(In reply to Richard Biener from comment #1) Thanks for your information. * I hoped somehow that my test section would not be optimised ���away��� (for a specific size comparison). * The additional compilation parameters ���-fno-������ might mean a target conflict for my use case then.