RFR(s): 8154210: Zero: Better byte behaviour

[Severin Gehwolf]

Hi,

[new thread for JDK 9].

Could somebody please review this Zero-only fix for jdk9/dev? The April
CPU patches for OpenJDK broke this (bug 8132051). Please see the bug
for details.

The fixes are:
- Fix the jump table which is used for Zero release builds, since
  JDK-8132051 introduced a new bytecode _fast_zputfield.
- Make sure that results, which are a subword of a stack slot
  (possibly narrowed) are always at the lowest address.

The fix was originally written by Andrew Haley.

Bug: https://bugs.openjdk.java.net/browse/JDK-8154210
webrev: http://cr.openjdk.java.net/~sgehwolf/webrevs/JDK-8154210/webrev.02/

Once reviewed and pushed, we'd need a similar fix[2] for JDK 8u Zero
which is currently broken too. I think this fix needs to get pushed to
jdk9/dev directly since jdk9/hs does not have the CPU fixes yet.

Thoughts?

Thanks,
Severin

[1] http://cr.openjdk.java.net/~sgehwolf/webrevs/JDK-8154210/webrev.01/