xlauko wrote: > I wonder whether we should immediately flatten multidimensional arrays in CIR > instead of casting pointers. Do we need this information for something? > > Maybe it's also better for `ArrayCtor`/`ArrayDtor` to use a pair of pointers > instead of relying on the `ArrayType` size in `LoweringPrepare`. That makes > easier to implement VLAs later on and it's possible that it could improve > code size when generating cleanups in the case of destructors because we > don't have to generate a new loop for each element in the array.
Not sure about specific usecase, but in general I would be for keeping as much info as we can if does not cost us much. https://github.com/llvm/llvm-project/pull/159820 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
