================
@@ -15941,15 +15945,18 @@ struct PaddingCalculator {
// the partially occupied bytes in either end, if present, their bit
// intervals need to be adjusted so that they count from the MSB instead.
//
- // FIXME: For over-sized bitfields in BE, Clang allocates padding bits
- // before the occupied bits. This violates the ABI rules, which say that
- // padding should be allocated after, regardless of endianness (Itanium C++
- // ABI ยง2.4, II.1(b)). The current code accommodates for Clang's current
- // behaviour though, and bumps Start forward to skip the leading padding
- // bits.
- const uint64_t Start =
- StartBitOffset + DeclaredSizeInBits - OccupiedSizeInBits;
- const uint64_t End = Start + OccupiedSizeInBits;
+ // Occupied bits are allocated first, and any padding follows them.
+ const uint64_t Start = StartBitOffset;
+ // A _BitInt(N) bit-field may be declared wider than N, e.g.
+ // `_BitInt(5) a : 6`. CodeGen stores the value in the low-order N bits of
+ // the field, which are allocated last on big-endian (01111100), so the
+ // leading bits are not value bits. Treat the field as occupied up to the
+ // storage size of _BitInt(N) so that the value bits are never cleared.
+ // Bits beyond the storage size are padding.
----------------
vhscampos wrote:
After the changes I suggested, I'm not sure we need to have this explanation
anymore.
https://github.com/llvm/llvm-project/pull/225494
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
