Re: [AMBER-Developers] Release candidate 5 for Amber20/AmberTools20

From: B. Lachele Foley <lfoley.ccrc.uga.edu>
Date: Thu, 30 Apr 2020 00:25:27 +0000

One of the people trying the Mac install had trouble installing cmake. He's using the instructions at the link below. I don't know Mac well at all, but it seems that cmake-gui down in Applications/CMake.app is symlinked to some other folder and it won't run properly to do the install.

Are there other/better instructions to give him?

https://tudat.tudelft.nl/installation/setupDevMacOs.html

:-) Lachele

Dr. B. Lachele Foley (she/her/hers)
Associate Research Scientist
Complex Carbohydrate Research Center
The University of Georgia
Athens, GA USA
lfoley.uga.edu
http://glycam.org
________________________________
From: Scott Brozell <sbrozell.iqb.rutgers.edu>
Sent: Wednesday, April 29, 2020 7:55 PM
To: AMBER Developers Mailing List <amber-developers.ambermd.org>
Subject: Re: [AMBER-Developers] Release candidate 5 for Amber20/AmberTools20

[EXTERNAL SENDER - PROCEED CAUTIOUSLY]


Hi,

I built and tested RC3 with intel/19.0.5 intelmpi/2019.5.
And i am at 36% in make install with RC5.
I searched my dir tree for mpigi and got no hits.


Details FWIW:

Currently Loaded Modules:
  1) xalt/latest 3) gcc-compatibility/8.4.0 5) mkl/2018.0.3 7) cmake/3.16.5
  2) modules/au2018 4) intel/19.0.5 6) intelmpi/2019.5

(Note to those actually reading this far; mkl/2018.0.3 was unintentional,
so this may be a rare test of OSC or Amber treatment of out of sync
mkl and compilers ...)

# Assume this is Linux:

cmake $AMBER_PREFIX/amber20_src \
    -DCMAKE_INSTALL_PREFIX=$AMBER_PREFIX/amber20 \
    -DCOMPILER=MANUAL \
  -DCMAKE_C_COMPILER=$OSC_CC \
  -DCMAKE_CXX_COMPILER=$OSC_CXX \
  -DCMAKE_Fortran_COMPILER=$OSC_F90 \
    -DMPI=TRUE -DCUDA=FALSE -DINSTALL_TESTS=TRUE \
    -DDOWNLOAD_MINICONDA=TRUE -DMINICONDA_USE_PY3=TRUE \
    -DCOLOR_CMAKE_MESSAGES=FALSE \
    2>&1 | tee cmake.log

OSC_CC=icc
OSC_COMPILER=intel
OSC_COMPILER_MAJOR=19
OSC_COMPILER_MINOR=0
OSC_CXX=icpc
OSC_F77=ifort
OSC_F90=ifort

scott

On Wed, Apr 29, 2020 at 04:11:04PM -0400, Daniel Roe wrote:
> For the parallel build, I can't get RC5 to get past the initial
> run_cmake phase with Intel 17.0.4 + Intel MPI + cmake 3.13.2:
>
> ```
> -- Could NOT find MPI_C (missing: MPI_mpigi_LIBRARY MPI_C_WORKS)
> -- Could NOT find MPI_CXX (missing: MPI_mpigi_LIBRARY MPI_CXX_WORKS)
> -- Could NOT find MPI_Fortran (missing: MPI_mpigi_LIBRARY MPI_Fortran_WORKS)
> -- Could NOT find MPI (missing: MPI_C_FOUND MPI_CXX_FOUND MPI_Fortran_FOUND)
> CMake Error at cmake/MPIConfig.cmake:12 (message):
> You requested MPI, but the MPI C library was not found. Please install one
> and try again, or set MPI_C_INCLUDE_PATH and MPI_C_LIBRARIES to point to
> your MPI.
> Call Stack (most recent call first):
> CMakeLists.txt:118 (include)
> ```
>
> Run after setting -DCOMILER to INTEL, and -DMPI and -DCUDA to TRUE.
> CMakeOutput.log is quite monolithic and it's not immediately clear to
> me what's going wrong (I can send it to interested parties if needed).
> The serial cmake (-DMPI=FALSE -DCUDA=FALSE) + install works. Note that
> the parallel cmake fails even after the serial cmake + install
> succeeds. Legacy configure works normally.
>

On Wed, Apr 29, 2020 at 06:54:02PM -0400, Brent Krueger wrote:
> Gerald -- CentOS 8 with MPICH does seem to be working for me. Build was
> find and tests are ongoing -- but most/all seem to be passing so far.
>
> However, I can NOT build CUDA (either 10.1 or 10.2) with CentOS 8
> successfully anymore. Both 10.1 and 10.2 worked fine with CentOS 8 for me
> with RC4 and RC3.
>
> Brent
>
>
> On Wed, Apr 29, 2020 at 4:57 PM David A Case <david.case.rutgers.edu> wrote:
>
> > On Wed, Apr 29, 2020, Dan Roe wrote:
> >
> > >For the parallel build, I can't get RC5 to get past the initial
> > >run_cmake phase with Intel 17.0.4 + Intel MPI + cmake 3.13.2:
> >
> > Is this a regression? That is, did earlier release candidates work? If
> > so, be sure to let Jamie know (see below), as this can probably be fixed.
> >
> > I don't have Intel MPI, (it's quite expensive) so can't test what you
> > are trying. I've built my own MPI using mpich and the intel compilers
> > (i.e. configure_mpich intel) -- will try that out soon.
> >
> > I can confirm that serial intel works pretty well (see wiki test results
> > site). I'm testing version 19.0.5.
> >
> >
> > >
> > >```
> > >-- Could NOT find MPI_C (missing: MPI_mpigi_LIBRARY MPI_C_WORKS)
> > >-- Could NOT find MPI_CXX (missing: MPI_mpigi_LIBRARY MPI_CXX_WORKS)
> > >-- Could NOT find MPI_Fortran (missing: MPI_mpigi_LIBRARY
> > MPI_Fortran_WORKS)
> > >-- Could NOT find MPI (missing: MPI_C_FOUND MPI_CXX_FOUND
> > MPI_Fortran_FOUND)
> > >CMake Error at cmake/MPIConfig.cmake:12 (message):
> > > You requested MPI, but the MPI C library was not found. Please install
> > one
> > > and try again, or set MPI_C_INCLUDE_PATH and MPI_C_LIBRARIES to point to
> > > your MPI.
> >
> > Be sure that you get problem reports to Jamie
> > <jsmith.crackofdawn.onmicrosoft.com>, since I doubt that he follows the
> > amber-developers mailing list.
> >
> > I can't find any Google hits on MPI_mpigi_LIBRARY. You might try
> > setting your CMAKE_PREFIX_PATH variable to the location of the intel MPI
> > library.
> >

_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Wed Apr 29 2020 - 17:30:02 PDT
Custom Search