Add a simple benchmark

benchmark/benchmark.jl

@@ -0,0 +1,25 @@
+using TrixiParticles
+using BenchmarkTools
+
+# Suppress the output of the simulation
+function suppress_simulation_output(f)
+    open("/dev/null", "w") do devnull
+        redirect_stdout(devnull) do
+            redirect_stderr(devnull) do
+                f()
+            end
+        end
+    end
+end
+
+# Benchmark using @btime and suppress output
+result = @belapsed suppress_simulation_output(() -> trixi_include(@__MODULE__,
+                                                                  joinpath(examples_dir(),
+                                                                           "fluid",
+                                                                           "dam_break_2d.jl"),
+                                                                  saving_callback=nothing,
+                                                                  saving_paper=nothing,
+                                                                  fluid_particle_spacing=0.01))
+
+# Print the timing result in nanoseconds
+println(result)

benchmark/plot_benchmark.jl

@@ -0,0 +1,33 @@
+using CSV
+using DataFrames
+using CairoMakie
+using FileIO
+
+csv_file = "benchmark_results.csv"
+
+if !isfile(csv_file)
+    println("Error: File '$csv_file' does not exist. Please make sure the file is available and try again.")
+    return
+end
+
+df = CSV.read(csv_file, DataFrame, header=1, skipto=3)
+
+longest_time = maximum(df.Time)
+df.Speedup = longest_time ./ df.Time
+
+x_ticks = df.Threads
+
+fig = Figure(size=(800, 600))
+ax = Axis(fig[1, 1],
+          title="Strong Scaling Speedup",
+          xlabel="Number of Threads",
+          ylabel="Speedup",
+          xticks=(x_ticks, string.(x_ticks)))
+
+line_plot = lines!(ax, df.Threads, df.Speedup, color=:blue, label="Speedup Curve")
+scatter_plot = scatter!(ax, df.Threads, df.Speedup, color=:red, markersize=10,
+                        label="Data Points")
+
+axislegend(ax, position=:rb)
+
+save("benchmark_speedup.svg", fig)

benchmark/run_strong_scaling.sh

@@ -0,0 +1,64 @@
+#!/bin/bash
+
+######################################################################################
+# Benchmark Script
+#
+# This script automates the benchmarking process for TrixiParticles by running the 
+# benchmark script (`benchmark/benchmark.jl`) with varying thread counts. It starts 
+# from a minimum number of threads (default: 1) and doubles the number of threads 
+# each iteration until the maximum number of threads (based on CPU or overridden) 
+# is reached.
+#
+# Output:
+# - The results are saved in a CSV file called "benchmark_results.csv", with columns:
+#     - Threads: Number of threads used during the benchmark run.
+#     - Time: The time taken to complete the benchmark (in seconds).
+# - Units are saved in the second row, indicating the time in seconds.
+#
+# Environment Variables:
+# - OVERRIDE_MAX_THREADS: Optionally, the maximum number of threads can be overridden
+#   by setting this environment variable. If not set, the system's CPU thread count 
+#   will be used.
+# - OVERRIDE_MIN_THREADS: Optionally, the minimum number of threads can be overridden
+#   by setting this environment variable. If not set, the default is 1.
+#
+# Example Usage:
+# - Run with default thread settings:
+#     ./benchmark_script.sh
+# - Override maximum threads:
+#     OVERRIDE_MAX_THREADS=8 ./benchmark_script.sh
+# - Override both minimum and maximum threads:
+#     OVERRIDE_MIN_THREADS=2 OVERRIDE_MAX_THREADS=8 ./benchmark_script.sh
+######################################################################################
+
+# Get the total number of CPU threads available on the system
+DEFAULT_MAX_THREADS=$(julia -e 'println(Sys.CPU_THREADS)')
+DEFAULT_MIN_THREADS=1
+
+MAX_THREADS=${OVERRIDE_MAX_THREADS:-$DEFAULT_MAX_THREADS}
+MIN_THREADS=${OVERRIDE_MIN_THREADS:-$DEFAULT_MIN_THREADS}
+
+# Initialize the output file with headers and units
+OUTPUT_FILE="benchmark_results.csv"
+echo "Threads,Time" > $OUTPUT_FILE
+echo "[1],[s]" >> $OUTPUT_FILE 
+
+CURRENT_THREADS=$MIN_THREADS
+
+while [ $CURRENT_THREADS -le $MAX_THREADS ]; do
+    echo "Running benchmark with $CURRENT_THREADS threads..."
+
+    # Run the benchmark using the specified number of threads and capture the timing result
+    BENCHMARK_TIME=$(julia --project=. --threads=$CURRENT_THREADS benchmark/benchmark.jl)
+
+    # Append the number of threads and timing result to the CSV file
+    echo "$CURRENT_THREADS,$BENCHMARK_TIME" >> $OUTPUT_FILE
+
+    CURRENT_THREADS=$((CURRENT_THREADS * 2))
+done
+
+echo "Benchmarking complete. Results saved to $OUTPUT_FILE."
+
+julia plot_benchmark.jl
+
+echo "Plotting complete. Plot saved as benchmark_speedup.svg."

benchmark_results.csv

@@ -0,0 +1,7 @@
+Threads,Time_ns
+4,44.130856019
+8,26.929279594
+16,17.867562992
+32,12.621583571
+64,11.243286411
+128,22.692239428