aboutsummaryrefslogtreecommitdiff
path: root/Carpet/CarpetWeb/scaling-improved/results.gnuplot
diff options
context:
space:
mode:
Diffstat (limited to 'Carpet/CarpetWeb/scaling-improved/results.gnuplot')
-rw-r--r--Carpet/CarpetWeb/scaling-improved/results.gnuplot48
1 files changed, 48 insertions, 0 deletions
diff --git a/Carpet/CarpetWeb/scaling-improved/results.gnuplot b/Carpet/CarpetWeb/scaling-improved/results.gnuplot
new file mode 100644
index 000000000..83062f7bf
--- /dev/null
+++ b/Carpet/CarpetWeb/scaling-improved/results.gnuplot
@@ -0,0 +1,48 @@
+set terminal postscript eps enhanced color colortext
+set output "results.eps"
+
+#set size 0.5
+
+set logscale x
+set xtics (1, 4, 16, 64, 256, "1k" 1024, "4k" 4096, "16k" 16384)
+set xrange [0.7:24576]
+set yrange [0:]
+
+set title "Cactus Benchmark"
+set xlabel "number of cores"
+set ylabel "time per RHS evaluation [{/Symbol m}s]"
+
+set key bottom
+
+# 25^3 grid points per core
+# 128 time steps
+# 9 refinement levels
+# a factor 2 because of the Berger-Oliger subcycling
+# 4 Runge-Kutta substeps per time step
+# 1e6 us per second
+p \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && $7== 1 && $8==10) print; }' results-hlrb2.out" u 5:($9/25**3/128/2/4*1e6) t "HLRB II" w lp lt 1 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6== 4 || $6==$5) && $7== 1 && $8==50) print; }' results-kraken.out" u 5:($9/25**3/128/2/4*1e6) t "Kraken (NT=1)" w lp lt 2 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6== 8 || $6==$5) && $7== 1 && $8== 0) print; }' results-queenbee.out" u 5:($9/25**3/128/2/4*1e6) t "Queen Bee (NT=1)" w lp lt 3 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6== 8 || $6==$5) && $7== 2 && $8== 0) print; }' results-queenbee.out" u 5:($9/25**3/128/2/4*1e6) t "Queen Bee (NT=2)" w lp lt 3 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6== 8 || $6==$5) && $7== 4 && $8== 0) print; }' results-queenbee.out" u 5:($9/25**3/128/2/4*1e6) t "Queen Bee (NT=4)" w lp lt 3 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6== 8 || $6==$5) && $7== 8 && $8== 0) print; }' results-queenbee.out" u 5:($9/25**3/128/2/4*1e6) t "Queen Bee (NT=8)" w lp lt 3 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6==16 || $6==$5) && $7== 1 && $8==71) print; }' results-ranger.out" u 5:($9/25**3/128/2/4*1e6) t "Ranger (NT=1)" w lp lt 4 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6==16 || $6==$5) && $7== 2 && $8==71) print; }' results-ranger.out" u 5:($9/25**3/128/2/4*1e6) t "Ranger (NT=2)" w lp lt 4 lw 3, \
+"< awk '/^Bench_McLachlan_Carpet_9lev/ { if ($2==128 && $3==25 && ($6==16 || $6==$5) && $7== 4 && $8==71) print; }' results-ranger.out" u 5:($9/25**3/128/2/4*1e6) t "Ranger (NT=4)" w lp lt 4 lw 3
+
+# HLRB 2: have results up to 2048 cores; runs with 4080 cores failed,
+# reason unknown. Job with updated executable and fewer output files
+# submitted; still fails for unknown reason. Could be out of memory.
+
+# Intrepid: no allocation.
+
+# Kraken: have results up to 8192 cores; runs with up 18048 cores did
+# fail, reason unknown. (out of memory? did work with NT=2 and
+# NT=4.)
+
+# Queen Bee: have results up to 1024 cores; runs with up to 2048 cores
+# are submitted.
+
+# Ranger: have results up to 12288 cores; runs with more cores depend
+# on permissions from system administrators.