Re: [arts-users] Another newbie build question
Hi Geoff, In the past, "Native" didn't work on ARM Macs, only on x86 machines. Out of curiosity, since you mentioned it, I just tried to compile with Native again and it seems to be fixed now (tested with gcc-13, clang 17 from homebrew and clang 16 from conda on M1). In theory, "Native" should give at least a slight speed boost, but we haven't done any specific benchmarks comparing Release vs Native. Cheers, Oliver > On 14. Dec 2023, at 14:15, Smith, Geoffrey CIV USN NRL (7233) Washington DC > (USA) wrote: > > All — > > Since I am starting with a fresh build and am moving to the dev version, I am > curious about another possibility. Is there any experience with the native > build option for Apple ARM64 M2 Max architecture? Certainly at first I will > not be desperate to eke out every last drop of performance, but since I am > starting fresh I would not mind exploring the available options. > > Any advice, positive or negative, is appreciated. > > -geoff > > — > Dr. Geoffrey B. Smith > Head, Ocean Applications Section > Remote Sensing Division > Naval Research Laboratory > > geoffrey.sm...@nrl.navy.mil > 202.404.6324 smime.p7s Description: S/MIME cryptographic signature
[arts-users] Another newbie build question
All — Since I am starting with a fresh build and am moving to the dev version, I am curious about another possibility. Is there any experience with the native build option for Apple ARM64 M2 Max architecture? Certainly at first I will not be desperate to eke out every last drop of performance, but since I am starting fresh I would not mind exploring the available options. Any advice, positive or negative, is appreciated. -geoff — Dr. Geoffrey B. Smith Head, Ocean Applications Section Remote Sensing Division Naval Research Laboratory geoffrey.sm...@nrl.navy.mil 202.404.6324
Re: [arts-users] New ARTS user/Tests question
Thanks for the download link, and suggestion for moving to the dev version noted. I will go ahead and do that before I go any farther. Thanks! -geoff > On Dec 14, 2023, at 3:54 AM, Lemke, Oliver > wrote: > > Hi Geoff, > > Those tests fail because they need additional input files from the > arts-xml-data package. For arts 2.4 you can find the download links here: > https://www.radiativetransfer.org/tools/ > Unpack the arts-xml-data-2.4.0.tar.gz in the same folder where your arts > folder is located, rename it from arts-xml-data-2.4.0 to arts-xml-data and > cmake will detect the location automatically. > > Since you wrote that you're just starting with ARTS and RT, I have another > recommendation for you. Instead of using ARTS 2.4, consider to use the > current development version of ARTS instead. We're working on a 2.6 release > which will happen soon(ish). Since there are quite a few changes, especially > a stronger move towards Python integration, I think it would be easier for > you to start with that. We also made improvements to the documentation > especially on how to get started with ARTS. Also, once 2.6 is released you > then don't have to adapt your scripts or relearn things for the new version. > > You can install the latest pre-release (2.5.12) directly in a Miniforge3[1] > Python environment and don't need to compile it yourself. You can find the > current documentation here: > > https://atmtools.github.io/arts-docs-2.6/installation.html > > Cheers, > Oliver > > [1] https://github.com/conda-forge/miniforge > > >> On 13. Dec 2023, at 19:13, Smith, Geoffrey CIV USN NRL (7233) Washington DC >> (USA) wrote: >> >> Hello All — >> >> I am working with a fresh install of 2.4.x on a MacBook Pro M2 running macOS >> 14.1.2. >> >> Thanks to a tweak by Oliver Lemke I successfully (I believe) built ARTS and >> am testing. >> >> I installed (I believe successfully) PyARTS as well. >> >> I ran ‘make check’ and 100% of all tests passed successfully. >> >> I then ran ‘make check-all’ and 96% of tests passed. I have zero experience >> with radiative transfer or ARTS, but am getting started. >> >> Can anyone offer any insight regarding the 6 tests that failed? Should I be >> concerned and try and fix? Are they known/common failures? OS specific? >> >> Here is the relevant part of the report: >> >> 96% tests passed, 6 tests failed out of 160 >> >> Total Test time (real) = 546.90 sec >> >> The following tests FAILED: >> 27 - arts.ctlfile.slow.artscomponents.absorption.TestIsoRatios (Failed) >> 28 - python.arts.ctlfile.slow.artscomponents.absorption.TestIsoRatios >> (Failed) >> 41 - arts.ctlfile.slow.artscomponents.clearsky.TestBatch (Failed) >> 42 - python.arts.ctlfile.slow.artscomponents.clearsky.TestBatch (Failed) >> 69 - arts.ctlfile.slow.artscomponents.nlte.testRotationalConvergence (Failed) >> 70 - python.arts.ctlfile.slow.artscomponents.nlte.testRotationalConvergence >> (Failed) >> Errors while running CTest >> make[3]: *** [CMakeFiles/check-all] Error 8 >> make[2]: *** [CMakeFiles/check-all.dir/all] Error 2 >> make[1]: *** [CMakeFiles/check-all.dir/rule] Error 2 >> make: *** [check-all] Error 2 >> >> >> Thanks much for any info and/or reassurances — >> >> Geoff > > > >
Re: [arts-users] New ARTS 2.4 build Mac OS 14.1.2
Hi Geoff,
I strongly recommend to not use the Apple Clang compiler because it does not
support OpenMP. This will prevent your ARTS calculations from running on more
than a single core.
You can pass the compiler when running cmake in the build directory (start with
an empty build directory before switching compilers):
cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=gcc-13
-DCMAKE_CXX_COMPILER=g++-13 ..
Cheers,
Oliver
> On 13. Dec 2023, at 18:27, Smith, Geoffrey CIV USN NRL (7233) Washington DC
> (USA) wrote:
>
> Oliver - Thanks for the help! My next question was going to be ‘please
> refresh my memory on how to use the home-brew gcc instead of the apple
> default’, but now I can re-clone the repo and test that instead?
>
> -Geoff
>
>> On Dec 13, 2023, at 4:36 AM, Lemke, Oliver
>> wrote:
>>
>> Hi again,
>>
>> I've committed a fix to the ARTS 2.4.x branch which should resolve the
>> problem.
>>
>> Cheers,
>> Oliver
>>
>>
>>> On 13. Dec 2023, at 08:04, Lemke, Oliver
>>> wrote:
>>>
>>> Hi Geoff,
>>>
>>> Unfortunately, the binary_function has been removed in recent versions of
>>> the Clang compiler. As a workaround, you can change `binary_function` in
>>> line 48 of src/sorting.h to `std::__binary_function`. That fixes the
>>> compilation with Clang.
>>>
>>> Alternatively, you can use gcc-13 from homebrew as a compiler which still
>>> works without modifying the code.
>>>
>>> Cheers,
>>> Oliver
>>>
>>>
On 12. Dec 2023, at 20:27, Smith, Geoffrey CIV USN NRL (7233) Washington
DC (USA) wrote:
Hi —
Sorry if this question is too basic; I did some searching of the mail
archive and did not see a solution, but I certainly could have missed it.
I am trying to build a fresh clone of ARTS 2.4 on a clean, updated,
MacBook Pro M2 running 14.1.2. I am getting an error during make…. I am
attempting to include the relevant parts here:
[ 12%] Building CXX object src/CMakeFiles/matpack.dir/lin_alg.cc.o
[ 13%] Building CXX object src/CMakeFiles/matpack.dir/logic.cc.o
In file included from /Users/gbsmith/local/arts/src/logic.cc:38:
/Users/gbsmith/local/arts/src/sorting.h:48:26: error: no template named
'binary_function'; did you mean '__binary_function'?
class IndexComp : public binary_function {
^~~
__binary_function
/Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX14.0.sdk/usr/include/c++/v1/__functional/binary_function.h:49:1:
note: '__binary_function' declared here
using __binary_function = __binary_function_keep_layout_base<_Arg1, _Arg2,
_Result>;
^
1 error generated.
make[2]: *** [src/CMakeFiles/matpack.dir/logic.cc.o] Error 1
make[1]: *** [src/CMakeFiles/matpack.dir/all] Error 2
I am not sure what to try next except move to the development version.
Any advice appreciated; any additional information that might help please
just let me know.
Thanks much,
Geoff Smith
Remote Sensing Division
Naval Research Laboratory
geoffrey.sm...@nrl.navy.mil
smime.p7s
Description: S/MIME cryptographic signature
Re: [arts-users] New ARTS user/Tests question
Hi Geoff, Those tests fail because they need additional input files from the arts-xml-data package. For arts 2.4 you can find the download links here: https://www.radiativetransfer.org/tools/ Unpack the arts-xml-data-2.4.0.tar.gz in the same folder where your arts folder is located, rename it from arts-xml-data-2.4.0 to arts-xml-data and cmake will detect the location automatically. Since you wrote that you're just starting with ARTS and RT, I have another recommendation for you. Instead of using ARTS 2.4, consider to use the current development version of ARTS instead. We're working on a 2.6 release which will happen soon(ish). Since there are quite a few changes, especially a stronger move towards Python integration, I think it would be easier for you to start with that. We also made improvements to the documentation especially on how to get started with ARTS. Also, once 2.6 is released you then don't have to adapt your scripts or relearn things for the new version. You can install the latest pre-release (2.5.12) directly in a Miniforge3[1] Python environment and don't need to compile it yourself. You can find the current documentation here: https://atmtools.github.io/arts-docs-2.6/installation.html Cheers, Oliver [1] https://github.com/conda-forge/miniforge > On 13. Dec 2023, at 19:13, Smith, Geoffrey CIV USN NRL (7233) Washington DC > (USA) wrote: > > Hello All — > > I am working with a fresh install of 2.4.x on a MacBook Pro M2 running macOS > 14.1.2. > > Thanks to a tweak by Oliver Lemke I successfully (I believe) built ARTS and > am testing. > > I installed (I believe successfully) PyARTS as well. > > I ran ‘make check’ and 100% of all tests passed successfully. > > I then ran ‘make check-all’ and 96% of tests passed. I have zero experience > with radiative transfer or ARTS, but am getting started. > > Can anyone offer any insight regarding the 6 tests that failed? Should I be > concerned and try and fix? Are they known/common failures? OS specific? > > Here is the relevant part of the report: > > 96% tests passed, 6 tests failed out of 160 > > Total Test time (real) = 546.90 sec > > The following tests FAILED: > 27 - arts.ctlfile.slow.artscomponents.absorption.TestIsoRatios (Failed) > 28 - python.arts.ctlfile.slow.artscomponents.absorption.TestIsoRatios (Failed) > 41 - arts.ctlfile.slow.artscomponents.clearsky.TestBatch (Failed) > 42 - python.arts.ctlfile.slow.artscomponents.clearsky.TestBatch (Failed) > 69 - arts.ctlfile.slow.artscomponents.nlte.testRotationalConvergence (Failed) > 70 - python.arts.ctlfile.slow.artscomponents.nlte.testRotationalConvergence > (Failed) > Errors while running CTest > make[3]: *** [CMakeFiles/check-all] Error 8 > make[2]: *** [CMakeFiles/check-all.dir/all] Error 2 > make[1]: *** [CMakeFiles/check-all.dir/rule] Error 2 > make: *** [check-all] Error 2 > > > Thanks much for any info and/or reassurances — > > Geoff smime.p7s Description: S/MIME cryptographic signature
