Windows Anaconda nightly build failing on SAPT0 calculations

1

In reference to the 1.5 dev version of psi4 (installed as per my post here) psi4 is failing with the error:

PSIO_ERROR: unit = 193, errval = 8

when attempting to run SAPT0 calculations that have been successfully run in previous versions of psi4. In particular the calculation I am attempting was previously run with Psi4 1.3.2 on WSL.

The last few lines of the output file are as follows:

   SAPT0
Ed Hohenstein
 6 June 2009

  Orbital Information

NSO        =       336
NMO        =       336
NRI        =      1116
NOCC A     =        37
NOCC B     =        13
FOCC A     =         9
FOCC B     =         3
NVIR A     =       299
NVIR B     =       323

The only error message in Anaconda prompt is PSIO_ERROR: unit = 193, errval = 8

Any ideas what is going wrong?

Tried changing scratch directory from default, no success.

2

As an update I had:

memory 4 GB

At the top of my input file. Removing this entirely, or setting to larger value (5, 6 or 40 GB) allows the calculation to proceed past the point above and does not result in an error message. Basically anything greater than 4 GB of memory works, or the default 500 MB…

Using the default 500 MB memory, the calculation appears to use ~ 500 MB. When setting a value larger than 4, the calculation tops out at around 1.1 GB.

3

Although the message you reported is the only error message explicitly written to stderr, there is a throw statement that would have been triggered and that is not mentioned in your report.

errval = 8 is Psi trying to close a file that’s already closed. Although the file opening and closing logic may depend on memory estimates, an error like this shouldn’t be directly dependent on memory allocation. unit = 193 says this is about the alpha-alpha block of density-fitted integrals.

That’s about as far as I can go without an input or output file that demonstrates the bug.

4

There is no other error or statement with this issue, not in the Anaconda prompt, nor in the output file.
using --verbose doesn’t give anything extra either.

(psi4environment) *****\working_directory\WORK>psi4 -n 10 -input test_input.dat -output test_output.dat
PSIO_ERROR: unit = 193, errval = 8

The following input file generates the error using a simple water dimer as an example.

memory 4 GB

molecule water_dimer {
     0 1
     O  -1.551007  -0.114520   0.000000
     H  -1.934259   0.762503   0.000000
     H  -0.599677   0.040712   0.000000
     --
     0 1
     O   1.350625   0.111469   0.000000
     H   1.680398  -0.373741  -0.758561
     H   1.680398  -0.373741   0.758561

     units angstrom
     no_reorient
     symmetry c1
}

set basis jun-cc-pvdz

energy('sapt0')

Changing memory to the following values results in the same error:
2.2 GB
2.3 GB
2.4 GB
2.5 GB
3 GB
3.5 GB
4 GB
4.1 GB
4.2 GB

Any memory value less than or equal to 2.1 GB or greater than or equal to 4.3 GB does not result in an error.

5

That’s a much simpler input file than I expected.

When I try to reproduce this on my Mac, I can’t, but you already said this ran fine on WSL, so this isn’t very informative. Unfortunately, I don’t have a Windows machine, so the best I can do is transfer this to our issue tracker, where the SAPT developers will be more likely to see this.

Can you try again without -n 10? I’m curious if parallelization is involved.

6

Same behaviour with or without parallelisation.

I also have a much more complex SAPT calculation that I’m trying to run. No value of memory I have yet found will allow the calculation to complete. All 4, 8 25 & 40 GB all fail with the same error.

I have 1.4.1 & 1.5a1 running on a linux box which runs without error, so I will switch to that for the time being.