OpenFOAM plugin library selection issue
November 14, 2015 at 04:33 #1294
I had install cuda 7.0 and I cannot find “lcudart, lcublas, lcusparse” in “/usr/local/cuda/lib”. When I add “-lcudart -lcublas -lcusparse -L/usr/local/cuda/lib \” in my “Make/options”, I got the error “/usr/bin/ld: cannot find -lcudart -lcublas -lcusparse”
When I remove that line, the Laplace example can run perfectly and the results are showing below. I would like to know if I success execute the program on GPU. What does the libraries “lcudart, lcublas, lcusparse” do?
| ========= | |
| \\ / F ield | OpenFOAM: The Open Source CFD Toolbox |
| \\ / O peration | Version: 3.0.0 |
| \\ / A nd | Web: http://www.OpenFOAM.org |
| \\/ M anipulation | |
Build : 3.0.0-8b097f6d8dd9
Exec : ./Laplace
Date : Nov 13 2015
Time : 22:14:33
Host : “cchang11-All-Series”
PID : 7189
Case : /home/cchang11/paralution-1.0.0/src/examples/OpenFOAM/Laplace
nProcs : 1
sigFpe : Enabling floating point exception trapping (FOAM_SIGFPE).
fileModificationChecking : Monitoring run-time modified files using timeStampMaster
allowSystemOperations : Allowing user-supplied system call operations
// * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * //
Create mesh for time = 0
Reading field T
Reading diffusivity DT
This version of PARALUTION is released under GPL.
By downloading this package you fully agree with the GPL license.
Number of CPU cores: 8
Host thread affinity policy – thread mapping on every core
Number of GPU devices in the system: 1
PARALUTION ver B1.0.0
PARALUTION platform is initialized
Accelerator backend: GPU(CUDA)
Selected GPU device: 0
Device number: 0
Device name: GeForce GTX 970
totalGlobalMem: 4095 MByte
compute capability: 5.2
Time = 1
PreSolveTime = 17.77 s PreSolveClockTime = 17 s
paralution_AMG(None): Solving for T, Initial residual = 1, Final residual = 9.30564e-11, No Iterations 13
PostSolveTime = 31.14 s PostSolveClockTime = 23 s
Contents of my Make/files are:
EXE = ./Laplace
Contents of my Make/options are:
EXE_INC = \
EXE_LIBS = \
Thank you very much,
November 19, 2015 at 16:42 #1298
- This topic was modified 2 years, 8 months ago by world2005.
The output looks good to me, your GPU is recognized by PARALUTION. Please check the user manual for detailed informations on the different objects info() functions, to check if the objects reside on the GPU during computation. For details on the different CUDA libraries, please have a look at the NVIDIA documentations.
You must be logged in to reply to this topic.