Prepare a Linux desktop PC for CPU sensitive benchmarking.

benchmark-prepare.sh

#!/bin/bash

set -e

# https://llvm.org/docs/Benchmarking.html
# https://smalldatum.blogspot.com/2024/10/managing-cpu-frequency-for-amd-on.html
# sudo apt install linux-tools-common linux-tools-generic linux-cloud-tools-generic cpuset
# disable hyper threading
for i in $(seq 16 31); do
    echo 0 | sudo tee /sys/devices/system/cpu/cpu$i/online
done

sudo cpupower idle-set -D1

sudo cpupower frequency-set --governor performance
cpupower frequency-info -o proc
echo 0 | sudo tee /sys/devices/system/cpu/cpufreq/boost
echo 0 | sudo tee /proc/sys/kernel/randomize_va_space
# echo 1 | sudo tee /sys/devices/system/cpu/intel_pstate/no_turbo
#taskset -c 0 ./mybenchmark

benchmark-restore.sh

#!/bin/bash

set -e

# sudo apt install linux-tools-common linux-tools-generic linux-cloud-tools-generic cpuset
# disable hyper threading
for i in $(seq 16 31); do
    echo 1 | sudo tee /sys/devices/system/cpu/cpu$i/online
done

sudo cpupower idle-set -E

sudo cpupower frequency-set --governor powersave
cpupower frequency-info -o proc
echo 1 | sudo tee /sys/devices/system/cpu/cpufreq/boost
# echo 0 | sudo tee /proc/sys/kernel/randomize_va_space
# echo 0 | sudo tee /sys/devices/system/cpu/intel_pstate/no_turbo
#taskset -c 0 ./mybenchmark

cpu_benchmark.sh

#!/bin/bash

SCRIPTPATH="$( cd -- "$(dirname "$0")" >/dev/null 2>&1 ; pwd -P )"

# CPU性能稳定性测试脚本
# 用途：多次运行CPU基准测试并记录性能随时间的变化

# 配置参数
TEST_COUNT=20                 # 测试次数
TEST_DURATION=30             # 每次测试的持续时间(秒)
PRIME_NUMBERS=10000          # 计算的质数范围
RESULTS_FILE="cpu_bench_results.csv"  # 结果保存文件
SLEEP_BETWEEN_TESTS=0       # 测试间隔(秒)


$SCRIPTPATH/benchmark-prepare.sh

# 检查sysbench是否安装
if ! command -v sysbench &> /dev/null; then
    echo "错误: sysbench未安装，请先安装sysbench"
    echo "在Ubuntu/Debian上可以使用: sudo apt install sysbench"
    echo "在CentOS/RHEL上可以使用: sudo yum install sysbench"
    exit 1
fi

# 创建结果文件头
echo "timestamp,test_number,events_per_second,cpu_latency" > $RESULTS_FILE

echo "开始CPU性能稳定性测试..."
echo "将进行 $TEST_COUNT 次测试，每次测试 $TEST_DURATION 秒"
echo "结果将保存到: $RESULTS_FILE"

# echo "[*] Shielding: move system slices off CPU0..."
# systemctl set-property --runtime init.scope AllowedCPUs=1-31
# systemctl set-property --runtime system.slice AllowedCPUs=1-31
# systemctl set-property --runtime user.slice AllowedCPUs=1-31

# 运行多次测试
for ((i=1; i<=$TEST_COUNT; i++))
do
    # 获取当前CPU频率设置（可选）
    echo "当前CPU频率信息:"
    if [ -f /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor ]; then
        echo "CPU调速器: $(cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor)"
    fi
    if [ -f /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq ]; then
        echo "当前频率: $(($(cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq)/1000)) MHz"
    fi

    echo "=============================================="
    echo "运行测试 #$i - $(date)"
    
    # 运行sysbench CPU测试
    echo "运行CPU基准测试..."
    # result=$(systemd-run --collect --scope -p AllowedCPUs=0 --same-dir -- sysbench cpu --cpu-max-prime=$PRIME_NUMBERS --time=$TEST_DURATION run)
    result=$(sysbench cpu --cpu-max-prime=$PRIME_NUMBERS --time=$TEST_DURATION run)

    # 提取性能数据
    events_per_second=$(echo "$result" | grep "events per second:" | awk '{print $4}')
    cpu_latency=$(echo "$result" | grep "avg:" | awk '{print $2}')
    timestamp=$(date +"%Y-%m-%d %H:%M:%S")
    
    # 记录结果
    echo "$timestamp,$i,$events_per_second,$cpu_latency" >> $RESULTS_FILE
    echo "测试完成: $events_per_second 事件/秒, CPU Latency: $cpu_latency%"
    
    # 如果不是最后一次测试，则等待
    if [ $i -lt $TEST_COUNT ]; then
        echo "等待 $SLEEP_BETWEEN_TESTS 秒后进行下一次测试..."
        sleep $SLEEP_BETWEEN_TESTS
    fi
done

echo "=============================================="
echo "所有测试完成! 结果已保存到 $RESULTS_FILE"

# 简单结果分析
echo "性能结果摘要:"
awk -F',' 'NR>1 {print $3}' $RESULTS_FILE | awk '
BEGIN {
    min = 9999999
    max = 0
    sum = 0
    count = 0
}
{
    if ($1 < min) min = $1
    if ($1 > max) max = $1
    sum += $1
    count++
}
END {
    printf "测试次数: %d\n", count
    printf "平均性能: %.2f 事件/秒\n", sum/count
    printf "最佳性能: %.2f 事件/秒\n", max
    printf "最差性能: %.2f 事件/秒\n", min
    printf "性能波动: %.2f%%\n", (max-min)/max*100
}' $RESULTS_FILE

# echo "[*] Restoring AllowedCPUs..."
# systemctl revert init.scope
# systemctl revert system.slice
# systemctl revert user.slice
# systemctl set-property --runtime init.scope AllowedCPUs=0-32
# systemctl set-property --runtime system.slice AllowedCPUs=0-32
# systemctl set-property --runtime user.slice AllowedCPUs=0-32

$SCRIPTPATH/benchmark-restore.sh