tests of ABINIT 8.8.2 with CUDA almost fails.  [SOLVED]

option, parallelism,...

Moderators: fgoudreault, mcote

Forum rules
Please have a look at ~abinit/doc/config/build-config.ac in the source package for detailed and up-to-date information about the configuration of Abinit 8 builds.
For a video explanation on how to build Abinit 7.x for Linux, please go to: http://www.youtube.com/watch?v=DppLQ-KQA68.
IMPORTANT: when an answer solves your problem, please check the little green V-like button on its upper-right corner to accept it.
Locked
ssutoh
Posts: 2
Joined: Mon Sep 03, 2018 7:31 am

tests of ABINIT 8.8.2 with CUDA almost fails.

Post by ssutoh » Tue Sep 04, 2018 6:31 am

tests of ABINIT with CUDA almost fails.

Hi everybody.

All of the tests of ABINIT 8.8.2 have passed,
but the tests of ABINIT with CUDA almost fails.

The following shows how to build and test execute.

(1) build
# cat build.ac
FC=mpiifort
CC=mpiicc
CXX=mpiicpc
enable_mpi="yes"
enable_mpi_io="yes"
enable_gpu="yes"
with_gpu_flavor="cuda-double"
with_gpu_prefix="/usr/local/cuda-9.0"
# ./configure --with-config-file="./build.ac"
# make


Information on my GPU card and compiler is as follows.

# deviceQuery
・・・
Device 0: "Tesla V100-PCIE-16GB"
CUDA Driver Version / Runtime Version 9.0 / 9.0
CUDA Capability Major/Minor version number: 7.0
・・・

# mpiifort --version
ifort (IFORT) 17.0.4 20170411

# mpiicc --version
icc (ICC) 17.0.4 20170411

# mpiicpc --version
icpc (ICC) 17.0.4 20170411


(2) test_fast
# make test_fast
・・・
==> The run finished cleanly.
Moreover, comparison of the total energy, and other (few) relevant quantities with reference values has been successful.
This does not mean that no problem is present, however.
Please run the complete set of ABINIT tests to gain a better confidence in your installation.
・・・


(3) tests
# python runtests.py
・・・
[fast][t00][np=1]: succeeded [file=t00.out]
[fast][t01][np=1]: succeeded [file=t01.out]
[fast][t02][np=1]: succeeded [file=t02.out]
[fast][t04][np=1]: failed: erroneous lines 8 > 0 [file=t04.out]
No YAML Error found in [fast][t04][np=1]
[fast][t03][np=1]: failed: erroneous lines 8 > 0 [file=t03.out]
No YAML Error found in [fast][t03][np=1]
[fast][t17][np=1]: failed: erroneous lines 40 > 0 [file=t17.out]
No YAML Error found in [fast][t17][np=1]
・・・


(4) tests gpu
# python runtests.py gpu -j4
...
Summary: failed=5, succeeded=0, passed=0, skipped=2, disabled=0
...


I get lots of tests failures.
Especially the fatal thing is that all gpu tests have failed and none of the successful gpu tests succeeded.
If there are any other particular points, please point them out.

User avatar
jbeuken
Posts: 365
Joined: Tue Aug 18, 2009 9:24 pm
Contact:

Re: tests of ABINIT 8.8.2 with CUDA almost fails.  [SOLVED]

Post by jbeuken » Wed Nov 21, 2018 8:53 am

sorry, but Abinit8 with CUDA 9 is not yet full supported...
we know that there are still some problems...
jmb
------
Jean-Michel Beuken
Computer Scientist

ssutoh
Posts: 2
Joined: Mon Sep 03, 2018 7:31 am

Re: tests of ABINIT 8.8.2 with CUDA almost fails.

Post by ssutoh » Wed Dec 05, 2018 9:09 am

jbeuken wrote:sorry, but Abinit8 with CUDA 9 is not yet full supported...
we know that there are still some problems...
jmb


Thank you for your answer.
Since I understood the situation, I will close this inquiry.

Locked