Andrew Haley noticed that the String indexOfChar intrinsic is broken on x86:
http://mail.openjdk.java.net/pipermail/hotspot-dev/2016-May/023218.html

I verified this by running test/compiler/intrinsics/string/TestStringIntrinsics.java with "-XX:+PrintCompilation -XX:+PrintInlining" and the log shows that C2 fails to inline the indexOfChar intrinsic:
                              @ 31 java.lang.StringUTF16::indexOfChar (30 bytes) failed to inline (intrinsic)