diff --git a/doc/sphinx/01_branson/branson.rst b/doc/sphinx/01_branson/branson.rst index e3d93942..faa5ee16 100644 --- a/doc/sphinx/01_branson/branson.rst +++ b/doc/sphinx/01_branson/branson.rst @@ -294,25 +294,6 @@ Strong scaling performance of Branson Crossroads 200M Particles is provided with Branson Strong Scaling Performance on Crossroads 200M particles -Multi-node scaling ------------------- - -The results of the scaling runs performed on rocinante hbm partition nodes are presented below. -Branson was built with intel oneapi 2023.1.0 and cray-mpich 8.1.25. -These runs used 32, 64, and 96 nodes with 110 tasks per node. -These runs use 85 million photons per node for a problem size using 25% of the total avalable memory across nodes. - -.. figure:: branson_roci_scale.png - :align: center - :scale: 50% - :alt: - -.. csv-table:: Multi Node Scaling Branson - :file: branson_roci_scale.csv - :align: center - :widths: 10, 10, 10, 10 - :header-rows: 1 - AMD Epyc + Nvidia A100 ---------------------- Throughput performance of Branson on AMD Epyc + Nvidia A100 (using a single GPU) is provided within the @@ -331,6 +312,25 @@ following table and figure. Branson Throughput Performance on AMD Epyc + Nvidia A100 +Multi-node scaling on Crossroads +================================ + +The results of the scaling runs performed on rocinante hbm partition nodes are presented below. +Branson was built with intel oneapi 2023.1.0 and cray-mpich 8.1.25. +These runs used 32, 64, and 96 nodes with 110 tasks per node. +These runs use 85 million photons per node for a problem size using 25% of the total avalable memory across nodes. + +.. figure:: branson_roci_scale.png + :align: center + :scale: 50% + :alt: + +.. csv-table:: Multi Node Scaling Branson + :file: branson_roci_scale.csv + :align: center + :widths: 10, 10, 10, 10, 10 + :header-rows: 1 + References ========== diff --git a/doc/sphinx/01_branson/branson_roci_badnodes_scale.csv b/doc/sphinx/01_branson/branson_roci_badnodes_scale.csv new file mode 100644 index 00000000..39af9275 --- /dev/null +++ b/doc/sphinx/01_branson/branson_roci_badnodes_scale.csv @@ -0,0 +1,4 @@ +Iteration,Photons,Nodes,Photons/s,Photons/s +1,2720,32,9.19E+07,2.87E+06 +1,5440,64,1.92E+07,3.00E+05 +1,8160,96,2.80E+07,2.91E+05 \ No newline at end of file diff --git a/doc/sphinx/01_branson/branson_roci_scale.csv b/doc/sphinx/01_branson/branson_roci_scale.csv index 08023132..a60ab714 100644 --- a/doc/sphinx/01_branson/branson_roci_scale.csv +++ b/doc/sphinx/01_branson/branson_roci_scale.csv @@ -1,8 +1,4 @@ -Photons,32,64,96 -200,1.25e+08,2.42e+08,2.41e+08 -400,1.26e+08,2.39e+08,3.42e+08 -800,1.25e+08,2.39e+08,3.53e+08 -1000,1.26e+08,2.37e+08,3.6e+08 -2000,1.27e+08,2.46e+08,3.58e+08 -4000,1.27e+08,2.52e+08,3.67e+08 -8000,,2.54e+08,3.77e+08 +Iteration,Photons,Nodes,Photons/s,Photons/s +1,2720,32,9.20E+07,2.87E+06 +1,5440,64,1.89E+08,2.95E+06 +1,8160,96,2.73E+08,2.85E+06 \ No newline at end of file diff --git a/doc/sphinx/01_branson/branson_roci_scale_photonrange.csv b/doc/sphinx/01_branson/branson_roci_scale_photonrange.csv new file mode 100644 index 00000000..08023132 --- /dev/null +++ b/doc/sphinx/01_branson/branson_roci_scale_photonrange.csv @@ -0,0 +1,8 @@ +Photons,32,64,96 +200,1.25e+08,2.42e+08,2.41e+08 +400,1.26e+08,2.39e+08,3.42e+08 +800,1.25e+08,2.39e+08,3.53e+08 +1000,1.26e+08,2.37e+08,3.6e+08 +2000,1.27e+08,2.46e+08,3.58e+08 +4000,1.27e+08,2.52e+08,3.67e+08 +8000,,2.54e+08,3.77e+08 diff --git a/doc/sphinx/01_branson/cpu.gp b/doc/sphinx/01_branson/cpu.gp index 2d669d86..8b000a97 100644 --- a/doc/sphinx/01_branson/cpu.gp +++ b/doc/sphinx/01_branson/cpu.gp @@ -33,9 +33,23 @@ set output "cpu_200M.png" plot "cpu_200M.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2 # Scaling Output -set output "branson_roci_scale.png" +set output "branson_roci_scale_range.png" set xrange [200:8000] unset logscale xy set key title "Number of Nodes" -plot "branson_roci_scale.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2, "" using 1:4 with line linestyle 3 +plot "branson_roci_scale_photonrange.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2, "" using 1:4 with line linestyle 3 + +# SCALING PLOTS, Y IS FOM PER NODE +set xrange [32:96] +set yrange [2.5e6:3.5e6] +set xlabel "Number of Nodes" +set format y "%.1e" +set ylabel "FOM/node" +unset logscale xy +set output "branson_roci_scale.png" +set title "Branson Multi Node Scaling" font "serif,22" +plot "branson_roci_scale.csv" using 3:5 with linespoints linestyle 1 +set output "branson_roci_scale_badnodes.png" +set title "Branson Multi Node Scaling" font "serif,22" +plot "branson_roci_badnodes_scale.csv" using 3:5 with linespoints linestyle 1 diff --git a/doc/sphinx/02_amg/amg.rst b/doc/sphinx/02_amg/amg.rst index 15dce633..f8b06794 100644 --- a/doc/sphinx/02_amg/amg.rst +++ b/doc/sphinx/02_amg/amg.rst @@ -346,28 +346,6 @@ Approximate results of the FOM for varying memory usages on Crossroads are provi Varying memory usage (estimated) for Problem 1 and 2 - -Multi-node scaling on Crossroads -================================ - -The results of the scaling runs performed on rocinante hbm partition are presented below. -Amg and hypre were built with intel oneapi 2023.1.0 and cray-mpich 8.1.25. -These runs used 32, 64, and 96 nodes with 108 tasks per node. -Problems 1 and 2 were run with problem sizes per MPI process, `-n`, of 25,25,125 and 40,40,200 respectively to use 15% of available memory. -The product of the x,y,z process topology must equal the number of processors. -In this case, x=y=24 for all node counts and z was set to 6, 12, and 18 for 32, 64, and 96 nodes respectively. - -.. figure:: cpu_scale_roci.png - :align: center - :scale: 50% - :alt: - -.. csv-table:: Multi Node Scaling AMG problem 1 and 2 - :file: amg_scale_roci.csv - :align: center - :widths: 10, 10, 10 - :header-rows: 1 - V-100 ===== @@ -413,6 +391,26 @@ The FOMs of AMG2023 on V100 for Problem 2 is provided in the following table and AMG2023 FOM on V100 for Problem 2 (7-pt stencil, AMG-PCG) +Multi-node scaling on Crossroads +================================ + +The results of the scaling runs performed on rocinante hbm partition are presented below. +Amg and hypre were built with intel oneapi 2023.1.0 and cray-mpich 8.1.25. +These runs used 32, 64, and 96 nodes with 108 tasks per node. +Problems 1 and 2 were run with problem sizes per MPI process, `-n`, of 25,25,125 and 40,40,200 respectively to use 15% of available memory. +The product of the x,y,z process topology must equal the number of processors. +In this case, x=y=24 for all node counts and z was set to 6, 12, and 18 for 32, 64, and 96 nodes respectively. + +.. figure:: cpu_scale_roci.png + :align: center + :scale: 50% + :alt: + +.. csv-table:: Multi Node Scaling AMG problem 1 and 2 + :file: amg_scale_roci.csv + :align: center + :widths: 10, 10, 10, 10, 10 + :header-rows: 1 References ========== diff --git a/doc/sphinx/02_amg/amg_scale_roci.csv b/doc/sphinx/02_amg/amg_scale_roci.csv index c1066b55..8e5585be 100644 --- a/doc/sphinx/02_amg/amg_scale_roci.csv +++ b/doc/sphinx/02_amg/amg_scale_roci.csv @@ -1,4 +1,4 @@ -NumNodes,Problem1,Problem2 -96,7.3e+09,3.01e+09 -64,5.44e+09,2.09e+09 -32,3.29e+09,1.19e+09 +NumNodes,Problem1,Problem2,Problem1,Problem2 +96,7.30E+09,3.01E+09,7.60E+07,3.14E+07 +64,5.44E+09,2.09E+09,8.50E+07,3.27E+07 +32,3.29E+09,1.19E+09,1.03E+08,3.72E+07 \ No newline at end of file diff --git a/doc/sphinx/02_amg/cpu.gp b/doc/sphinx/02_amg/cpu.gp index 81edf9f7..97e6c4d4 100644 --- a/doc/sphinx/02_amg/cpu.gp +++ b/doc/sphinx/02_amg/cpu.gp @@ -44,11 +44,13 @@ set output "roci_2_320.png" set title "AMG2023 Strong Scaling for Problem 2, 320 x 320 x 320" font "serif,22" plot "roci_2_320.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2 +# SCALING PLOTS, Y IS FOM PER NODE set xrange [32:96] set xlabel "Number of Nodes" set format y "%.1e" +set ylabel "FOM/node" unset logscale xy set output "cpu_scale_roci.png" set title "AMG Multi Node Scaling" font "serif,22" -plot "amg_scale_roci.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2 +plot "amg_scale_roci.csv" using 1:4 with linespoints linestyle 1, "" using 1:5 with line linestyle 2