Skip to content

Commit

Permalink
Amg roci results with plots and table.
Browse files Browse the repository at this point in the history
  • Loading branch information
dmageeLANL committed Feb 6, 2024
1 parent bace28a commit 96c1ca0
Show file tree
Hide file tree
Showing 4 changed files with 36 additions and 3 deletions.
2 changes: 1 addition & 1 deletion doc/sphinx/00_intro/introduction.rst
Original file line number Diff line number Diff line change
Expand Up @@ -170,7 +170,7 @@ Microbenchmark Overview
.. _GlobalRunRules:

Run Rules Synopsis
===============
==================

Single node benchmarks will require respondent to provide estimates on

Expand Down
24 changes: 22 additions & 2 deletions doc/sphinx/02_amg/amg.rst
Original file line number Diff line number Diff line change
Expand Up @@ -225,7 +225,7 @@ The second figure provides memory use on 1 node of CTS-1 (Quartz) using 4 MPI ta


Strong Scaling on Crossroads
----------------------------
============================

We present strong scaling results for varying problem sizes on Crossroads with HBM below. The code was configured and compiled using hypre v2.29.0 with MPI only and optimization -O2.

Expand Down Expand Up @@ -347,9 +347,29 @@ 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 nodesare presented below.
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 of 25,25,50 and 36,36,72 respectively.
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 problem 1 and 2
:file: amg_scale_roci.csv
:align: center
:widths: 10, 10, 10
:header-rows: 1

V-100
-----
=====

We have also performed runs on 1 NVIDIA V-100 GPU increasing the problem size n x n x n.
For these runs hypre 2.29.0 was configured as follows:
Expand Down
4 changes: 4 additions & 0 deletions doc/sphinx/02_amg/amg_scale_roci.csv
Original file line number Diff line number Diff line change
@@ -0,0 +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
9 changes: 9 additions & 0 deletions doc/sphinx/02_amg/cpu.gp
Original file line number Diff line number Diff line change
Expand Up @@ -20,6 +20,7 @@ set key autotitle columnheader

set style line 1 linetype 6 dashtype 1 linecolor rgb "#FF0000" linewidth 2 pointtype 6 pointsize 3
set style line 2 linetype 1 dashtype 2 linecolor rgb "#FF0000" linewidth 2
set style line 3 linetype 6 dashtype 1 linecolor rgb "#0000FF" linewidth 2 pointtype 6 pointsize 3

plot "roci_1_120.csv" using 1:2 with linespoints linestyle 1, "" using 1:3 with line linestyle 2

Expand All @@ -43,3 +44,11 @@ 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

set xrange [32:96]
set xlabel "Number of Nodes"
set format y "%.1e"
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

0 comments on commit 96c1ca0

Please sign in to comment.