Some of the tests regressed with a fix for the vectorization of shifts. The riscv cost models need to be adjusted to avoid the unprofitable optimization. The failure of these tests has been known since 2024-03-13, without a forthcoming fix, so I suggest we consider it expected by now. Adjust the tests to reflect that expectation.
So in the beginning those were added when we could avoid vectorization by higher register move costs but that's not true any more and hasn't been in a while. We still shouldn't vectorize those and, IIRC, the "burden of proof" lies more in the middle end than in the back end (because we don't vectorize with a fixed vector length). Even back when the tests started regressing we could _somehow_ prevent vectorization by adjusting costs but it never seemed right.
But can't we just keep them FAILing for now? Just because time has passed our expectation hasn't changed. To me that's more obvious than xfail.
-- Regards Robin
