[petsc-dev] Kokkos/Crusher perforance
Mark Adams
mfadams at lbl.gov
Tue Jan 25 12:49:48 CST 2022
>
>
> Note that Mark's logs have been switching back and forth between
> -use_gpu_aware_mpi and changing number of ranks -- we won't have that
> information if we do manual timing hacks. This is going to be a routine
> thing we'll need on the mailing list and we need the provenance to go with
> it.
>
GPU aware MPI crashes sometimes so to be safe, while debugging, I had it
off. It works fine here so it has been on in the last tests.
Here is a comparison.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20220125/9127260e/attachment.html>
-------------- next part --------------
Script started on 2022-01-25 13:44:31-05:00 [TERM="xterm-256color" TTY="/dev/pts/0" COLUMNS="296" LINES="100"]
13:44[;1m[35m[;0m [34madams/aijkokkos-gpu-logging *= [33mcrusher:[32m/gpfs/alpine/csc314/scratch/adams/petsc/src/snes/tests/data[0m$ exitbash -x run_crusher_jac.sbatchexit[Kbash -x run_crusher_jac.sbatch
+ '[' -z '' ']'
+ case "$-" in
+ __lmod_vx=x
+ '[' -n x ']'
+ set +x
Shell debugging temporarily silenced: export LMOD_SH_DBG_ON=1 for this output (/usr/share/lmod/lmod/init/bash)
Shell debugging restarted
+ unset __lmod_vx
+ NG=8
+ NC=1
+ date
Tue 25 Jan 2022 01:44:38 PM EST
+ EXTRA='-dm_view -log_viewx -ksp_view -use_gpu_aware_mpi true'
+ HYPRE_EXTRA='-pc_hypre_boomeramg_relax_type_all l1scaled-Jacobi -pc_hypre_boomeramg_interp_type ext+i -pc_hypre_boomeramg_coarsen_type PMIS -pc_hypre_boomeramg_no_CF'
+ HYPRE_EXTRA='-pc_hypre_boomeramg_no_CF true -pc_hypre_boomeramg_strong_threshold 0.75 -pc_hypre_boomeramg_agg_nl 1 -pc_hypre_boomeramg_coarsen_type HMIS -pc_hypre_boomeramg_interp_type ext+i '
+ for REFINE in 5
+ for NPIDX in 1
+ let 'N1 = 1 * 1'
++ bc -l
+ PG=2.00000000000000000000
++ printf %.0f 2.00000000000000000000
+ PG=2
+ let 'NCC = 8 / 1'
+ let 'N4 = 2 * 1'
+ let 'NODES = 1 * 1 * 1'
+ let 'N = 1 * 1 * 8'
+ echo n= 8 ' NODES=' 1 ' NC=' 1 ' PG=' 2
n= 8 NODES= 1 NC= 1 PG= 2
++ printf %03d 1
+ foo=001
+ srun -n8 -N1 --ntasks-per-gpu=1 --gpu-bind=closest -c 8 ../ex13 -dm_plex_box_faces 2,2,2 -petscpartitioner_simple_process_grid 2,2,2 -dm_plex_box_upper 1,1,1 -petscpartitioner_simple_node_grid 1,1,1 -dm_refine 5 -dm_view -log_viewx -ksp_view -use_gpu_aware_mpi true -dm_mat_type aijkokkos -dm_vec_type kokkos -pc_type jacobi
+ tee jac_out_001_kokkos_Crusher_5_1_noview.txt
DM Object: box 8 MPI processes
type: plex
box in 3 dimensions:
Number of 0-cells per rank: 35937 35937 35937 35937 35937 35937 35937 35937
Number of 1-cells per rank: 104544 104544 104544 104544 104544 104544 104544 104544
Number of 2-cells per rank: 101376 101376 101376 101376 101376 101376 101376 101376
Number of 3-cells per rank: 32768 32768 32768 32768 32768 32768 32768 32768
Labels:
celltype: 4 strata with value/size (0 (35937), 1 (104544), 4 (101376), 7 (32768))
depth: 4 strata with value/size (0 (35937), 1 (104544), 2 (101376), 3 (32768))
marker: 1 strata with value/size (1 (12474))
Face Sets: 3 strata with value/size (1 (3969), 3 (3969), 6 (3969))
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Solve time: 0.34211
#PETSc Option Table entries:
-benchmark_it 2
-dm_distribute
-dm_mat_type aijkokkos
-dm_plex_box_faces 2,2,2
-dm_plex_box_lower 0,0,0
-dm_plex_box_upper 1,1,1
-dm_plex_dim 3
-dm_plex_simplex 0
-dm_refine 5
-dm_vec_type kokkos
-dm_view
-ksp_converged_reason
-ksp_max_it 200
-ksp_norm_type unpreconditioned
-ksp_rtol 1.e-12
-ksp_type cg
-ksp_view
-log_viewx
-mg_levels_esteig_ksp_max_it 10
-mg_levels_esteig_ksp_type cg
-mg_levels_ksp_chebyshev_esteig 0,0.05,0,1.05
-mg_levels_ksp_type chebyshev
-mg_levels_pc_type jacobi
-options_left
-pc_gamg_coarse_eq_limit 100
-pc_gamg_coarse_grid_layout_type compact
-pc_gamg_esteig_ksp_max_it 10
-pc_gamg_esteig_ksp_type cg
-pc_gamg_process_eq_limit 400
-pc_gamg_repartition false
-pc_gamg_reuse_interpolation true
-pc_gamg_square_graph 0
-pc_gamg_threshold -0.01
-pc_type jacobi
-petscpartitioner_simple_node_grid 1,1,1
-petscpartitioner_simple_process_grid 2,2,2
-petscpartitioner_type simple
-potential_petscspace_degree 2
-snes_max_it 1
-snes_rtol 1.e-8
-snes_type ksponly
-use_gpu_aware_mpi true
#End of PETSc Option Table entries
WARNING! There are options you set that were not used!
WARNING! could be spelling mistake, etc!
There are 15 unused database options. They are:
Option left: name:-log_viewx (no value)
Option left: name:-mg_levels_esteig_ksp_max_it value: 10
Option left: name:-mg_levels_esteig_ksp_type value: cg
Option left: name:-mg_levels_ksp_chebyshev_esteig value: 0,0.05,0,1.05
Option left: name:-mg_levels_ksp_type value: chebyshev
Option left: name:-mg_levels_pc_type value: jacobi
Option left: name:-pc_gamg_coarse_eq_limit value: 100
Option left: name:-pc_gamg_coarse_grid_layout_type value: compact
Option left: name:-pc_gamg_esteig_ksp_max_it value: 10
Option left: name:-pc_gamg_esteig_ksp_type value: cg
Option left: name:-pc_gamg_process_eq_limit value: 400
Option left: name:-pc_gamg_repartition value: false
Option left: name:-pc_gamg_reuse_interpolation value: true
Option left: name:-pc_gamg_square_graph value: 0
Option left: name:-pc_gamg_threshold value: -0.01
+ srun -n8 -N1 --ntasks-per-gpu=1 --gpu-bind=closest -c 8 ../ex13 -dm_plex_box_faces 2,2,2 -petscpartitioner_simple_process_grid 2,2,2 -dm_plex_box_upper 1,1,1 -petscpartitioner_simple_node_grid 1,1,1 -dm_refine 5 -dm_view -log_viewx -ksp_view -use_gpu_aware_mpi true -use_gpu_aware_mpi false -dm_mat_type aijkokkos -dm_vec_type kokkos -pc_type jacobi
+ tee jac_out_001_kokkos_Crusher_5_1.txt
DM Object: box 8 MPI processes
type: plex
box in 3 dimensions:
Number of 0-cells per rank: 35937 35937 35937 35937 35937 35937 35937 35937
Number of 1-cells per rank: 104544 104544 104544 104544 104544 104544 104544 104544
Number of 2-cells per rank: 101376 101376 101376 101376 101376 101376 101376 101376
Number of 3-cells per rank: 32768 32768 32768 32768 32768 32768 32768 32768
Labels:
celltype: 4 strata with value/size (0 (35937), 1 (104544), 4 (101376), 7 (32768))
depth: 4 strata with value/size (0 (35937), 1 (104544), 2 (101376), 3 (32768))
marker: 1 strata with value/size (1 (12474))
Face Sets: 3 strata with value/size (1 (3969), 3 (3969), 6 (3969))
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Linear solve did not converge due to DIVERGED_ITS iterations 200
KSP Object: 8 MPI processes
type: cg
maximum iterations=200, initial guess is zero
tolerances: relative=1e-12, absolute=1e-50, divergence=10000.
left preconditioning
using UNPRECONDITIONED norm type for convergence test
PC Object: 8 MPI processes
type: jacobi
type DIAGONAL
linear system matrix = precond matrix:
Mat Object: 8 MPI processes
type: mpiaijkokkos
rows=2048383, cols=2048383
total: nonzeros=127263527, allocated nonzeros=127263527
total number of mallocs used during MatSetValues calls=0
not using I-node (on process 0) routines
Solve time: 0.370544
#PETSc Option Table entries:
-benchmark_it 2
-dm_distribute
-dm_mat_type aijkokkos
-dm_plex_box_faces 2,2,2
-dm_plex_box_lower 0,0,0
-dm_plex_box_upper 1,1,1
-dm_plex_dim 3
-dm_plex_simplex 0
-dm_refine 5
-dm_vec_type kokkos
-dm_view
-ksp_converged_reason
-ksp_max_it 200
-ksp_norm_type unpreconditioned
-ksp_rtol 1.e-12
-ksp_type cg
-ksp_view
-log_viewx
-mg_levels_esteig_ksp_max_it 10
-mg_levels_esteig_ksp_type cg
-mg_levels_ksp_chebyshev_esteig 0,0.05,0,1.05
-mg_levels_ksp_type chebyshev
-mg_levels_pc_type jacobi
-options_left
-pc_gamg_coarse_eq_limit 100
-pc_gamg_coarse_grid_layout_type compact
-pc_gamg_esteig_ksp_max_it 10
-pc_gamg_esteig_ksp_type cg
-pc_gamg_process_eq_limit 400
-pc_gamg_repartition false
-pc_gamg_reuse_interpolation true
-pc_gamg_square_graph 0
-pc_gamg_threshold -0.01
-pc_type jacobi
-petscpartitioner_simple_node_grid 1,1,1
-petscpartitioner_simple_process_grid 2,2,2
-petscpartitioner_type simple
-potential_petscspace_degree 2
-snes_max_it 1
-snes_rtol 1.e-8
-snes_type ksponly
-use_gpu_aware_mpi false
#End of PETSc Option Table entries
WARNING! There are options you set that were not used!
WARNING! could be spelling mistake, etc!
There are 15 unused database options. They are:
Option left: name:-log_viewx (no value)
Option left: name:-mg_levels_esteig_ksp_max_it value: 10
Option left: name:-mg_levels_esteig_ksp_type value: cg
Option left: name:-mg_levels_ksp_chebyshev_esteig value: 0,0.05,0,1.05
Option left: name:-mg_levels_ksp_type value: chebyshev
Option left: name:-mg_levels_pc_type value: jacobi
Option left: name:-pc_gamg_coarse_eq_limit value: 100
Option left: name:-pc_gamg_coarse_grid_layout_type value: compact
Option left: name:-pc_gamg_esteig_ksp_max_it value: 10
Option left: name:-pc_gamg_esteig_ksp_type value: cg
Option left: name:-pc_gamg_process_eq_limit value: 400
Option left: name:-pc_gamg_repartition value: false
Option left: name:-pc_gamg_reuse_interpolation value: true
Option left: name:-pc_gamg_square_graph value: 0
Option left: name:-pc_gamg_threshold value: -0.01
+ date
Tue 25 Jan 2022 01:46:55 PM EST
13:46[;1m[35m[;0m [34madams/aijkokkos-gpu-logging *= [33mcrusher:[32m/gpfs/alpine/csc314/scratch/adams/petsc/src/snes/tests/data[0m$ exit
exit
Script done on 2022-01-25 13:47:22-05:00 [COMMAND_EXIT_CODE="0"]
More information about the petsc-dev
mailing list