Accelerate Unsafe CAS Intrinsics on Power and X

[IBMJimmyk]

Adds lbarx, lharx, stbcx_r and sthcx_r instruction support for Power. These instructions are not currently in use but will be needed for future enhancements targetting CAS on byte and short length data.

Updates willBeEvaluatedAsCallByCodeGen to reflect the new acceleration of Unsafe compareAndExchange methods on X.

Edit:
X changes are dependent on this OpenJ9 PR:
eclipse-openj9/openj9#19991
Behavior of inlineCompareAndSwapNative and inlineCompareAndSwapObjectNative in J9TreeEvaluator.cpp must match behavior in willBeEvaluatedAsCallByCodeGen in OMRCodeGenerator.cpp

[IBMJimmyk]

@hzongaro The compiler/x/codegen/OMRCodeGenerator.cpp change goes along with the OpenJ9 changes for accelerating Unsafe CAS on X and unfortunately is inside OMR instead of OpenJ9. I also need to double check if the change to callDoesAnImplicitAsyncCheck is correct.

@zlwang I need a review on the Power instructions being added.

Also, would it be preferred if this PR was split into two? I was working on the Power and X accelerations at the same time and the end result just happened to be 2 fairly disjoint changes to OMR.

[hzongaro]

I think the changes for x/codegen/OMRCodeGenerator.cpp and compiler/optimizer/RedundantAsyncCheckRemoval.cpp look correct.

[IBMJimmyk]

@zlwang did you get a chance to look over these changes?

[zl-wang]

LGTM

[hzongaro]

Sorry for the late comment. May I ask you to reformat your commit comment so that lines do not exceed 72 characters in length, per https://github.com/eclipse/omr/blob/master/CONTRIBUTING.md#commit-guidelines?

[hzongaro]

Jenkins build all

[IBMJimmyk]

I fixed the commit message.

[hzongaro]

Jenkins build all

[hzongaro]

@IBMJimmyk, there is a failure in the Linux ppc64le testing:

14:39:12  31: �[0;32m[----------] �[m14 tests from Special/PPCDirectEncodingTest
14:39:12  31: free(): invalid next size (normal)
14:39:12  31/31 Test #31: compunittest ......................Child aborted***Exception:   1.72 sec

Is that related to the Binary Encoder test change that is part of this pull request?

[IBMJimmyk]

It doesn't seem like it should be. I changed PPCTrg1MemEncodingTest, PPCRecordFormSanityTest and PPCMemSrc1EncodingTest not PPCDirectEncodingTest. Furthermore, the test passed under the AIX build and it should be the same test.

[hzongaro]

It looks like the failure in PPCDirectEncodingTest is due to known issue #6571.