t8_time_set_join_by_vertices.cxx

t8_time_set_join_by_vertices

Overview

This benchmark specifically measures the performance of joining mesh sets by their vertices, a common operation in mesh processing tasks.

---

Sample Usage

mpirun -np 4 ./t8_time_set_join_by_vertices -f mymesh

Replace mymesh with the prefix of your mesh file.

---

Sample Output

Statistics for t8_cmesh_from_msh_file
...

(Output will display detailed timing statistics similar to the example below.)

---

Interpretation

Statistics for t8_cmesh_from_msh_file

This block shows how long it took on each MPI process (rank) to read and parse the mesh file.

• Global number of values:
  4 means you ran the program with 4 MPI processes (ranks 0, 1, 2, and 3).

• Mean value (std. dev.):
  Average time (in seconds) each process spent reading the mesh file, with standard deviation as a percentage.
  Example:

  0.00453795 (0.000308 = 6.79%)
  

  Average time is 0.00454 seconds, with standard deviation 0.00031.

• Minimum attained at rank:
  Fastest time was on rank 2: 0.00400532 seconds.

• Maximum attained at rank:
  Slowest time was on rank 3: 0.00473786 seconds.

• Summary:
  List of average time(s).

• Maximum:
  Maximum time measured (0.00473786 seconds).

• ntrees = 409:
  The mesh consists of 409 trees (subdomains or mesh blocks).

---

Statistics for t8_cmesh_set_join_by_vertices

This block shows the timings for building the face connectivity by joining mesh sets via their vertices.

• Global number of values:
  Again, 4 MPI ranks.

• Mean value (std. dev.):
  Average time spent on this operation, with standard deviation.
  Example:

  0.00116814 (0.000124 = 10.6%)
  

  Average time is 0.00117 seconds.

• Minimum attained at rank:
  Fastest rank was 2: 0.000953369 seconds.

• Maximum attained at rank:
  Slowest was rank 3: 0.00124739 seconds.

• Summary / Maximum:
  Same as before, summary and max timing values.

---

Summary Interpretation

• The mesh reading step (t8_cmesh_from_msh_file) took, on average, ~0.00454 seconds.
• The join-by-vertices step (t8_cmesh_set_join_by_vertices) was faster, ~0.00117 seconds.
• There is some variability (see std. dev.), but it is not extreme.
• Times are reported per rank; the slowest rank determines parallel efficiency.
• 409 trees indicates the size or decomposition of your mesh.

---

Bottom Line

• These numbers help you benchmark your mesh processing speed.
• Running with larger meshes or more MPI ranks will change times and scaling behavior.
• You can use these stats to identify bottlenecks or compare with future optimizations.