[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: db5.3 package maintainance


2018-08-14 20:45 Karsten Merker:

It uses "Zero" instead of proper JIT support with Hotspot, so it's
slower, but it works.

Great news!  I wasn't aware that (basic) RISC-V support is already
part of upstream OpenJDK - from a recent posting on sw-dev
I had the impression that the existing RISC-V support for OpenJDK
is only available out-of-tree.

AFAIUI, the thing is that this "Zero" project contains Zero-assembly,
thus the name, so it's completely generic and CPU-agnostic, without
special optimisations, and it's aimed precisely for bootstrapping and
these kind of situations.  It's upstream OpenJDK, no modifications

With Adrian's confirmation that everything should work or otherwise is
considered a bug, well, it's excellent news indeed, and Java for RISC-V
mostly "a solved problem" except for Hotspot/JIT (it's been like that
for a while already, I think that May or June).

I think that the messages that you mention were some people working on
fully supporting "Hotspot" and JIT, so, the "regular Java".  I haven't
followed that closely, because this has been going back a few years and
in fact in ~2015 or so undergraduates (Martin Maas for example) were
working on some JDK, I think Jikes, for RISC-V, but sometimes there's
progress and sometimes "regress", I think that that one was eventually

But for the time being, even if Java in riscv64 is slow (or, let's say,
slower than usual :P) it does work, and it satisfies all
build-dependencies.  This is not the case of db5.3 because depending on
openjdk-8 or something, but if moved to require openjdk-10 it probably
builds fine, and the same for others that still depend on openjdk-8 or
-9 (which should happen on its own before Buster's freeze, I hope).

Manuel A. Fernandez Montecelo <manuel.montezelo@gmail.com>

Reply to: