Error with build eclm-parflowGPU with intel.psmpi, LAPACK routines?

Follow-up on issue #100

When building the combination eclm-parflowGPU with jsc.2025.intel.psmpi I get the following error:

[ 99%] Building Fortran object eclm/CMakeFiles/eclm.dir/cime_driver.F90.o
[100%] Linking Fortran executable eclm.exe
ld: ../clm5/libclm.a(SoilWaterMovementMod.F90.o): in function `dgtsv_.t253890p.t253891p.t253892p.t253893p.t253894p.t253895p.t253896p.t253897p':
/p/scratch/cslts/gonzalez5/TSMP2/BUILDS/TSMP2_autom_builds/models/eCLM/src/clm5/biogeophys/SoilWaterMovementMod.F90:(.text+0x280855): undefined reference to `dgtsv_'
ld: ../clm5/libclm.a(UrbBuildTempOleson2015Mod.F90.o): in function `dgesv_.t58174p.t58175p.t58176p.t58177p.t58178p.t58179p.t58180p.t58181p':
/p/scratch/cslts/gonzalez5/TSMP2/BUILDS/TSMP2_autom_builds/models/eCLM/src/clm5/biogeophys/UrbBuildTempOleson2015Mod.F90:(.text+0x237ae5): undefined reference to `dgesv_'
ld: ../clm5/libclm.a(BandDiagonalMod.F90.o): in function `dgbsv_.t839p.t840p.t841p.t842p.t843p.t844p.t845p.t846p.t847p.t848p':
/p/scratch/cslts/gonzalez5/TSMP2/BUILDS/TSMP2_autom_builds/models/eCLM/src/clm5/biogeophys/BandDiagonalMod.F90:(.text+0x28eaf): undefined reference to `dgbsv_'
gmake[5]: *** [eclm/CMakeFiles/eclm.dir/build.make:500: eclm/eclm.exe] Error 1
gmake[4]: *** [CMakeFiles/Makefile2:582: eclm/CMakeFiles/eclm.dir/all] Error 2
gmake[3]: *** [Makefile:136: all] Error 2
gmake[2]: *** [CMakeFiles/eCLM.dir/build.make:112: eCLM/src/eCLM-stamp/eCLM-install] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:113: CMakeFiles/eCLM.dir/all] Error 2
gmake: *** [Makefile:91: all] Error 2

Do these undefined reference to 'dgesv_', 'dgbsv_','dgtsv_' errors mean that the LAPACK routines are not being linked into the final executable when using intel.psmpi environment?

NOTE: the eclm-parflowGPU builds for gnu.psmpi and gnu.openmpi environments work fine.

Thanks for reporting this Ana. The workaround is to simply module load ScaLAPACK/2.2.0-fb in env/jsc.2025.intel.psmpi. While this fixes the eCLM build, there will be issues in the ParFlowGPU build. It would be great if you can also note the other build errors so we could address them.

@kvrigor

@kvrigor Here the next error:

[100%] Linking Fortran executable eclm.exe
/p/software/fs/jurecadc/stages/2025/software/intel-compilers/2024.2.0-CUDA-12/compiler/2024.2/lib/clang/19/lib/x86_64-unknown-linux-gnu/libclang_rt.msan.a(msan.cpp.o): in function `__msan::flags()':
msan.cpp:(.text._ZN6__msan5flagsEv+0x7): relocation truncated to fit: R_X86_64_PC32 against `.bss._ZN6__msanL10msan_flagsE'
/p/software/fs/jurecadc/stages/2025/software/intel-compilers/2024.2.0-CUDA-12/compiler/2024.2/lib/clang/19/lib/x86_64-unknown-linux-gnu/libclang_rt.msan.a(msan.cpp.o): in function `__msan::PrintWarningWithOrigin(unsigned long, unsigned long, unsigned int)':
msan.cpp:(.text._ZN6__msan22PrintWarningWithOriginEmmj+0x37): relocation truncated to fit: R_X86_64_PC32 against symbol `__msan::msan_report_count' defined in .bss._ZN6__msan17msan_report_countE section in /p/software/fs/jurecadc/stages/2025/software/intel-compilers/2024.2.0-CUDA-12/compiler/2024.2/lib/clang/19/lib/x86_64-unknown-linux-gnu/libclang_rt.msan.a(msan.cpp.o)
msan.cpp:(.text._ZN6__msan22PrintWarningWithOriginEmmj+0x5b): relocation truncated to fit: R_X86_64_PC32 against symbol `__msan::msan_inited' defined in .bss._ZN6__msan11msan_initedE section in /p/software/fs/jurecadc/stages/2025/software/intel-compilers/2024.2.0-CUDA-12/compiler/2024.2/lib/clang/19/lib/x86_64-unknown-linux-gnu/libclang_rt.msan.a(msan.cpp.o)
msan.cpp:(.text._ZN6__msan22PrintWarningWithOriginEmmj+0x6b): failed to convert GOTPCREL relocation against '_ZN11__sanitizer21common_flags_dont_useE'; relink with --no-relax
gmake[5]: *** [eclm/CMakeFiles/eclm.dir/build.make:504: eclm/eclm.exe] Error 1
gmake[4]: *** [CMakeFiles/Makefile2:582: eclm/CMakeFiles/eclm.dir/all] Error 2
gmake[3]: *** [Makefile:136: all] Error 2
gmake[2]: *** [CMakeFiles/eCLM.dir/build.make:112: eCLM/src/eCLM-stamp/eCLM-install] Error 2
gmake[1]: *** [CMakeFiles/Makefile2:113: CMakeFiles/eCLM.dir/all] Error 2
gmake: *** [Makefile:91: all] Error 2

@AGonzalezNicolas

@AGonzalezNicolas this works for me: ./build_tsmp2.sh eclm parflow --env env/jsc.2025.intel.psmpi. You can test the binaries here: /p/scratch/cslts/shared_data/TSMP2_S2025/JURECADC_eCLM-ParFlow_RELEASE-Intel

Note that we can't build eCLM-ParflowGPU using Intel compilers since there is no equivalent Hypre library built on the same toolchain.

TSMP2/env/jsc.2025.intel.psmpi

Lines 38 to 45 in 2d5bfeb

    
           if [[ "$1" == "--parflowgpu" ]]; then 
        
             # WARNING: 
        
             # 
        
             # Loading Hypre/2.31.0 under Intel will stubbornly load Hypre/2.31.0-cpu 
        
             # since Hypre/2.31.0 is only available under the GCC toolchain. This wrong 
        
             # Hypre will trigger a linker failure in ParFlow. I suggest using jsc.2025.gnu.openmpi when 
        
             # building ParFlow.

Ok. Since no eclm-parflowGPU can't be built with jsc.2025.intel.psmpi, this issue can be closed.

Navigation Menu

Repository navigation

Error with build eclm-parflowGPU with `intel.psmpi`, LAPACK routines? #113

Follow-up on issue #100

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	if [[ "$1" == "--parflowgpu" ]]; then
	# WARNING:
	#
	# Loading Hypre/2.31.0 under Intel will stubbornly load Hypre/2.31.0-cpu
	# since Hypre/2.31.0 is only available under the GCC toolchain. This wrong
	# Hypre will trigger a linker failure in ParFlow. I suggest using jsc.2025.gnu.openmpi when
	# building ParFlow.

Error with build eclm-parflowGPU with intel.psmpi, LAPACK routines? #113

Description

Follow-up on issue #100

Activity

kvrigor commented on Nov 28, 2025

AGonzalezNicolas commented on Nov 28, 2025

kvrigor commented on Jun 17, 2026

AGonzalezNicolas commented on Jun 18, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Error with build eclm-parflowGPU with `intel.psmpi`, LAPACK routines? #113