make install-missing file?

[Jack Wieting]

I am having a little issue trying to build openmm from source. I got so far as to run make test and all the tests passed, but then when I do make install errors pop up when it tries to link the shared library libOpenMM.dylib. I think some file is missing but I am not sure, all the errors are cuda related. Ive listed some below, any help would be greatly appreciated

Undefined symbols:
"_cudaEventSynchronize", referenced from:
"GpuLJ14Softcore::GpuLJ14Softcore()", referenced from:
"_cudaMemcpyFromSymbol", referenced from:
"_cudaLaunch", referenced from:
"GpuGBVISoftcore::setQuinticUpperLimit(float)", referenced from:
_gpuSetGBVISoftcoreParameters in kCalculateGBVISoftcoreBornSum.cu_OpenMMFreeEnergyCuda_generated.cpp.o
"GpuGBVISoftcore::setQuinticLowerLimitFactor(float)", referenced from:
_gpuSetGBVISoftcoreParameters in kCalculateGBVISoftcoreBornSum.cu_OpenMMFreeEnergyCuda_generated.cpp.o
"GpuNonbondedSoftcore::GpuNonbondedSoftcore()", referenced from:
_gpuSetNonbondedSoftcoreParameters in kCalculateNonbondedSoftcore.cu_OpenMMFreeEnergyCuda_generated.cpp.o
"GpuNonbondedSoftcore::~GpuNonbondedSoftcore()", referenced from:
_gpuDeleteNonbondedSoftcoreParameters in kCalculateNonbondedSoftcore.cu_OpenMMFreeEnergyCuda_generated.cpp.o
"_cudaGetLastError", referenced from:

[Peter Eastman]

Those errors are all coming from the free energy plugin. Just disable building it by setting OPENMM_BUILD_FREE_ENERGY_PLUGIN and OPENMM_BUILD_FREE_ENERGY_CUDA_LIB to false.

Peter

[Jack Wieting]

Thanks for the response Peter, but the errors are still there. I have included some more of the error below. Any more ideas would be greatly appreciated.

Linking CXX shared library libOpenMM.dylib
Undefined symbols:
"_cudaEventSynchronize", referenced from:
kApplyCCMA(_gpuContext*, float4*, bool)in kCCMA.cu_OpenMMCuda_generated.cpp.o
"_cudaMemcpyFromSymbol", referenced from:
GetBrownianUpdateSim(_gpuContext*) in kBrownianUpdate.cu_OpenMMCuda_generated.cpp.o
GetCalculateAndersenThermostatSim(_gpuContext*) in kCalculateAndersenThermostat.cu_OpenMMCuda_generated.cpp.o
GetCalculateCDLJForcesSim(_gpuContext*) in kCalculateCDLJForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateCDLJObcGbsaForces1Sim(_gpuContext*) in kCalculateCDLJObcGbsaForces1.cu_OpenMMCuda_generated.cpp.o
GetCalculateCustomAngleForcesSim(_gpuContext*) in kCalculateCustomAngleForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateCustomBondForcesSim(_gpuContext*) in kCalculateCustomBondForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateCustomExternalForcesSim(_gpuContext*) in kCalculateCustomExternalForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateCustomNonbondedForcesSim(_gpuContext*) in kCalculateCustomNonbondedForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateCustomTorsionForcesSim(_gpuContext*) in kCalculateCustomTorsionForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateGBVIBornSumSim(_gpuContext*) in kCalculateGBVIBornSum.cu_OpenMMCuda_generated.cpp.o
GetCalculateGBVIForces2Sim(_gpuContext*) in kCalculateGBVIForces2.cu_OpenMMCuda_generated.cpp.o
GetCalculateLocalForcesSim(_gpuContext*) in kCalculateLocalForces.cu_OpenMMCuda_generated.cpp.o
GetCalculateObcGbsaBornSumSim(_gpuContext*) in kCalculateObcGbsaBornSum.cu_OpenMMCuda_generated.cpp.o
GetCalculateObcGbsaForces2Sim(_gpuContext*) in kCalculateObcGbsaForces2.cu_OpenMMCuda_generated.cpp.o
GetCalculatePMESim(_gpuContext*) in kCalculatePME.cu_OpenMMCuda_generated.cpp.o
GetCCMASim(_gpuContext*) in kCCMA.cu_OpenMMCuda_generated.cpp.o
GetForcesSim(_gpuContext*) in kForces.cu_OpenMMCuda_generated.cpp.o
GetLangevinUpdateSim(_gpuContext*) in kLangevinUpdate.cu_OpenMMCuda_generated.cpp.o
GetRandomSim(_gpuContext*) in kRandom.cu_OpenMMCuda_generated.cpp.o
GetSettleSim(_gpuContext*) in kSettle.cu_OpenMMCuda_generated.cpp.o
GetShakeHSim(_gpuContext*) in kShakeH.cu_OpenMMCuda_generated.cpp.o
GetVerletUpdateSim(_gpuContext*) in kVerletUpdate.cu_OpenMMCuda_generated.cpp.o
"_cudaLaunch", referenced from:
__device_stub__Z12bitonicSortDI4int2EvPT_(int2*) in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z22assignElementToBucketDI4int2EvPT_S2_iPjff(int2*, int2*, int, unsigned int*, float, float)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z22assignElementToSlicesDI4int2EvPT_iPjS3_ffi(int2*, int, unsigned int*, unsigned int*, float, float, int)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z34assignElementToSlicesNearlySortedDI4int2EvPT_iPjS3_ffii(int2*, int, unsigned int*, unsigned int*, float, float, int, int)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z14calDifferenceDI4int2EvPT_Pfi(int2*, float*, int)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z10reduceMinDI4int2EvPT_ii(int2*, int, int)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z10reduceMaxDI4int2EvPT_ii(int2*, int, int)in bbsort.cu_OpenMMCuda_generated.cpp.o
__device_stub__Z10reduceSumDPfii(float*, int, int)in bbsort.cu_OpenMMCuda_generated.cpp.o

[Jack Wieting]

Here is some more information. To compile cuda programs I need to type:

nvcc --include-path /usr/local/cuda/include -L/usr/local/cuda/lib -lcuda -lcudart -lm -lpthread

before the source fils. If I type:

nvcc --include-path /usr/local/cuda/include

I get the same errors that I do now compiling openMM. Any ideas on how I can express this in the make file?

Thanks,
Jack