pmatos added a comment. In D104797#2855322 <https://reviews.llvm.org/D104797#2855322>, @pmatos wrote:
> After @arsenm 's commit > https://github.com/llvm/llvm-project/commit/990278d026d680942c859be70836ad34a9a716f7, > MachineOperands store LLT for size instead of uint64_t and the constructor > called from `SelectionDAG::getLoad`, calls `MachineMemOperand` which does: > > : MachineMemOperand(ptrinfo, f, > s == ~UINT64_C(0) ? LLT() : LLT::scalar(8 * s), a, > AAInfo, Ranges, SSID, Ordering, FailureOrdering) {} > > Because `s` is zero for reference types, `LLT::scalar(0)` is called, which > asserts because scalar LLTs cannot have zero size. Why do we have the check > `s == ~UINT64_C(0)`, instead of `ss == UINT64_C(0)` given opaque types and > therefore reference types are zero sized? OK, sort of replying to myself `~UINT64_T(0)` is the size for invalid types. I think at this point we need `LLT` support for opaque types. Maybe we couldn't have opaque types stored in memory but we now can with reference types in WebAssembly. Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D104797/new/ https://reviews.llvm.org/D104797 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
