Skip to content

Unexpected change in performance and correctness check output since tag v5.5.0 #29

@gshipman

Description

@gshipman

Hi folks,

Using v.5.5.0 I get output from:

mpirun -n 112 --map-by numa --bind-to core ./test_driver -c 20 -B global -d 14,14,14 --sweep_kernel=0 --benchmark_problem=1 -t 1
=================================================================
Test driver finished time steps
=================================================================
Average throughput of single iteration of iterative solver was 6.30546e+09 unknowns calculated per second.
Throughput of iterative solver was 5.81149e+08 unknowns calculated per second.
(average throughput of single iteration * # iterations for solver to produce answer

Total number of flux solver iterations for run: 217
Total wall time for run: 167.098 seconds.
=================================================================

RESULT CHECK PASSED: Energy check (this is relative to total energy) 1.04324e-15 within tolerance of +/- 1e-09; check './unnamed.csv' for tally details

Compared to latest head I get output from:

mpirun -n 112 --map-by numa --bind-to core ./test_driver -c 20 -B global -d 14,14,14 --sweep_kernel=0 --benchmark_problem=1 -t 1
=================================================================
Test driver finished time steps
Simulation time progressed: 3.503887e-03
Number of cycles: 20
Total wall time for run: 7.056378e+01 seconds.
=================================================================

=================================================================
Test driver throughput statistics

Definition of flux solver throughput:
  # unknowns calculated * simulation time progressed / wall time
Definition of single flux iteration throughput:
  # flux solver throughput / number of flux iterations taken

=================================================================
Problem throughput: 2.410993e+05
=================================================================

Is this expected change in behavior?
Is there an easy way of comparing these two runs in terms of performance?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions