Alright, I'm trying to install the software on a High Performance Computer, however when running it with the example test file, it gives a segmentation fault. Here is the output.

/Vina-GPU --config ./input_file_example/2bm2_config.txt
Reading input ... done.
Setting up the scoring function ... done.
Analyzing the binding site ... done.
Using random seed: 721271727

Platform: NVIDIA CUDA
Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Build kernels from source
OpenCL version: 3.0
Search depth is set to 6
Segmentation fault (core dumped)

I put it into the debugger and got this as an output.

Device: Tesla V100S-PCIE-32GB
[New Thread 0x2aaaaf02b700 (LWP 38126)]
[New Thread 0x2aaaaf22c700 (LWP 38128)]
[New Thread 0x2aaaaf42d700 (LWP 38129)]
[New Thread 0x2aaaaf62e700 (LWP 38130)]
[New Thread 0x2aaaaf82f700 (LWP 38131)]
[New Thread 0x2aaaafa30700 (LWP 38132)]

Build kernels from source
OpenCL version: 3.0
[New Thread 0x2aaaafc31700 (LWP 38135)]
Search depth is set to 6

Program received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x2aaaaf02b700 (LWP 38126)]
0x00002aaaadcc5fa0 in ?? () from /lib64/libnvidia-opencl.so.1
Missing separate debuginfos, use: debuginfo-install glibc-2.17-325.el7_9.x86_64 nvidia-driver-branch-495-cuda-libs-495.29.05-1.el7.x86_64 ocl-icd-2.2.12-1.el7.x86_64
(gdb) bt
#0  0x00002aaaadcc5fa0 in ?? () from /lib64/libnvidia-opencl.so.1
#1  0x00002aaaadde74c0 in ?? () from /lib64/libnvidia-opencl.so.1
#2  0x00002aaaadd8974f in ?? () from /lib64/libnvidia-opencl.so.1
#3  0x00002aaaadc8649c in ?? () from /lib64/libnvidia-opencl.so.1
#4  0x00002aaaaddfef96 in ?? () from /lib64/libnvidia-opencl.so.1
#5  0x00002aaaadd83376 in ?? () from /lib64/libnvidia-opencl.so.1
#6  0x00002aaaabc53ea5 in start_thread () from /lib64/libpthread.so.0
#7  0x00002aaaac17eb0d in clone () from /lib64/libc.so.6

The computer hardware specs are,

• 4 NVIDIA Tesla V100's
• 2x AMD EPYC 7601 processors (32 cores each with hyper-threading)

We are running this in "CentOS Linux 7 (Core)".

I will continue to investigate this issue further. However by the looks of things, this looks more like an issue with our Cuda or other libraries we are linking to instead of an issue with Vina-GPU.

When I install the software ,I meet the problem

Reading input ... done.
Setting up the scoring function ... done.
Analyzing the binding site ... done.
Using random seed: -181661601

Platform: NVIDIA CUDA
Platform 0 device name:Tesla V100-SXM2-16GB

Platform 0 global memory size:16.945512 GB

Platform 0 local memory size:49.152000 KB

Search depth is set to 5
Segmentation fault (core dumped)

Have you tested your program in the A100 (Ampere Architecture) + Linux ?If did ,could you tell me the version of CUDA, GCC , boost, OS .

I've recently compiled Vina-GPU however when I run the command: ./Vina-GPU --config ./input_file_example/2bm2_config.txt

I get the error: `Reading input ... done. Setting up the scoring function ... done. Analyzing the binding site ... done. Using random seed: -108247774

Platform: NVIDIA CUDA Device: Tesla T4

Search depth is set to 6 Err-30:CL_INVALID_VALUE`

My gcc and g++ are both 10.3.0 and cuda is 11.1 (installed via conda)

Background to how I've compiled it:

1. I've installed the following packages conda install -c conda-forge rdkit conda install -c hcc adfr-suite

2. Downloaded and decompressed Boost and Vina-GPU and ran ulimit -s 8192

3. Changed the Makefile content as: BOOST_LIB_PATH=/content/boost_1_77_0 OPENCL_LIB_PATH=/usr/local/cuda and then cd Vina-GPU-main/ && make clean && make I've got several warning and deprecation messages but no errors and "Vina-GPU" was generated.

Any ideas about what might be causing the problem?

Hi, I am not able to compile the code on Linux system. When I do make source, I get the following error messages:

/usr/bin/ld: cannot find -lboost_program_options /usr/bin/ld: cannot find -lboost_system /usr/bin/ld: cannot find -lboost_filesystem collect2: error: ld returned 1 exit status make: *** [source] Error 1

Because I do not have sudo privileges, I downloaded boost 1.80 and stored it under the root folder of the repository. I have also installed CUDA toolkit (11.7) under upper folder relative to the repository. I have changed Makefile accordingly, so its first several lines read:

BOOST_LIB_PATH=boost_1_80_0 OPENCL_LIB_PATH=../cuda_11_7 OPENCL_VERSION=-DOPENCL_3_0 GPU_PLATFORM=-DNVIDIA_PLATFORM

Using the same settings I actually can compile the code in an Ubuntu 20.04.2LTS machine which I have root privileges. But I cannot compile the code on the cluster that has Red Hat 4.8.5-44. I am using GCC version 10.2.

Any help is greatly appreciated!

Hi! I'm using CPU version of Vina (v 1.2.3) with exhaustiveness=8 and all available CPU on linux (12 cores on AMD 5990x). then I use Vina-GPU, how much "threads" i need to set to get the same (or related) docking accuracy? how exhaustiveness *cpus is related to threads?

(base) [dalg@localhost Vina-GPU-main]$ ./Vina-GPU --config ./input_file_example/2bm2_config.txt ################################################################# # If you used Vina-GPU in your work, please cite: # #Shidi, Tang, Chen Ruiqi, Lin Mengru, Lin Qingde, # #Zhu Yanxiang, Wu Jiansheng, Hu Haifeng, and Ling Ming. # #Accelerating AutoDock VINA with GPUs. ChemRxiv (2021).Print. # # And also the origin AutoDock Vina paper: # # O. Trott, A. J. Olson, # # AutoDock Vina: improving the speed and accuracy of docking # # with a new scoring function, efficient optimization and # # multithreading, Journal of Computational Chemistry 31 (2010) # # 455-461 # # DOI 10.1002/jcc.21334 # #################################################################

Reading input ... done. Setting up the scoring function ... done. Analyzing the binding site ... done. Using random seed: 1167282562 段错误 (核心已转储)

OS centos 8 stream CUDA 11.4 GPU Tesla V100-PCIE gcc 8.5.0

When I ran this command, ./Vina-GPU --config ./input_file_example/2bm2_config.txt

I got this output,

Reading input ... done.
Setting up the scoring function ... done.
Analyzing the binding site ... done.
Using random seed: -1907456286

Platform: NVIDIA CUDA
Search depth is set to 5
Segmentation fault (core dumped)`

I am using Ubuntu 20.04, made sure the boost and cuda tool kit was the exact same version as was said in the readme.

I put the same command in the debugger and got this for an output.

`Using random seed: -705894464
[New Thread 0x7ffff5bf6700 (LWP 2570)]

Platform: NVIDIA CUDA[New Thread 0x7fffde835700 (LWP 2571)]
[New Thread 0x7fffde034700 (LWP 2572)]
[New Thread 0x7fffdd833700 (LWP 2573)]
[New Thread 0x7fffdd032700 (LWP 2574)]
[New Thread 0x7fffdc831700 (LWP 2575)]
[New Thread 0x7fffd7fff700 (LWP 2576)]
[New Thread 0x7fffd77fe700 (LWP 2577)]

[New Thread 0x7fffd6ffd700 (LWP 2578)]
Search depth is set to 5

Thread 2 "Vina-GPU" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7ffff5bf6700 (LWP 2570)]
0x00007ffff78444a7 in __GI_fseek (fp=0x0, offset=0, whence=2) at fseek.c:35
35	fseek.c: No such file or directory.
(gdb) bt
#0  0x00007ffff78444a7 in __GI_fseek (fp=0x0, offset=0, whence=2) at fseek.c:35
#1  0x00005555555f3116 in SetupBuildProgramWithBinary(_cl_context*, _cl_device_id**, char const*) ()
#2  0x00005555555b765a in monte_carlo::operator()(model&, boost::ptr_vector<output_type, boost::heap_clone_allocator, void>&, precalculate const&, igrid const&, precalculate const&, igrid const&, vec const&, vec const&, incrementable*, boost::random::mersenne_twister_engine<unsigned int, 32ul, 624ul, 397ul, 31ul, 2567483615u, 11ul, 4294967295u, 7ul, 2636928640u, 15ul, 4022730752u, 18ul, 1812433253u>&) const ()
#3  0x00005555555c5e77 in boost::detail::thread_data<boost::reference_wrapper<parallel_for<parallel_iter<parallel_mc_aux, boost::ptr_vector<parallel_mc_task, boost::heap_clone_allocator, void>, parallel_mc_task, true>::aux, true>::aux> >::run() ()
#4  0x00005555555f49c2 in thread_proxy ()
#5  0x00007ffff79ce609 in start_thread (arg=<optimized out>) at pthread_create.c:477
#6  0x00007ffff78d8293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
(gdb) f 6
#6  0x00007ffff78d8293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
95	../sysdeps/unix/sysv/linux/x86_64/clone.S: No such file or directory.

Dears

I am getting the following error:

Platform: NVIDIA CUDA Build kernels from sourceError: failed to open file : ./OpenCL/src/kernels/code_head.cppError: failed to open file : ./OpenCL/src/kernels/mutate_conf.cppError: failed to open file : ./OpenCL/src/kernels/matrix.cppError: failed to open file : ./OpenCL/src/kernels/quasi_newton.cppError: failed to open file Segmentation fault (core dumped)

My installation output is this:

gcc -o Vina-GPU -I/home/zapata/boost_1_77_0 -I/home/zapata/boost_1_77_0/boost -I./lib -I./OpenCL/inc -I/usr/local/cuda-11.5/include ./main/main.cpp -O3 ./lib/.cpp ./OpenCL/src/wrapcl.cpp /home/zapata/boost_1_77_0/libs/thread/src/pthread/thread.cpp /home/zapata/boost_1_77_0/libs/thread/src/pthread/once.cpp -lboost_program_options -lboost_system -lboost_filesystem -lOpenCL -lstdc++ -lm -lpthread -L/home/zapata/boost_1_77_0/stage/lib -L/usr/local/cuda-11.5/lib64 -DOPENCL_3_0 -DNVIDIA_PLATFORM -DBUILD_KERNEL_FROM_SOURCE In file included from ./lib/monte_carlo.cpp:34: /home/zapata/boost_1_77_0/boost/progress.hpp:23:108: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp>" ) ^ In file included from /home/zapata/boost_1_77_0/boost/progress.hpp:25, from ./lib/monte_carlo.cpp:34: /home/zapata/boost_1_77_0/boost/timer.hpp:21:70: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp>" ) ^ ./lib/monte_carlo.cpp: In member function ‘void monte_carlo::operator()(model&, output_container&, const precalculate&, const igrid&, const precalculate&, const igrid&, const vec&, const vec&, incrementable, rng&) const’: ./lib/monte_carlo.cpp:481:66: warning: narrowing conversion of ‘(fl)((const monte_carlo*)this)->monte_carlo::hunt_cap.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] float hunt_cap_float[3] = {hunt_cap[0], hunt_cap[1], hunt_cap[2]}; ^ ./lib/monte_carlo.cpp:481:66: warning: narrowing conversion of ‘(fl)((const monte_carlo*)this)->monte_carlo::hunt_cap.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] ./lib/monte_carlo.cpp:481:66: warning: narrowing conversion of ‘(fl)((const monte_carlo*)this)->monte_carlo::hunt_cap.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] ./lib/monte_carlo.cpp:482:79: warning: narrowing conversion of ‘authentic_v.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] float authentic_v_float[3] = { authentic_v[0],authentic_v[1], authentic_v[2] }; ^ ./lib/monte_carlo.cpp:482:79: warning: narrowing conversion of ‘authentic_v.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] ./lib/monte_carlo.cpp:482:79: warning: narrowing conversion of ‘authentic_v.vec::operator’ from ‘fl’ {aka ‘double’} to ‘float’ inside { } [-Wnarrowing] In file included from ./lib/parallel_progress.h:26, from ./lib/parallel_mc.cpp:26: /home/zapata/boost_1_77_0/boost/progress.hpp:23:108: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp>" ) ^ In file included from /home/zapata/boost_1_77_0/boost/progress.hpp:25, from ./lib/parallel_progress.h:26, from ./lib/parallel_mc.cpp:26: /home/zapata/boost_1_77_0/boost/timer.hpp:21:70: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp>" ) ^ In file included from ./lib/parallel_progress.h:26, from ./lib/parallel_progress.cpp:25: /home/zapata/boost_1_77_0/boost/progress.hpp:23:108: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp> or <boost/timer/progress_display.hpp>" ) ^ In file included from /home/zapata/boost_1_77_0/boost/progress.hpp:25, from ./lib/parallel_progress.h:26, from ./lib/parallel_progress.cpp:25: /home/zapata/boost_1_77_0/boost/timer.hpp:21:70: note: #pragma message: This header is deprecated. Use the facilities in <boost/timer/timer.hpp> instead. BOOST_HEADER_DEPRECATED( "the facilities in <boost/timer/timer.hpp>" ) ^ ./OpenCL/src/wrapcl.cpp: In function ‘_cl_program* SetupBuildProgramWithBinary(cl_context, _cl_device_id**, const char*)’: ./OpenCL/src/wrapcl.cpp:261:10: warning: ignoring return value of ‘size_t fread(void*, size_t, size_t, FILE*)’, declared with attribute warn_unused_result [-Wunused-result] fread(binary_buffer, sizeof(char), program_size, program_handle); ~~~~~^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

My cuda version is 11.5 Kernel 4.15.0-162-generic Ubuntu 18.04 DOPENCL_3_0

I would appreciate your help to figure out how to solve this issue.

Thanks in advance!

Hi,

Just wondering if there were any other flags or special config arugments to better support docking compounds with macrocycles? The failure rate for docking macrocycle compounds is pretty high ~14%. Is there anything more than the default parameters that might help bring this docking failure rate down?

Thank you.

So while testing the program we noticed that it wasn't using all of the gpus on our system. It looks like Vina-GPU is successful detecting all the gpus based on this output.

Using random seed: 1833385468

Platform: NVIDIA CUDA
Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Device: Tesla V100S-PCIE-32GB

Search depth is set to 6

However, Vina-GPU is using only one gpu based on some diagnostics tools that I used. It's probably something to do with how things are configured in our system that I'll need to find, however I just wanted to confirm real quick if Vina-GPU is designed to fully utilize multiple gpus in a system or just a singular one. I'll assume its designed to use multiple, however if it isn't designed with multiple gpus in mind, that would be nice to know so I don't go on a wild goose chase.

How to run vina-gpu on windows 10. I tried to use this software, and it felt unusable. The error appeared Platform: NVIDIA CUDACould not open file: C:\Users\ADMINI~1\AppData\Local\Temp\dep-c25713.d. How did this error happen? ? How should I configure my computer to use vina-gpu on windows 10? Can you describe in detail how to set these steps? E.g: add ./lib ./OpenCL/inc $(YOUR_BOOST_LIBRARY_PATH)/boost $(YOUR_CUDA_TOOLKIT_LIBRARY_PATH)/CUDA/v11.5/include in the include directories;

add $(YOUR_BOOST_LIBRARY_PATH)/stage/lib $(YOUR_CUDA_TOOLKIT_PATH)/CUDA/lib/Win32in the addtional library;

add OpenCL.lib in the additional dependencies。

Hi,

I was wondering if there were any parameters that could be given to reduce the memory consumption of Vina-GPU during docking to avoid the Err-6:CL_OUT_OF_HOST_MEMORY as I have other concurrent processes running on my GPU and can only afford about 4GB for Vina-GPU... Thank you.

While doing large scale virtual screening using Vina-GPU, many molecules fail because of the assertion error:

CUDA_VISIBLE_DEVICES=1 ./Vina-GPU --receptor protein.pdbqt --ligand 1.pdbqt --num_modes 1 
--center_x 7.2395000000000005 --center_y -0.6745000000000001 --center_z 31.628500000000003
--size_x 9.583 --size_y 19.809 --size_z 12.779 --thread 10000 --out res.pdbqt: 

Vina-GPU: ./lib/coords:31 fl rmsd_upper_bound(const vecv&, const vecv&): Assertion `a.size() == b.size()` failed

@Glinttsd Do you have any idea of this? Is there any configuration that can prevent it or does retry work?

Thanks in advance!

Hi,

I am running autodock-vina tutorial data (https://github.com/ccsb-scripps/AutoDock-Vina/tree/develop/example/flexible_docking) and Vina-GPU gives an error when running with the --flex option.

Any ideas?

Vina-GPU: ./lib/monte_carlo.cpp:309: void monte_carlo::operator()(model&, output_container&, const precalculate&, const igrid&, const precalculate&, const igrid&, const vec&, const vec&, incrementable*, rng&) const: Assertion `m.num_other_pairs() == 0' failed.

Hi, I tried to install all the dependencies, e.g. Boost, OpenCL, CUDA, etc., according to README instructions, but I didn't succeed.

The hardest part I am facing is the OpenCL lib, which is not in the CUDA folder as instructed after I installed NVidia driver and CUDA library.

I wonder if you could write in detail about how to install the dependencies on a newly provisioned Linux system (e.g. Ubuntu 20.04 or 22.04) Or could you help to provide a Dockerfile in order for us to build the Docker images?

Dear Glinttsd I faced with this "Err-6:CL_OUT_OF_HOST_MEMORY" issue on your test system and any of mine.

Info: Ubuntu 20.04, boost_1_80

NVIDIA-SMI 510.60.02 Driver Version: 510.60.02 CUDA Version: 11.6 NVIDIA 3080 967MiB / 10240MiB

total used free shared buff/cache available Mem: 31Gi 10Gi 7,9Gi 550Mi 12Gi 19Gi Swap: 2,0Gi 439Mi 1,6Gi

Hi I have problem, how to solved this: ./Vina-GPU --config ./input_file_example/2bm2_config.txt ./Vina-GPU: error while loading shared libraries: libboost_program_options.so.1.79.0: cannot open shared object file: No such file or directory