Performance regression between 8u77 and 8u91 on arm32

[Gottfried Haider]

Hello all,

I am very new to Java development, and OpenJDK as well, so please
forgive any egregious mistakes or misconceptions.

I am the author of a piece of C code that exposes the some of the
GStreamer framework through JNI to Processing for the Raspberry Pi
(armv6hf). Switching from 8u77 to 8u91 massively degrades the
performance for video playback. We're using the runtime part of
Oracle's JDK builds.

Do you have any advise for me how to troubleshoot or debug this?
Anything that changed in Oracle's builds recently? (I saw that ARM is
curiously absent from the list of their 8u92 builds. I did report to
bugs.java.com.)

Thanks in advance
Gottfried