x86: check xcr0 in AOT compiles

[BradleyWood]

Issue: #19408

[dsouzai]

J9::X86::CPU::detect pretty much looks like a copy of OMR::X86::CPU::detect; why not just call OMR::X86::CPU::detect (from J9::X86::CPU::detect) and update the OMR copy with any differences?

[BradleyWood]

OMR::X86::CPU::detect returns only features enabled by OMR. If OpenJ9, tries to use extra features that OMR doesn't have enabled, then we can't just manually enabled these features in OpenJ9 if CPUID supports them. OpenJ9 doesn't know why a feature is or is/isn't enabled in OMR. It could be because OMR hasn't enabled it, or because the target CPU/OS doesn't support it. We end of back at the same spot, forced to check XCR0 again.

Your suggestion probably won't cause any issues unless someone tries to enabled a feature in OpenJ9 that OMR doesn't enable. Fixing this kind of thing would require changes in both OMR and OpenJ9, which I don't think there is time for at the moment. There are a few customers using windows server 2012 (which is technically out of support) that this affects.

[JamesKingdon]

Hi Brad, thanks for fixing this. We have another customer case open at the moment for this problem and it looks like we might see more after WebSphere started shipping the recent JVM release.
@0xdaryl for info

[dsouzai]

At the risk of sounding unsympathetic, the original issue was opened almost a year ago and the last customer update was back in Sept, so I personally think we should fix this properly. However, I don't want to explicitly prevent this from going forward, so I'll defer this to @0xdaryl or @hzongaro.