<div dir="auto">Banana Pi M64 comes with 64bit arm. They are ~100 €/$. Probably, Jonas can set up the testing once he is back😀<div dir="auto"><br><div dir="auto">-Joachim</div></div></div><div class="gmail_extra"><br><div class="gmail_quote">Am 20.12.2017 10:16 schrieb "Cownie, James H via Openmp-dev" <<a href="mailto:openmp-dev@lists.llvm.org">openmp-dev@lists.llvm.org</a>>:<br type="attribution"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">> I think we don't have access to an ARM machine to test ourselves<br>
There was someone on Twitter the other day running OpenMP code on his Raspberry Pi (using "Raspbian",<br>
I assume since he had a Linux environment) so it may be possible to add ARM to your testing environment<br>
at little monetary cost. (Of course the time cost to get it all working may be non-trivial, but the<br>
hardware cost seems under 100Euro).<br>
<br>
That would be a 32bit ARM, not AARCH64, but still better than nothing.<br>
<br>
-- Jim<br>
<br>
Jim Cownie <<a href="mailto:james.h.cownie@intel.com">james.h.cownie@intel.com</a>><br>
SSG/DPD/TCAR (Technical Computing, Analyzers, and Runtimes)<br>
Tel: <a href="tel:%2B44%20117%209071438" value="+441179071438">+44 117 9071438</a><br>
<br>
-----Original Message-----<br>
From: Openmp-dev [mailto:<a href="mailto:openmp-dev-bounces@lists.llvm.org">openmp-dev-bounces@<wbr>lists.llvm.org</a>] On Behalf Of Jonas Hahnfeld via Openmp-dev<br>
Sent: Tuesday, December 19, 2017 6:52 PM<br>
To: Paul Osmialowski <<a href="mailto:pawel.osmialowski@foss.arm.com">pawel.osmialowski@foss.arm.<wbr>com</a>><br>
Cc: <a href="mailto:openmp-dev@lists.llvm.org">openmp-dev@lists.llvm.org</a><br>
Subject: Re: [Openmp-dev] Enabling OMPT by default<br>
<br>
Am 2017-12-19 15:15, schrieb Paul Osmialowski:<br>
> Nah, it blocks forever on nested_lwt.c when I started it on 96-cores<br>
> ARMv8.1 machine.<br>
<br>
Can you get us a stack trace? I think we don't have access to an ARM<br>
machine to test ourselves but we might be able to guess based on the<br>
implementation.<br>
<br>
> BTW, the missing part of callbacks.h for tests I added like this:<br>
><br>
> diff --git a/runtime/test/ompt/callback.h<br>
> b/runtime/test/ompt/callback.h<br>
> index 8481fb4..9a560aa 100755<br>
> --- a/runtime/test/ompt/callback.h<br>
> +++ b/runtime/test/ompt/callback.h<br>
> @@ -105,6 +105,11 @@ ompt_label_##id:<br>
> #define print_possible_return_<wbr>addresses(addr) \<br>
> printf("%" PRIu64 ": current_address=%p\n",<br>
> ompt_get_thread_data()->value, \<br>
> ((char *)addr) - 8)<br>
> +#elif KMP_ARCH_AARCH64<br>
> +// On AArch64 the NOP instruction is 4 bytes long.<br>
> +#define print_possible_return_<wbr>addresses(addr) \<br>
> + printf("%" PRIu64 ": current_address=%p\n",<br>
> ompt_get_thread_data()->value, \<br>
> + ((char *)addr) - 4)<br>
> #else<br>
> #error Unsupported target architecture, cannot determine address<br>
> offset!<br>
> #endif<br>
<br>
Interesting that we don't have a load instruction at all for AArch64.<br>
Sigh, that's why we have to customize it for every architecture...<br>
<br>
><br>
> On 18/12/2017 13:56, Jonas Hahnfeld via Openmp-dev wrote:<br>
>> Does it pass the tests? I doubt that because it needs modifications as<br>
>> Joachim pointed out (had to do the same for PPC64).<br>
>><br>
>> Jonas<br>
>><br>
>> Am 2017-12-18 14:52, schrieb Pawel Osmialowski:<br>
>>> Guys,<br>
>>><br>
>>> For more than a year I was always building libomp with<br>
>>> -DLIBOMP_OMPT_SUPPORT=ON for our AArch64 machines and I'd happily see<br>
>>> it enabled by default.<br>
>>><br>
>>> Cheers,<br>
>>> Paul<br>
>>><br>
>>><br>
>>> On Mon, 2017-12-18 at 14:23 +0100, Jonas Hahnfeld wrote:<br>
>>>> Adding Paul, maybe he can help here. Though switching the default<br>
>>>> for<br>
>>>> x86/x86_64 and ppc64 doesn't mean it has to work on AArch64 for the<br>
>>>> 6.0<br>
>>>> release, it will just stay deactivated as for MIPS(64).<br>
>>>><br>
>>>> Jonas<br>
>>>><br>
>>>> Am 2017-12-18 13:02, schrieb Cownie, James H via Openmp-dev:<br>
>>>> > The ARM folks should chip in. This should clearly be tested on<br>
>>>> > AARCH64<br>
>>>> > as well!<br>
>>>> ><br>
>>>> > -- Jim<br>
>>>> ><br>
>>>> > Jim Cownie <<a href="mailto:james.h.cownie@intel.com">james.h.cownie@intel.com</a>><br>
>>>> > SSG/DPD/TCAR (Technical Computing, Analyzers, and Runtimes)<br>
>>>> > Tel: <a href="tel:%2B44%20117%209071438" value="+441179071438">+44 117 9071438</a><br>
>>>> ><br>
>>>> > -----Original Message-----<br>
>>>> > From: Openmp-dev [mailto:<a href="mailto:openmp-dev-bounces@lists.llvm.org">openmp-dev-bounces@<wbr>lists.llvm.org</a>] On<br>
>>>> > Behalf<br>
>>>> > Of Joachim Protze via Openmp-dev<br>
>>>> > Sent: Monday, December 18, 2017 11:58 AM<br>
>>>> > To: Openmp-dev <<a href="mailto:openmp-dev@lists.llvm.org">openmp-dev@lists.llvm.org</a>><br>
>>>> > Subject: [Openmp-dev] Enabling OMPT by default<br>
>>>> ><br>
>>>> > Dear all,<br>
>>>> ><br>
>>>> > OpenMP 5.0 will introduce OMPT, the OpenMP tools interface. The<br>
>>>> > current<br>
>>>> > implementation in the LLVM/OpenMP runtime is at the level of<br>
>>>> > OpenMP<br>
>>>> > TR6,<br>
>>>> > as published in November (with some known issues, were we have<br>
>>>> > patches<br>
>>>> > under review).<br>
>>>> > Currently, this feature is deactivated by default<br>
>>>> > (LIBOMP_OMPT_SUPPORT).<br>
>>>> > According to overhead measurements performed by Intel and on some<br>
>>>> > of<br>
>>>> > our<br>
>>>> > machines, this feature introduces a worst case runtime overhead of<br>
>>>> > ~4%,<br>
>>>> > but for most cases basically no overhead.<br>
>>>> > As I understood, some of the build-bots run the tests when the<br>
>>>> > feature<br>
>>>> > is enabled. We locally test the changes continuously with different<br>
>>>> > compiler versions (icc/gcc/clang), but only on x86_64 architecture.<br>
>>>> > Further, this feature was tested on ppc64.<br>
>>>> ><br>
>>>> > How should we proceed to enable this feature by default, and get it<br>
>>>> > tested on other architectures?<br>
>>>> > For the 6.0 release, I would like to enable OMPT at least for the<br>
>>>> > supported platforms x86/x86_64 and ppc64.<br>
>>>> ><br>
>>>> ><br>
>>>> > We would also appreciate collaboration with vendors/users of other<br>
>>>> > platforms, to ensure the code is running there.<br>
>>>> ><br>
>>>> > Typically, only the architecture specific offset needs to be added:<br>
>>>> > <a href="https://github.com/llvm-mirror/openmp/blob/master/runtime/test/ompt" rel="noreferrer" target="_blank">https://github.com/llvm-<wbr>mirror/openmp/blob/master/<wbr>runtime/test/ompt</a><br>
>>>> > /callback.h#L95<br>
>>>> ><br>
>>>> > Looking forward to your feedback,<br>
>>>> ><br>
>>>> > Joachim<br>
>>>> > ______________________________<wbr>_________________<br>
>>>> > Openmp-dev mailing list<br>
>>>> > <a href="mailto:Openmp-dev@lists.llvm.org">Openmp-dev@lists.llvm.org</a><br>
>>>> > <a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/openmp-dev</a><br>
>>>> > ------------------------------<wbr>------------------------------<wbr>-----<br>
>>>> > ----<br>
>>>> > Intel Corporation (UK) Limited<br>
>>>> > Registered No. 1134945 (England)<br>
>>>> > Registered Office: Pipers Way, Swindon SN3 1RJ<br>
>>>> > VAT No: 860 2173 47<br>
>>>> ><br>
>>>> > This e-mail and any attachments may contain confidential material<br>
>>>> > for<br>
>>>> > the sole use of the intended recipient(s). Any review or<br>
>>>> > distribution<br>
>>>> > by others is strictly prohibited. If you are not the intended<br>
>>>> > recipient, please contact the sender and delete all copies.<br>
>>>> > ______________________________<wbr>_________________<br>
>>>> > Openmp-dev mailing list<br>
>>>> > <a href="mailto:Openmp-dev@lists.llvm.org">Openmp-dev@lists.llvm.org</a><br>
>>>> > <a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/openmp-dev</a><br>
>>> IMPORTANT NOTICE: The contents of this email and any attachments are<br>
>>> confidential and may also be privileged. If you are not the intended<br>
>>> recipient, please notify the sender immediately and do not disclose<br>
>>> the contents to any other person, use it for any purpose, or store or<br>
>>> copy the information in any medium. Thank you.<br>
>> ______________________________<wbr>_________________<br>
>> Openmp-dev mailing list<br>
>> <a href="mailto:Openmp-dev@lists.llvm.org">Openmp-dev@lists.llvm.org</a><br>
>> <a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/openmp-dev</a><br>
______________________________<wbr>_________________<br>
Openmp-dev mailing list<br>
<a href="mailto:Openmp-dev@lists.llvm.org">Openmp-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/openmp-dev</a><br>
------------------------------<wbr>------------------------------<wbr>---------<br>
Intel Corporation (UK) Limited<br>
Registered No. 1134945 (England)<br>
Registered Office: Pipers Way, Swindon SN3 1RJ<br>
VAT No: 860 2173 47<br>
<br>
This e-mail and any attachments may contain confidential material for<br>
the sole use of the intended recipient(s). Any review or distribution<br>
by others is strictly prohibited. If you are not the intended<br>
recipient, please contact the sender and delete all copies.<br>
______________________________<wbr>_________________<br>
Openmp-dev mailing list<br>
<a href="mailto:Openmp-dev@lists.llvm.org">Openmp-dev@lists.llvm.org</a><br>
<a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/openmp-dev" rel="noreferrer" target="_blank">http://lists.llvm.org/cgi-bin/<wbr>mailman/listinfo/openmp-dev</a><br>
</blockquote></div></div>