Failing jdk testcase: java/lang/Math/WorstCaseTests.java

[Roland Westrelin]

> cc'ing Rroland Westrelin, as this may be caused by some of the changes in hs24-b12. Note that jdk8/tl is only sync with b11 and that probably explains why we aren't seeing them here.

Yes, a change we made to speed up pow/exp calculation by relying on x86 fpu instructions broke this test. Is there an existing CR for this failure already? Otherwise I'll file it.

Roland.