External Package Installation on Apple M1 Chip

mathieumorlighem

I have not tried python but it should theoretically be possible to add python to configure.sh and have both MATLAB and python support.
Best
Mathieu

lawrence-bird

mathieumorlighem

Thanks Matthieu. If I give it a go, I'll report back.

I grabbed the latest svn revision (26945), but I don't see an install-3.17-mac-m1.sh for PETSc. Has this been merged into the trunk yet, or should I be looking somewhere else outside of the petsc external packages directory?

Thanks.

Lawrence

mathieumorlighem

it is not merged yet but install-3.15-mac-arm.sh should work just fine? If not, let me know and I will send it to you by email.
Mathieu

lawrence-bird

mathieumorlighem

Hi Mathieu:

I gave this a whirl, but I had no luck. However, I was able to get my python build to work with inversions (an issue I was having previously). Can I ask what MacOS version you had success with this on? I think some of my issues might be due to Monterey - My system updated last week to 12.3, and some of the reported issues that the update corrected were memory-related issues (potentially responsible for my Seg Faults). I'm definitely speculating here.

I did try to update the PETSc Version number in the install-3.15-mac-arm.sh script to install Version 3.17.0 (the latest), but when running make I was returned the following error. I just thought I'd pass this along FYI.

CXX cores/libISSMCore_la-controltao_core.lo ./cores/controltao_core.cpp:97:2: warning: 'TaoSetInitialVector' is deprecated [-Wdeprecated-declarations] TaoSetInitialVector(tao,X->pvector->vector); ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petsctao.h:226:1: note: 'TaoSetInitialVector' has been explicitly marked deprecated here PETSC_DEPRECATED_FUNCTION("Use TaoSetSolution() (since version 3.17)") static inline PetscErrorCode TaoSetInitialVector(Tao t,Vec v) { return TaoSetSolution(t,v); } ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petscconf.h:11:53: note: expanded from macro 'PETSC_DEPRECATED_FUNCTION' #define PETSC_DEPRECATED_FUNCTION(why) __attribute((deprecated)) ^ ./cores/controltao_core.cpp:104:2: warning: 'TaoSetObjectiveAndGradientRoutine' is deprecated [-Wdeprecated-declarations] TaoSetObjectiveAndGradientRoutine(tao,FormFunctionGradient,(void*)&user); ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petsctao.h:240:1: note: 'TaoSetObjectiveAndGradientRoutine' has been explicitly marked deprecated here PETSC_DEPRECATED_FUNCTION("Use TaoSetObjectiveAndGradient() (since version 3.17)") static inline PetscErrorCode TaoSetObjectiveAndGradientRoutine(Tao t,PetscErrorCode (*f)(Tao,Vec,PetscReal*,Vec,void*),void *c) { return TaoSetObjectiveAndGradient(t,NULL,f,c); } ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petscconf.h:11:53: note: expanded from macro 'PETSC_DEPRECATED_FUNCTION' #define PETSC_DEPRECATED_FUNCTION(why) __attribute((deprecated)) ^ ./cores/controltao_core.cpp:112:2: error: use of undeclared identifier 'TaoGetSolutionVector' TaoGetSolutionVector(tao,&X->pvector->vector); ^ ./cores/controltao_core.cpp:114:2: warning: 'TaoGetGradientVector' is deprecated [-Wdeprecated-declarations] TaoGetGradientVector(tao,&G->pvector->vector); ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petsctao.h:238:1: note: 'TaoGetGradientVector' has been explicitly marked deprecated here PETSC_DEPRECATED_FUNCTION("Use TaoGetGradient() (since version 3.17)") static inline PetscErrorCode TaoGetGradientVector(Tao t,Vec *v) { return TaoGetGradient(t,v,NULL,NULL); } ^ /Users/lbir0005/issm/trunk/externalpackages/petsc/install/include/petscconf.h:11:53: note: expanded from macro 'PETSC_DEPRECATED_FUNCTION' #define PETSC_DEPRECATED_FUNCTION(why) __attribute((deprecated)) ^ 3 warnings and 1 error generated. make[2]: *** [cores/libISSMCore_la-controltao_core.lo] Error 1 make[2]: *** Waiting for unfinished jobs.... make[1]: *** [install-recursive] Error 1 make: *** [install-recursive] Error 1

I got things to work using Version 3.15.2 (i.e., the version number listed in install-3.15-mac-arm.sh.

mathieumorlighem

lawrence-bird You are almost there! Can you replace src/c/cores/controltao_core.cpp with the file I sent you by email? We basically need to tell ISSM how to deal with PETSc's updated API. Nothing major.

lawrence-bird

mathieumorlighem

Hi Mathieu:

I had a flurry of activity with this today and I was able to get things working! I replaced the controltao_core.cpp file and ISSM successfully compiled.

I jumped into the inversion tutorial to make sure that inversions were behaving, and I'm getting a crash when trying to execute an inversion. When running runme.m in the inversion tutorial, I get the following error:

[proxy:0:0@MU00186957X] HYDU_create_process (utils/launch/launch.c:73): execvp error on file /Users/jenkins/workspace/macOS-Binaries-MATLAB/bin/issm.exe (No such file or directory) [proxy:0:0@MU00186957X] HYDU_create_process (utils/launch/launch.c:73): execvp error on file /Users/jenkins/workspace/macOS-Binaries-MATLAB/bin/issm.exe (No such file or directory) source /Users/lbir0005/missm/trunk//etc/environment.sh && cd /Users/lbir0005/missm/trunk//execution && rm -rf ./Square-05-17-2022-15-16-11-40903 && mkdir Square-05-17-2022-15-16-11-40903 && cd Square-05-17-2022-15-16-11-40903 && mv ../Square-05-17-2022-15-16-11-40903.tar.gz ./ && tar -zxf Square-05-17-2022-15-16-11-40903.tar.gz && source Square.queue : Signal 127

The following command is in the Square.queue file:
#!/bin/sh mpiexec -np 2 /Users/jenkins/workspace/macOS-Binaries-MATLAB/bin/issm.exe StressbalanceSolution /Users/lbir0005/missm/trunk//execution/Square-05-17-2022-15-16-11-40903 Square

Could this be related to the controltao_core.cpp file? Or maybe an absolute path issue stemming from the binary build? It's also entirely possible that I'm missing something somewhere.

Thanks for any advice on this!

Lawrence

adamjhepburn

mathieumorlighem

Hi Mathieu,

Is this file available publically? I will be compiling ISSM on an M1 pro mac and I'm trying to get ahead of these issues...

Thanks!
Adam

mathieumorlighem

Hi Lawrence,

you are close! What is happening here is that you used the MATLAB files in the binaries... (which you should not do!). I forgot to mention that with --enable-development the MATLAB files stay in place and are not moved to bin/ (see Developers section in https://issm.jpl.nasa.gov/documentation/addpath/). What you need to do to fix this is:

cd $ISSM_DIR
rm bin/* 
make install

and to launch MATLAB, you should do this:

matlab -r "addpath $ISSM_DIR/src/m/dev; devpath;"

Let me know if that works on your end!
Mathieu

mathieumorlighem

Hi Adam
you can download the Mac binaries, they should work just fine on M1. If you want to install ISSM from source, the new version of ISSM has all the files you need.
Don't hesitate to bug me/us if you run into any issues
All the best
Mathieu

Chao_Jiang

mathieumorlighem
Hi Mathieu,

I use this configure.sh file you provided to compile my ISSM.I can successfully compile and load ISSM to Matlab. However, I found that sometimeswhen I launch the matlab and run my model, the following warning will appearand then I cannot run the inversion. Is this issue related to the line ‘--without-Love --without-kml --without-Sealevelchange \’ in configure.sh file?

Thanks,
Chao

mathieumorlighem

Chao_Jiang

Hello Chao

this probably happens when you load an "old" model (which did not have this field). This bug has been corrected in the development branche and is easy to implement: open love.m and change the function loadobj so that it looks like this:

      function self = loadobj(self) % {{{
         % This function is directly called by matlab when a model object is
         % loaded. Update old properties here

         if isstruct(self),
            oldself=self;
            %Assign property values from struct
            self=structtoobj(love(),oldself);
            if isfield(oldself,'int_steps_per_layers')
               self.min_integration_steps = self.int_steps_per_layers;
            end
         end
      end% }}}

I hope this helps!
Mathieu

Chao_Jiang

mathieumorlighem
Hi Mathieu,

Thanks for your help! It works now but I got an other error when running the model.

May I know how can I fix this issue?

Many thanks!
Chao

mathieumorlighem

hum... that's weird. Can you check if md.love is actually of class love? If it is 0 you can just do

mds.love = love();

Mathieu

Chao_Jiang

mathieumorlighem
Thanks Mathieu! I think I have fixed this issue and can run my model. Much appreciate your time and help!

Regards,
Chao

justinquinn

lawrence-bird Hi Lawrence, I just submitted a fix for this to our development branch if you have access to it. Otherwise, I can walk you through the two minor changes needed to fix it.

« Previous Page