Re: [AMBER] installing CUDA driver with ubuntu 20.04

From: Enrico Martinez via AMBER <amber.ambermd.org>
Date: Wed, 3 Aug 2022 18:05:02 +0200

Dear David,
I've just tried to install amber-22 on my ubuntu machine with cuda
11.7 and Driver Version: 515.43.04.
First, as you indicated below, I edited cmake/CudaConfig.cmake
indicating the version 11.7 of the cuda driver.

then I untar amber-tools and amber to the same directory and run cmake
indicating cuda driver:

cd amber22_src/build
./run_cmake -DCUDA=TRUE -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-11.7


that correctly detected cuda 11.7 at the beginning of the compilation
however in the end, cmake reported that cuda has not been detected:

-- Features:
-- MPI: OFF
-- OpenMP: OFF
-- CUDA: OFF
-- Build Shared Libraries: ON
-- Build GUI Interfaces: ON
-- Build Python Programs: ON
-- -Python Interpreter: Internal Miniconda (version 3.9)
-- Build Perl Programs: ON
-- Build configuration: RELEASE
-- Target Processor: x86_64
-- Build Documentation: OFF
-- Sander Variants: normal LES API LES-API
-- Install location: /home/enrico/amber22/
-- Installation of Tests: ON

-- Compilers:
-- C: GNU 9.4.0 (/usr/bin/gcc)
-- CXX: GNU 9.4.0 (/usr/bin/g++)
-- Fortran: GNU 9.4.0 (/usr/bin/gfortran)

-- Building Tools:
-- addles amberlite ambpdb antechamber cifparse cphstats cpptraj emil
etc few gbnsr6 gem.pmemd kmmd leap lib mdgx mm_pbsa mmpbsa_py moft nab
ndiff-2.00 nfe-umbrella-slice nmode nmr_aux packmol_memgen paramfit
parmed pbsa pdb4amber pmemd pymsmt pysander pytraj reduce rism sander
saxs sebomd sff sqm xray xtalutil

-- NOT Building Tools:
-- quick - BUILD_QUICK is not enabled
-- tcpb-cpp - BUILD_TCPB is not enabled
-- tcpb-cpp/pytcpb - BUILD_TCPB is not enabled
-- reaxff_puremd - BUILD_REAXFF_PUREMD is not enabled
-- gpu_utils - Requires CUDA

May you suggest some solution in order that I could use GPUs with
pmemd ? Do I need to additionally activate the mpi option for the
compilation assuming that I do not plan to use both GPUs together?
Many thanks in advance
Cheers,
Enrico

Il giorno mer 20 lug 2022 alle ore 16:31 David A Case
<david.case.rutgers.edu> ha scritto:
>
> On Tue, Jul 19, 2022, Enrico Martinez via AMBER wrote:
> >
> >nvcc --version
> >nvcc: NVIDIA (R) Cuda compiler driver
> >Copyright (c) 2005-2022 NVIDIA Corporation
> >Built on Tue_May__3_18:49:52_PDT_2022
> >Cuda compilation tools, release 11.7, V11.7.64
> >Build cuda_11.7.r11.7/compiler.31294372_0
>
> To use Cuda 11.7, you will need to edit cmake/CudaConfig.cmake and change
> references to "11.6" to "11.7". An update to do this is in the works, but
> it's quite a simple change.
>
> ...good luck...dac
>

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 04 2022 - 13:37:56 PDT
Custom Search