llvm dev - Feb 2016 - [cfe-dev] [lldb-dev] [3.8 Release] Release Candidate 2 source and binaries available

On 6 February 2016 at 01:02, C Bergström <cfe-dev at lists.llvm.org>
wrote:
> I worked on the port of llvm-OpenMP-formally-known-as-Intel to Aarch64
> - can it be included there as well? (I'm not sure what's precisely
> involved - I'm willing to do my best fixing any bugs which pop up -
Hi Chris,

That'd entail me enabling them for AArch64. I have to say, I've done
that a few months ago and it all went smoothly, but I didn't know
anyone was interested (and there were some failures in the OMP tests).

I can enable them on the next release candidate and keep them on for
the next releases, but I can't support it myself, so if you're willing
to handle that, I'm ok with releasing them.

Also, it's not trivial to run the OMP tests, making it a bit harder to
monitor regressions between releases. For now, I'll just build them
and hope you or someone can spot regressions, but ultimately, it'd be
good to have it as part of the release process, maybe inside the
test-suite?

cheers,
--renato

  Original Message  
From: Renato Golin
Sent: Monday, February 8, 2016 21:13
To: C Bergström
Cc: Hans Wennborg; llvm-dev; release-testers at lists.llvm.org; cfe-dev;
openmp-dev (openmp-dev at lists.llvm.org); LLDB Dev
Subject: Re: [cfe-dev] [lldb-dev] [3.8 Release] Release Candidate 2 source and
binaries available

On 6 February 2016 at 01:02, C Bergström <cfe-dev at lists.llvm.org>
wrote:
> I worked on the port of llvm-OpenMP-formally-known-as-Intel to Aarch64
> - can it be included there as well? (I'm not sure what's precisely
> involved - I'm willing to do my best fixing any bugs which pop up -
Hi Chris,

That'd entail me enabling them for AArch64. I have to say, I've done
that a few months ago and it all went smoothly, but I didn't know
anyone was interested (and there were some failures in the OMP tests).

I can enable them on the next release candidate and keep them on for
the next releases, but I can't support it myself, so if you're willing
to handle that, I'm ok with releasing them.

Also, it's not trivial to run the OMP tests, making it a bit harder to
monitor regressions between releases. For now, I'll just build them
and hope you or someone can spot regressions, but ultimately, it'd be
good to have it as part of the release process, maybe inside the
test-suite?

‎

Imho it's critical to get parallel programming working on ARMv8 ( even if
it's crappy OMP) to start. Please enable it and I'll run the tests
against our internal QA and we can informally handle testing / guarding against
regressions.

Thanks

On 8 February 2016 at 22:56,  <cbergstrom at pathscale.com>
wrote:
> Imho it's critical to get parallel programming working on ARMv8 ( even
if it's crappy OMP) to start. Please enable it and I'll run the tests
against our internal QA and we can informally handle testing / guarding against
regressions.
Ok, I'll do it with rc3.

cheers,
--renato

On 8 February 2016 at 22:56,  <cbergstrom at pathscale.com>
wrote:
> Imho it's critical to get parallel programming working on ARMv8 ( even
if it's crappy OMP) to start. Please enable it and I'll run the tests
against our internal QA and we can informally handle testing / guarding against
regressions.
So, I got some dependency errors while building OpenMP on my AArch64 box:

cd /home/rengolin/devel/llvm/release/rc3/Phase3/openmp/src &&
/usr/bin/perl
/home/rengolin/devel/llvm/release/rc3/openmp.src/runtime/tools/check-depends.pl
--os=lin --arch=aarch64
--expected="libdl.so.2,libgcc_s.so.1,libpthread.so.0"
/home/rengolin/devel/llvm/release/rc3/Phase3/openmp/src/libomp.so
check-depends.pl: (i) Dependencies:
check-depends.pl: (i)     libc.so.6
check-depends.pl: (i)     libdl.so.2
check-depends.pl: (i)     libpthread.so.0
check-depends.pl: (x) Unexpected dependencies:
check-depends.pl: (x)     libc.so.6

I remember those problems when I tried last year, and they should be
not too hard to work around. However, we're late in the release and I
don't want to delay it even more for an experimental feature.

It should be fine to compile it separately and upload later. But it
would be better if you could work on the release and OMP scripts to
make it work out of the box, so we can do the next releases without
delay.

I'll continue without OMP for now.

cheers,
--renato