[RFR] jdk9.0.4+12 + functionality

[Edward Nevill]

Hi all,

There is a long history as to why we have ended up with two arm64 ports and two arm32 ports. I do not intend to go into the history of how and why we got here.

The situation at the moment is as follows

In mainline code we have

aarch64 port
arm64 port
arm32 port

These are supported (I don't want a semantic argument about what that means) from jdk9 onwards.

In aarch32 port area we have

jdk8u
jdk9

These consist of the aarch32 port of the template interpreter contributed by Linaro and the C1 compiler contributed by Azul.

At the moment the only JIT support for jdk8 is in the aarch32 jdk8u repo and this has been maintained primarily by Azul.

Azul have quite reasonably added C2 support along with other features (intrinsics, SIMD support, additional floating point support, ...) and and looking to push this to the aarch32/jdk9 repo.

I agree with the sentiment expressed on the list that we do not want an additional arm32 port going into mainstream.

What I propose is

- I ask ops to create a new jdk9u repo in the aarch32 area (the aarch32  C2 work is based on jdk9u and it makes more sense to do this than update the jdk9 repo to jdk9u). The jdk9 repo can then be frozen.

- The Azul C2 aarch32 port can be pushed here as a staging area.

- We evaluate the additional features in the aarch32 C2 port to see what could be usefully extracted and pushed to mainline.

Andrey, if you could draw up a concise list of what additional features there are in the aarch32 C2 port which you think could be usefully contributed to mainstream.

All the best,
Ed.