's a very good thing. Reordering trapping instructions changes the semantics of control flow regardless of whether the language allows it. A transformation that (legally) changes control flow needs to explicitly update the CFG.

-Andy



More information about the llvm-dev mailing list