atillack
commented
2 years ago @randymojica Thank you for your very detailed issue post. Both nvcc: not found and Error: clGetPlatformIDs(): -1001 in the output indicate that something is still not quite working yet with the Cuda and/or Nvidia driver installation. What is the output of nvidia-smi?
Also, it looks like your second .bashrc should use export CUDAROOT=/usr/local/cuda-11.7 and with a bit of luck all that's needed additionally is export LD_LIBRARY_PATH="$CUDAROOT/lib64:$LD_LIBRARY_PATH" ;-)
randymojica
Also, after adding 
tobigithub
Good Afternoon! I have been working with AutoDock4 since a few years ago and now that there is a recent and updated version I really wanted to give it a try. I tried to compile AutoDock-GPU from source and failed, and downloaded de binary adgpu-v1.5.3_linux_ocl_128wi released on Dec. 17, 2021, but it didn't work either.
To Reproduce When I tried to execute the binary downloaded from github, there is an error constantly popping up:
./adgpu-v1.5.3_linux_ocl_128wi --lfile ../acarbose_pubchem_minimized.pdbqt --ffile ../aGluc_3top_md_first_frame.maps.fld --nrun 50Running 1 docking calculation Kernel source used for development: ./device/calcenergy.cl Kernel string used for building: ./host/inc/stringify.h Kernel compilation flags: -I ./device -I ./common -DN128WI -cl-mad-enable Error: clGetPlatformIDs(): -1001
Expected behavior I expected for the compiled program to excecute like normal or the downloaded binary to run without problems as in the previous AutoDock 4.2.6, but that was not the case.
Information to help narrow down the bug
Version of AutoDock-GPU: v1.5.3 and v.1.5.release
Which operating system are you on?: WSL2 (tried on both Ubuntu 22.04 LTS and Ubuntu 20.04 LTS) running on Windows 11 Home version 22H2 build 22623.746
Which compiler, compiler version, and
makecompile options did you use? To compile, I tried multiple combination of options in both operative systems, e.g.:make DEVICE=CUDA NUMWI=256make DEVICE=CUDA NUMWI=128make DEVICE=CUDA NUMWI=64Resulting in:rm -f ./host/inc/performdocking.h ./host/src/performdocking.cpp Building adgpu_analysis ... g++ \ ./host/src/calcenergy.cpp ./host/src/getparameters.cpp ./host/src/main.cpp ./host/src/miscellaneous.cpp ./host/src/processgrid.cpp ./host/src/processligand.cpp ./host/src/processresult.cpp ./host/src/setup.cpp \ -std=c++11 -I./common -I./host/inc -DVERSION=\"v1.5-release\" \ -o./bin/adgpu_analysis \ -O3 -DTOOLMODE DEVICE is set to GPU CPU_INCLUDE_PATH is set to /usr/local/cuda-11.7/include/CL CPU_LIBRARY_PATH is set to /usr/local/cuda-11.7/lib64 GPU_INCLUDE_PATH is set to /usr/local/cuda-11.7/include GPU_LIBRARY_PATH is set to /usr/local/cuda-11.7/lib64 nvcc -DN256WI -use_fast_math --ptxas-options="-v" -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61 -gencode arch=compute_70,code=sm_70 -std=c++11 -I./common -I./host/inc -I/usr/local/cuda-11.7/include -I./cuda -c ./cuda/kernels.cu /bin/sh: 1: nvcc: not found make: [Makefile.Cuda:187: kernels] Error 127***
While using the next option:
make DEVICE=GPU NUMWI=256make DEVICE=GPU NUMWI=128Returned the outputs:
rm -f ./host/inc/performdocking.h ./host/src/performdocking.cpp Building adgpu_analysis ... g++ \ ./host/src/calcenergy.cpp ./host/src/getparameters.cpp ./host/src/main.cpp ./host/src/miscellaneous.cpp ./host/src/processgrid.cpp ./host/src/processligand.cpp ./host/src/processresult.cpp ./host/src/setup.cpp \ -std=c++11 -I./common -I./host/inc -DVERSION=\"v1.5-release\" \ -o./bin/adgpu_analysis \ -O3 -DTOOLMODE DEVICE is set to GPU CPU_INCLUDE_PATH is set to /usr/local/cuda-11.7/include/CL CPU_LIBRARY_PATH is set to /usr/local/cuda-11.7/lib64 GPU_INCLUDE_PATH is set to /usr/local/cuda-11.7/include GPU_LIBRARY_PATH is set to /usr/local/cuda-11.7/lib64 ./stringify_ocl_krnls.sh Stringified input header files: ./common/defines.h ./common/calcenergy_basic.h Stringified input non-kernel files: ./device/auxiliary_genetic.cl ./device/calcenergy.cl ./device/calcgradient.cl ./device/calcMergedEneGra.cl Stringified input kernel-files: ./device/kernel1.cl ./device/kernel2.cl ./device/kernel4.cl ./device/kernel3.cl ./device/kernel_sd.cl ./device/kernel_fire.cl ./device/kernel_ad.cl Stringified output file: host/inc/stringify.h ln -sf performdocking.h.OpenCL ./host/inc/performdocking.h ln -sf performdocking.cpp.OpenCL ./host/src/performdocking.cpp Building autodock_gpu_256wi ... g++ \ ./wrapcl/src/BufferObjects.cpp ./wrapcl/src/CommandQueues.cpp ./wrapcl/src/Programs.cpp ./wrapcl/src/Kernels.cpp ./wrapcl/src/Devices.cpp ./wrapcl/src/ImportSource.cpp ./wrapcl/src/Platforms.cpp ./wrapcl/src/Contexts.cpp ./wrapcl/src/ImportBinary.cpp ./wrapcl/src/listAttributes.cpp ./host/src/calcenergy.cpp ./host/src/getparameters.cpp ./host/src/main.cpp ./host/src/miscellaneous.cpp ./host/src/performdocking.cpp ./host/src/processgrid.cpp ./host/src/processligand.cpp ./host/src/processresult.cpp ./host/src/setup.cpp \ -std=c++11 -I./common -I./wrapcl/inc -I./host/inc -I./device -I/usr/local/cuda-11.7/include -L/usr/local/cuda-11.7/lib64 -DVERSION=\"v1.5-release\" \ -lOpenCL \ -o./bin/autodock_gpu_256wi \ -DGPU_DEVICE -DN256WI -O3 -DKRNL_SOURCE=./device/calcenergy.cl -DKRNL_DIRECTORY=./device -DKCMN_DIRECTORY=./common -DK1="gpu_calc_initpop" -DK2="gpu_sum_evals" -DK3="perform_LS" -DK4="gpu_gen_and_eval_newpops" -DK5="gradient_minSD" -DK6="gradient_minFire" -DK7="gradient_minAD"
But if I try to run the compiled program, it does not work:
./autodock_gpu_256wi -L ../acarbose_pubchem_minimized.pdbqt -M ../aGluc_3top_md_first_frame.maps.fld --nrun 50AutoDock-GPU version: v1.5-release Running 1 docking calculation Kernel source used for development: ./device/calcenergy.cl Kernel string used for building: ./host/inc/stringify.h Kernel compilation flags: -I ./device -I ./common -DN256WI -cl-mad-enable Error: clGetPlatformIDs(): -1001Which GPU(s) are you running on and is Cuda or OpenCL used? NVIDIA GeForce RTX3050Ti (Laptop Lenovo LEGION5) targeting for Cuda and the CPU is an Ryzen 7 5800H.
Which driver version and if applicable, which Cuda version are you using?: NVIDIA driver 516.94 and CUDA 11.7 following the instructions for installing CUDA in WSL2 from https://docs.nvidia.com/cuda/wsl-user-guide/index.html
When compiling AutoDock-GPU, are
GPU_INCLUDE_PATHandGPU_LIBRARY_PATHset? Are both environment variables set to the correct directories, i.e. corresponding to the correct Cuda version or OpenCL library?On Ubuntu 20.04 LTS I set the environment variables of the file .bashrc in home as:
Setting CUDA for AutoDock-GPU
export CUDAROOT=/usr/local/cuda-11.7 export GPU_INCLUDE_PATH=$CUDAROOT/include export GPU_LIBRARY_PATH=$CUDAROOT/lib64 export CPU_INCLUDE_PATH=$CUDAROOT/include/CL export CPU_LIBRARY_PATH=$CUDAROOT/lib64 The files compiled with DIVICE=GPU but the executable did not work.
Then I tried something slightly different on Ubuntu 22.04 LTS:
Setting environment variables for ADGPU
export CUDAROOT=/usr/local/cuda export PATH=$CUDAROOT/bin/:$PATH export GPU_INCLUDE_PATH=$CUDAROOT/include export GPU_LIBRARY_PATH=$CUDAROOT/lib64 The files did not compile, but again the program did not work.
Files:
aGluc_3top_md_first_frame.maps.zip