Problems with testing of dalton installation

Problems with Dalton installation? Find answers or ask for help here
Post Reply
Evg
Posts: 4
Joined: 03 May 2015, 16:47
First name(s): Evgeniy
Last name(s): Gromov
Affiliation: TC Universitu of Heidelberg
Country: Germany

Problems with testing of dalton installation

Post by Evg » 04 May 2015, 10:29

Dear Developers and Users of Dalton,

I am experiencing problems with testing my Dalton2015 installation. All first 42 tests failed.
Then tests 51-54 also failed and so on. In the LastTest.log.tmp file there is the following error:

---------------------------------------------------------------------------------------------------------------
Traceback (most recent call last):
File "/home/evgeniy/SYSAdmin/DALTON-Source/DALTON/test/dft_ac_grac/test", line 15, in ?
from runtest_dalton import Filter, TestRun
File "/home/evgeniy/SYSAdmin/DALTON-Source/DALTON/test/dft_ac_grac/../runtest_dalton.py", line 13
with open(log_file, 'w') as f:
^
SyntaxError: invalid syntax
<end of output>
Test time = 0.05 sec
----------------------------------------------------------
Test Failed.
------------------------------------------------------------------

The compilation was done with gfortran, with the setup command:

./setup --fc=gfortran --cc=gcc --cxx=g++ --int64

No blas/lapack was used so far. The version of gfortran is 4.8.1.

In fact I would like to use ACML. What settings should I specify for that? Did anybody
manage to successfully install Dalton2015 with ACML and gfortran?

Many thanks!
Evg

lyzhao
Posts: 74
Joined: 11 Nov 2013, 00:36
First name(s): Youzhao
Last name(s): Lan
Affiliation: Institute of Physical Chemistry
Country: China

Re: Problems with testing of dalton installation

Post by lyzhao » 04 May 2015, 11:22

File "/home/evgeniy/SYSAdmin/DALTON-Source/DALTON/test/dft_ac_grac/../runtest_dalton.py", line 13
with open(log_file, 'w') as f:
^
SyntaxError: invalid syntax
It seems that your python is too old.
try to update your python

Lan

taylor
Posts: 600
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: Problems with testing of dalton installation

Post by taylor » 04 May 2015, 15:52

I agree with lyzhao that it is almost certain that you need to update your version of python, and the comment I am about to make is quite independent of your failing tests. It is a more general remark about your build.

I notice you are building with 64-bit integers. Do you have a particular reason for doing this? For most purposes it is not necessary to have 64-bit integers (for example, packing of integral labels is automatically adjusted to accommodate the size of the basis), and some functionality will not benefit anyway (for example, MCSCF Hessians are limited to 255 basis functions currently irrespective of whether 32-bit or 64-bit integers are used). So in most situations 64-bit integers are unnecessary, and in some cases (particularly when building MPI-parallel versions) can be a handicap, because building e.g. OpenMPI as completely 64-bit integer aware is not always straightforward.

So unless you have a specific need for 64-bit integers --- as far as I know the only reference to this in the manual is some advantage in Cholesky decomposition calculations, but I admit to not knowing exactly what that advantage is; of course if you want to address more than 2 gigawords (i.e., 16GB) of memory you will need to go beyond 32-bit addressing --- it is probably easier and safer to just build with 32-bit integers. Qualifying my last sentence further, I know that the Intel compiler and associated libraries, because the highest level addressing is based on "words" not bytes, can address 16GB using 32-bit integers. I do not routinely use gfortran, and have not used it on anything with e.g. 16GB of memory, so I cannot be sure in that case.

Best regards
Pete

Evg
Posts: 4
Joined: 03 May 2015, 16:47
First name(s): Evgeniy
Last name(s): Gromov
Affiliation: TC Universitu of Heidelberg
Country: Germany

Re: Problems with testing of dalton installation

Post by Evg » 04 May 2015, 16:51

Many thanks for your replies!

Yes, indeed the problem was the too old python. After updating, it is better now, although a few tests still didn't pass.
I will report which exactly tests didn't pass later on when the testing is complete.

Concerning the 32 vs 64 issue, the main reason for 64 is that one can use more memory.
If you say that 16 GB should be enough for most of the calculations I will then opt for the
32 version, keeping in mind the 16 GB limit.

Best regards,
Evgeniy

Evg
Posts: 4
Joined: 03 May 2015, 16:47
First name(s): Evgeniy
Last name(s): Gromov
Affiliation: TC Universitu of Heidelberg
Country: Germany

Re: Problems with testing of dalton installation

Post by Evg » 04 May 2015, 17:45

Just a short follow-up:

is python needed/called during a normal dalton run (not testing run)?

taylor
Posts: 600
Joined: 15 Oct 2013, 05:37
First name(s): Peter
Middle name(s): Robert
Last name(s): Taylor
Affiliation: Tianjin University
Country: China

Re: Problems with testing of dalton installation

Post by taylor » 05 May 2015, 14:22

I should perhaps have qualified my earlier reply by saying that if one has a lot of memory and chooses to enable the "store integrals in memory" option then it may well be that 64-bit integers would be useful for addressing. Since I never enable that option, because it is largely irrelevant for the calculations I am interested in, I cannot comment.

I am not stating that 16GB will cover all eventualities, even without storing integrals. For example, one could construct a CASSCF calculation with 10 billion determinants: this would typically require 80GB for the current coefficient vector guess, another 80GB for the product of the Hamiltonian matrix and this guess, and 80GB for the exact diagonal of the Hamiltonian, which is needed for the original Davidson preconditioner. Such a calculation would therefore need at least 240GB of memory. In some sense this is the classic "how long is a piece of string" question! Without knowing what you are trying to calculate, and at what level, it is hardly possible to answer the question "is 16GB [say] enough"? But if you are planning calculations at the level of SCF or DFT, or CC with at least not gigantic systems, and/or the calculation of properties using the associated response methods, I would be inclined to suggest 16GB is enough.

Best regards
Pete

bast
Posts: 1210
Joined: 26 Aug 2013, 13:22
First name(s): Radovan
Last name(s): Bast
Affiliation: none
Country: Germany

Re: Problems with testing of dalton installation

Post by bast » 05 May 2015, 19:35

Evg wrote:Just a short follow-up:

is python needed/called during a normal dalton run (not testing run)?
no it is not. it is only used during configuration, testing, and source code generation.

Post Reply

Who is online

Users browsing this forum: No registered users and 1 guest