nigelp-xmos wrote: XCore: yes the current setting of the new flag, the default false value, is right for XCore. Misaligned loads can generate a function call. We do not have benchmarks as such, but comparing the code generated for the test files touched here, there is an improvement. In particular bitfield-ir.cpp is a big improvement.
In bitfield.cpp, two misaligned loads remain. One disappears if I remove the struct packed attribute. The other remains: N0::read31(). This is not a problem, and is not a degradation, it's a function call intended for the purpose; just mentioning for information. https://github.com/llvm/llvm-project/pull/65742 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits