Looks like Math.log is no longer intrinsified into C2 IR, so dead code / constant fold examples no longer work. It seems to have started since JDK 9 and JDK-8152907 that now defers the computation of log10 to quick stub. It does, however, limits the compiler optimizations around the Math.log, and samples fail. The samples should not rely on JDK internals, and should just use the custom written payload.
- duplicates
-
CODETOOLS-7902806 Update JMH samples for new JDKs
-
- Closed
-
- relates to
-
JDK-8152907 Update for x86 tan and log10 in the math lib
-
- Resolved
-
- links to
-
Review openjdk/jmh/60