Currently the aarch64 vm options like "UseSVE" is not synced with the relative cpu feature (i.e. "sve"). It means on the sve system, if people use "-XX:UseSVE=0", the "sve" feature in JVM is not masked. This will make the IR checks in IR framework complex. It has to both check the cpu features and vm options to make sure the test is run under sve environment. Here is an example:

```
@IR(failOn = IRNode.AND_V, applyIfCPUFeature = {"asimd", "true"}, applyIf = {"UseSVE", "0"})
```

See the discussion details here: https://github.com/openjdk/jdk/pull/14533#issuecomment-1600454917