ecp-copa/comd

Classical molecular dynamics proxy application.

comd's Issues

YAML output not compliant with standard

After several hours trying to debug an issue while plotting performance results, thinking there was an issue with my R script, I have realised that the YAML output is not compliant with standard version 1.2. As is, when using a YAML parser there is a Scanner error: mapping values are not allowed in this context.

The problematic chunk of the output is for example:

Performance Results For Rank 0:
  Timer: total
    CallCount:  1
    AvgPerCall:   6.0009
    Total:        6.0009
    PercentLoop:   101.79
  Timer: loop
    CallCount:  1
    AvgPerCall:   6.0009
    Total:        6.0009
    PercentLoop:   101.79

It should instead output:

Performance Results For Rank 0:
  - Timer: total
    CallCount:  1
    AvgPerCall:   6.0009
    Total:        6.0009
    PercentLoop:   101.79
  - Timer: loop
    CallCount:  1
    AvgPerCall:   6.0009
    Total:        6.0009
    PercentLoop:   101.79

Makefile open mp failed to compile

While running make in src-openmp after copying sample I am getting this error.

Rebuilding dependencies...
./generate_info_header CoMD-openmp-mpi "mpicc" "-std=c99 -fopenmp -DDOUBLE -DDO_MPI -g -O5   " " -lm "
CC: 
Compiler version: unknown
make[1]: *** [Makefile:105: depend] Error 127
mpicc -std=c99 -fopenmp -DDOUBLE -DDO_MPI -g -O5    -c ljForce.c -o ljForce.o
/bin/bash: mpicc: command not found
make: *** [Makefile:80: ljForce.o] Error 127

Issue with random number generator.

Posting this issue on behalf of Guillaume:

I looked at the random generator used in CoMD with great interest since I did the exact same thing (in F90) for one of our code.

In my code, I bumped into the problem that with very large simulations, my seed could wrapup through 0. And if seed == 0 then my rng and lcg61 will stay at 0 forever. And I wasn't exiting gasdev :-(

It means also that the actual interval for lcg61 is ]0., 1.].

A guard against that situation in mkSeed would be useful. And I wonder if a protection against zero would not be a good idea too in lcg61.

As an improvement I would suggest to move lcg61 in the header as "inline" so that it can be vectorized easily if used outside gasdev (at least on SNB, something's wrong on KNC).

Measure performance with problem-specific metrics.

Performance as measured by execution time in seconds is difficult to interpret between vendors or implementations. Please report performance in some more problem-specific metric such as atoms per second.

GitHub.io link broken

The link to the documentation on github.io does not resolve.

Nor does the main website

src-mpi/yamlOutput.c no error checking or read-only functionality

When attempting to run mpirun -np 1 bin/CoMD-mpi in a read-only (container) environment, the program segfaults.

After debugging in GDB I was able to trace the segfault to yamlOutput.c where I read your comment and learned that this behavior is intended.

It is rather disappointing to learn that I will not be able to run CoMD in its current state. If I get the time during the school year this fall, I will fork it and get creative. In the meantime, I'd like to suggest adding some form of error checking to the file I/O in yamlOutput.c, specifically at line 60:

yamlFile = fopen(filename, "w");

This will save others (with read-only container environments) some time when debugging a seemingly random segfault from mpirun.