lapack > 3.8.0 fails to build with IBM XL compiler #606
Comments
|
This looks to me to be failing in generic CMAKE test code for verifying that the compiler can build anything at all, way before anything LAPACK-specific. Were you able to compile 3.8.0 on the exact same system ? |
Hi Martin, Yes, 3.8.0 will work with a small patch. To see that, do: $ wget https://github.com/Reference-LAPACK/lapack/archive/refs/tags/v3.8.0.tar.gz
$ tar xvzf https://github.com/Reference-LAPACK/lapack/archive/refs/tags/v3.8.0.tar.gz
$ cd lapack-3.8.0
$ patch -p1 < /path/to/ibm-xl.patch
$ cd ..Then do a CMake build as above: $ mkdir build; cd build
$ cmake -G 'Unix Makefiles' -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo -DCMAKE_INTERPROCEDURAL_OPTIMIZATION:BOOL=OFF -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON -DCMAKE_INSTALL_RPATH_USE_LINK_PATH:BOOL=OFF -DBUILD_SHARED_LIBS:BOOL=OFF -DLAPACKE:BOOL=ON -DLAPACKE_WITH_TMG:BOOL=ON -DCBLAS=ON -DCMAKE_Fortran_COMPILER=xlf_r -DCMAKE_Fortran_FLAGS='-F/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/etc/xlf.cfg.rhel.7.8.gcc.8.3.0.cuda.10.1 -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -O3 -qnohot' -DBUILD_DEPRECATED:BOOL=ON -DBUILD_TESTING:BOOL=OFF /home/quellyn/GIT/lapack/3.8.0/lapack-3.8.0I've attached the output of that so you can see it was successful. This patch originated from a PR that @KineticTheory put into lapack some time ago. It was adopted into Spack and is used there successfully for building lapack 3.8.0 as shown above. However, it does not apply to 3.9.0. Based on @KineticTheory 's PR, it appears that all or some of it was accepted upstream. Even so, a build of lapack 3.9.0 fails: $ wget https://github.com/Reference-LAPACK/lapack/archive/refs/tags/v3.9.0.tar.gz
$ tar xvzf v3.9.0.tar.gz
$ mkdir build; cd build
$ cmake -G 'Unix Makefiles' -DCMAKE_BUILD_TYPE:STRING=RelWithDebInfo -DCMAKE_INTERPROCEDURAL_OPTIMIZATION:BOOL=OFF -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON -DCMAKE_INSTALL_RPATH_USE_LINK_PATH:BOOL=OFF -DBUILD_SHARED_LIBS:BOOL=OFF -DLAPACKE:BOOL=ON -DLAPACKE_WITH_TMG:BOOL=ON -DCBLAS=ON -DCMAKE_Fortran_COMPILER=xlf_r -DCMAKE_Fortran_FLAGS='-F/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/etc/xlf.cfg.rhel.7.8.gcc.8.3.0.cuda.10.1 -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -O3 -qnohot' -DBUILD_DEPRECATED:BOOL=ON -DBUILD_TESTING:BOOL=OFF /home/quellyn/GIT/lapack/3.9.0/lapack-3.9.0The errors are the same as I see in 3.10.0: -- The Fortran compiler identification is XL 16.1.1
-- The C compiler identification is XLClang 16.1.1.7
-- Check for working Fortran compiler: /projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r
-- Check for working Fortran compiler: /projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r - works
-- Detecting Fortran compiler ABI info
-- Detecting Fortran compiler ABI info - done
-- Checking whether /projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r supports Fortran 90
-- Checking whether /projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r supports Fortran 90 - yes
-- Check for working C compiler: /projects/opt/ppc64le/ibm/xlc-16.1.1.7/xlC/16.1.1/bin/xlc
-- Check for working C compiler: /projects/opt/ppc64le/ibm/xlc-16.1.1.7/xlC/16.1.1/bin/xlc - works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Checking if build type is 'Coverage'
-- Checking if build type is 'Coverage': 0
-- Build tests: OFF
-- Enabling fixed format F90/F95 with -qfixed
-- Looking for Fortran NONE - found
-- Looking for Fortran INT_CPU_TIME - found
-- Looking for Fortran EXT_ETIME - not found
-- Looking for Fortran EXT_ETIME_ - found
-- Looking for Fortran INT_ETIME - not found
-- --> Will use second_EXT_ETIME_.f and dsecnd_EXT_ETIME_.f as timing function.
-- Build deprecated routines: ON
-- Build single precision real: ON
-- Build double precision real: ON
-- Build single precision complex: ON
-- Build double precision complex: ON
-- Using supplied NETLIB BLAS implementation
-- CBLAS enable
-- Detecting Fortran/C Interface
Failed to compile
-- Verifying Fortran/C Compiler Compatibility
Failed to compile
CMake Warning (dev) at /projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface.cmake:309 (message):
No FortranCInterface mangling known for VerifyFortran
Call Stack (most recent call first):
/projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify/CMakeLists.txt:16 (FortranCInterface_HEADER)
This warning is for project developers. Use -Wno-dev to suppress it.
-- Verifying Fortran/C Compiler Compatibility - Failed
CMake Error at /projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface.cmake:383 (message):
The Fortran compiler:
/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r
and the C compiler:
/projects/opt/ppc64le/ibm/xlc-16.1.1.7/xlC/16.1.1/bin/xlc
failed to compile a simple test project using both languages. The output
was:
Change Dir: /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC
Run Build Command(s):/usr/bin/gmake VerifyFortranC && /projects/opt/ppc64le/p9/cmake/3.17.3/bin/cmake -S/projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify -B/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC --check-build-system CMakeFiles/Makefile.cmake 0
/usr/bin/gmake -f CMakeFiles/Makefile2 VerifyFortranC
gmake[1]: Entering directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
/projects/opt/ppc64le/p9/cmake/3.17.3/bin/cmake -S/projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify -B/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC --check-build-system CMakeFiles/Makefile.cmake 0
/projects/opt/ppc64le/p9/cmake/3.17.3/bin/cmake -E cmake_progress_start /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC/CMakeFiles 5
/usr/bin/gmake -f CMakeFiles/Makefile2 CMakeFiles/VerifyFortranC.dir/all
gmake[2]: Entering directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
/usr/bin/gmake -f CMakeFiles/VerifyFortran.dir/build.make CMakeFiles/VerifyFortran.dir/depend
gmake[3]: Entering directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
cd /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC && /projects/opt/ppc64le/p9/cmake/3.17.3/bin/cmake -E cmake_depends "Unix Makefiles" /projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify /projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC /home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC/CMakeFiles/VerifyFortran.dir/DependInfo.cmake
Scanning dependencies of target VerifyFortran
gmake[3]: Leaving directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
/usr/bin/gmake -f CMakeFiles/VerifyFortran.dir/build.make CMakeFiles/VerifyFortran.dir/build
gmake[3]: Entering directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
[ 20%] Building Fortran object CMakeFiles/VerifyFortran.dir/VerifyFortran.f.o
/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/xlf_r -I/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC -F/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/etc/xlf.cfg.rhel.7.8.gcc.8.3.0.cuda.10.1 -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/lib -Wl,-rpath,/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -L/projects/opt/ppc64le/ibm/xlf-16.1.1.7/lib -O3 -qnohot;-qnosave;-qstrict=none -qfixed -O -c /projects/opt/ppc64le/p9/cmake/3.17.3/share/cmake-3.17/Modules/FortranCInterface/Verify/VerifyFortran.f -o CMakeFiles/VerifyFortran.dir/VerifyFortran.f.o
/projects/opt/ppc64le/ibm/xlf-16.1.1.7/xlf/16.1.1/bin/.orig/xlf_r: 1501-294 (S) No input file specified. Please use -qhelp for more information.
/bin/sh: -qnosave: command not found
/bin/sh: -qstrict=none: command not found
gmake[3]: *** [CMakeFiles/VerifyFortran.dir/VerifyFortran.f.o] Error 127
gmake[3]: Leaving directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
gmake[2]: *** [CMakeFiles/VerifyFortran.dir/all] Error 2
gmake[2]: Leaving directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
gmake[1]: *** [CMakeFiles/VerifyFortranC.dir/rule] Error 2
gmake[1]: Leaving directory `/vast/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/FortranCInterface/VerifyC'
gmake: *** [VerifyFortranC] Error 2
Call Stack (most recent call first):
CBLAS/CMakeLists.txt:9 (FortranCInterface_VERIFY)
-- Configuring incomplete, errors occurred!
See also "/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/CMakeOutput.log".
See also "/home/quellyn/GIT/lapack/3.9.0/build/CMakeFiles/CMakeError.log".Thanks for looking at this, |
|
Hmm. The effect is the same, but the source of the error appears to be different - in the 3.9.0 log, the compiler aborts because it does not understand the |
Usually I perform all my builds within Spack, but I did these build attempts by hand in order to make the simplest possible reproducer. It's possible that I did not have something set correctly in that case. Let me repeat all the various build attempts (3.8.0, 3.9.0, and 3.10.0) and make sure that I am running all of them in the same way. |
|
Okay, I have repeated all the builds, one after another, on the same Power9 node with the same modules (for cmake and IBM XL) loaded. Here are the complete build directories with the CMake output logs, etc.: 3.8.0.tar.gz I apologize for uploading such large files but as you don't have ready access to a similar Power9 platform I wanted to give you as much as I could. |
Yes, I can see that now. I've created a new patch to address the |
|
Actually in 3.9.0, "something" appears to have inserted semicolon characters into the xlf command line between the qnohot/qnosave/qstrict options, turning it into separate, nonsensical commands. (So first xlf complains about not seeing a file to compile, then the shell complains about an unrecognized command "qnohot" etc.) I think this is related to how the options string is concatenated in cmake (i.e. I made such a mistake in a different project in the past) but I need to find where it happens. |
Thank you again for looking. Please let me know if there's anything I can try out on my Power9 platform here. I'm not much of a coder but I'm willing to try with some guidance :-) |
|
Any progress on this issue? Anything you would like me to try out on my end? |
|
Sorry, no idea why the 3.10.0 build would not find libgcc_s , and no time to look further. (The library seems to have been found in the other builds, the test that needs it is part of cmake and I saw no obvious difference in compiler arguments.) |
|
Hi Martin,
I've found the problem: Beginning with You can see the incorrect compiler behavior with a simple "Hello World" program: The
Indeed, that was the issue. In |
|
Ah, great. Yes please open a PR if you have the time. And sorry again for dropping out of this discussion - I do not have access to ppc systems with the xlf compiler, and I was busy with other tasks. |
|
(And looking at your PR for spack, perhaps the current set of tests for a supported "recursive" keyword could be retained as a catch-all in case the Fortran compiler is something else not on the list of known names ?) |
Additional build variants of TAU with combinatorial support CUDA and OMP profiling enabled-or-disabled are added to Summit. Module suffixes are used to distinguish the different build variants. Darshan-runtime without HDF5 support is added and made the default build. Modules for this variant are suffixed `-lite` and re-added to the default loaded DefApps modules. Adios 1.13.1 is installed for a special case request by a user. Imagemagick is installed at user request. An attempt to install the latest netlib-lapack using the XL compilers is added. These builds still fail due to a configure-time bug validating xlf when building the static versions of the libraries. XLF attempts to link GCC runtime libs into the build without providing the necessary library directories or without static libraries available. The failure output from a build attempt which forces the `-qstaticlink` flag is shown below, but `-qstaticlink` is not set for a normal build and it's presence has no effect on the error. A similar error is already reported upstream Reference-LAPACK/lapack#606: ``` Build flags: ;-O3;-qnohot;-qstaticlink Id flags: -v The output was: 1 exec: export(export,XL_CONFIG=/sw/summit/xl/16.1.1-10/xlf/16.1.1/etc/xlf.cfg.rhel.8.2.gcc.8.3.1.cuda.11.0:xlf_r,NULL) exec: export(export,XL_DIS=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/dis -o "CMakeFortranCompilerId.s" "CMakeFortranCompilerId.o",NULL) exec: /autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/cpp(/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/cpp,CMakeFortranCompilerId.F,/tmp/xlf.jWEE22/0hkYDH,-D__ibmxl__,-C,-q64,-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlmass/9.1.1/include,-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlsmp/5.1.1/include,-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/include,NULL) exec: /autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfentry(/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfentry,/tmp/xlf.jWEE22/0hkYDH,/tmp/xlf.jWEE22/vMULQf,/tmp/xlf.jWEE22/vMULQfF.lst,xlfsmsg.cat,xlfmsg.cat,CMakeFortranCompilerId.F,OSVAR(rhel.8.2),THREADED,NOZEROSIZE,ALIAS(intptr),POSITION(appendold),XLF90(noautodealloc,nosignedzero,oldpad),XLF77(intarg,intxor,noleadzero,gedit77,noblankpad,oldboz,softeof),XLF2003(nopolymorphic,nobozlitargs,nostopexcept,novolatile,noautorealloc,oldnaninf),XLF2008(nocheckpresence),64,GNU_VERSION(8.3.1),SAVEOPTFILE(/tmp/xlf.jWEE22/optf.lUZ5eB),SAVEVERSIONFILE(/tmp/xlf.jWEE22/version.Ck8wr9),OPT(3),NOHOT,WSTREAMS(/tmp/xlf.jWEE22/h1,/tmp/xlf.jWEE22/b1,/tmp/xlf.jWEE22/s1),DEFMSG(/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/msg/en_US),XFLAG(CPP),-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlmass/9.1.1/include,-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlsmp/5.1.1/include,-I/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/include,NULL) ** cmakefortrancompilerid === End of Compilation 1 === unlink: /tmp/xlf.jWEE22/0hkYDH exec: export(export,XL_FRONTEND=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfentry,NULL) exec: export(export,XL_ASTI=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfhot,NULL) exec: export(export,XL_BACKEND=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfcode,NULL) exec: export(export,XL_LINKER=/usr/bin/ld,NULL) exec: export(export,XL_BOLT=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/bolt,NULL) exec: /autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfhot(/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfhot,/tmp/xlf.jWEE22/h1,/tmp/xlf.jWEE22/h2,/tmp/xlf.jWEE22/b1,/tmp/xlf.jWEE22/b2,/tmp/xlf.jWEE22/s1,/tmp/xlf.jWEE22/s2,/tmp/xlf.jWEE22/vMULQf,/tmp/xlf.jWEE22/vMULQfA.lst,NULL) exec: export(export,XL_FRONTEND=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfentry,NULL) exec: export(export,XL_ASTI=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfhot,NULL) exec: export(export,XL_BACKEND=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfcode,NULL) exec: export(export,XL_LINKER=/usr/bin/ld,NULL) exec: export(export,XL_BOLT=/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/bolt,NULL) exec: /autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfcode(/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/exe/xlfcode,/tmp/xlf.jWEE22/h2,/tmp/xlf.jWEE22/b2,CMakeFortranCompilerId.o,/tmp/xlf.jWEE22/vMULQfB.lst,/tmp/xlf.jWEE22/s2,NULL) 1501-510 Compilation successful for file CMakeFortranCompilerId.F. unlink: /tmp/xlf.jWEE22/vMULQfA.lst exec: /usr/bin/ld(ld,--eh-frame-hdr,-Qy,-melf64lppc,-static,/usr/lib/gcc/ppc64le-redhat-linux/8/../../../../lib64/crt1.o,/usr/lib/gcc/ppc64le-redhat-linux/8/../../../../lib64/crti.o,/usr/lib/gcc/ppc64le-redhat-linux/8/crtbeginT.o,-L/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlsmp/5.1.1/lib,-L/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlmass/9.1.1/lib,-L/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/xlf/16.1.1/lib,-R/autofs/nccs-svm1_sw/summit/xl/16.1.1-10/lib,-L/usr/lib/gcc/ppc64le-redhat-linux/8,-L/usr/lib/gcc/ppc64le-redhat-linux/8/../../../../lib64,-L/lib/../lib64,-L/usr/lib/../lib64,-L/sw/summit/spack-envs/base/opt/linux-rhel8-ppc64le/xl-16.1.1-10/spectrum-mpi-10.4.0.3-20210112-v7qymniwgi6mtxqsjd7p5jxinxzdkhn3/lib,-L/ccs/proj/ccsstaff/olcf4_acceptance/devel/nccs_test_harness/libraries,-L/usr/lib/gcc/ppc64le-redhat-linux/8/../../..,--no-toc-optimize,--no-toc-optimize,--no-toc-optimize,--no-toc-optimize,--no-toc-optimize,--no-toc-optimize,CMakeFortranCompilerId.o,-lxlf90_r,-lxlopt,-lxlomp_ser,-lxl,-lxlfmath,-ldl,-lrt,-lpthread,-lm,-lc,-lgcc_eh,-lc,-lgcc,/usr/lib/gcc/ppc64le-redhat-linux/8/crtend.o,/usr/lib/gcc/ppc64le-redhat-linux/8/../../../../lib64/crtn.o,NULL) | exec: /usr/bin/awk(/usr/bin/awk,BEGIN{a=0;b=0;s=""};{if(($0!~/.*-224.*@tp.*/)&&($0!~/.*-34[45].*loadmap.*/||b==1||a==0)){b=1;print $0;}else{a=1;if((s=="")&& $0 ~ /.*-34[45].*loadmap.*/){s=$0}}};END{if(b==1 && s!=""){print s}},NULL) ld: cannot find -ldl ld: cannot find -lrt ld: cannot find -lpthread ld: cannot find -lm ld: cannot find -lc ld: cannot find -lc unlink: /tmp/xlf.jWEE22/optf.lUZ5eB unlink: /tmp/xlf.jWEE22/h1 unlink: /tmp/xlf.jWEE22/h2 unlink: /tmp/xlf.jWEE22/b1 unlink: /tmp/xlf.jWEE22/b2 unlink: /tmp/xlf.jWEE22/s1 unlink: /tmp/xlf.jWEE22/s2 unlink: /tmp/xlf.jWEE22/version.Ck8wr9 unlink: /tmp/xlf.jWEE22 ```
Be more explicit about selecting recursive fortran flags, as not every compiler will correctly reject an incorrect option (e.g., XL).
|
I think this issue can be closed. My troubles have been resolved by the above-mentioned PR to lapack. |
Beginning with lapack version 3.9.0, I am unable to build using the IBM XL compiler on a Power9 Linux platform.
To reproduce:
The above fails with:
I've included the CMake output and error logs:
CMakeError.log
CMakeOutput.log
I'm including @KineticTheory on this report, as he has an interest in this issue as well.
Thanks for looking!
Quellyn Snead
Checklist
The text was updated successfully, but these errors were encountered: