search for: gcvalue

...oot as if it were live for the entire function. > > It seems to me that "gc-rootness" should not be an aspect of an alloca, as > it currently is now, but rather an aspect of a Value, similar to the way > llvm.dbg.value() works. > > Let's imagine a new intrinsic, llvm.gcvalue(), which takes as its arguments > a Value and a metadata pointer, and returns that same value as its result. > The frontend can "wrap" any value, of any type, with llvm.gcvalue(), which > is a signal to LLVM that this value is a garbage collection root. It would > then be the...

...e approach and treat every root as if it were live for the entire function. It seems to me that "gc-rootness" should not be an aspect of an alloca, as it currently is now, but rather an aspect of a Value, similar to the way llvm.dbg.value() works. Let's imagine a new intrinsic, llvm.gcvalue(), which takes as its arguments a Value and a metadata pointer, and returns that same value as its result. The frontend can "wrap" any value, of any type, with llvm.gcvalue(), which is a signal to LLVM that this value is a garbage collection root. It would then be the responsibility of LL...

...tire >> function. >> >> It seems to me that "gc-rootness" should not be an aspect of an alloca, as >> it currently is now, but rather an aspect of a Value, similar to the way >> llvm.dbg.value() works. >> >> Let's imagine a new intrinsic, llvm.gcvalue(), which takes as its >> arguments a Value and a metadata pointer, and returns that same value as its >> result. The frontend can "wrap" any value, of any type, with llvm.gcvalue(), >> which is a signal to LLVM that this value is a garbage collection root. It >> wo...