Skip to content

parser.py doesn't work without Nvidia / CUDA #3

Description

@JuKu

Hi together,

I do a research project about MPI / OpenMPI 5.0.8 (with Peruse) and we try to trace things with ucTrace.
I used a slurm system without Nvidia / CUDA with 8 Nodes and ran LULESH 2.0 LULESH 2.0 with it:

#!/bin/bash
#SBATCH --nodes=1
#SBATCH --tasks-per-node=8
#SBATCH --cpus-per-task=1
#SBATCH --mem-per-cpu=2G
#SBATCH --output=lulesh-%j.out
#SBATCH --error=lulesh-%j.err
export LD_PRELOAD=/home/<USERNAME>/git/so-files-lib/preload.so
# This folder contains the built .so files from ucTrace folder
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/<USERNAME>/git/so-files-lib/
srun /home/<USERNAME>/git/LULESH/lulesh2.0

Everything worked and I got the 8 .asd.gz files.

After that, I tried to execute parse.py to parse the log files:

python3 parse.py my-outputs/ -p my_output.pkl -n 0

Badly, I got an error while the script was parsing the COMMS:

Traceback (most recent call last):
  File "/home/<USERNAME>/ucTrace/parser/parse.py", line 2017, in <module>
    main()
    ~~~~^^
  File "/home/<USERNAME>/ucTrace/parser/parse.py", line 1966, in main
    all_data, parsed_comms = parse_folder(args.folder_path, args)
                             ~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/<USERNAME>/ucTrace/parser/parse.py", line 1793, in parse_folder
    match_ucp_send_recv(parsed_comms, all_data)
    ~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/<USERNAME>/ucTrace/parser/parse.py", line 1253, in match_ucp_send_recv
    recvs = tmp_recvs[target_pid]
            ~~~~~~~~~^^^^^^^^^^^^
KeyError: "Error decoding hex string-- 'ascii' codec can't decode byte 0xf0 in position 8"

How can I fix this?
Thanks in advance!

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions