testsAndMisc/linux_configuration/i3-configuration/i3blocks/gpu_monitor.sh

81 lines
2.2 KiB
Bash
Raw Normal View History

#!/bin/bash
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
# i3blocks GPU monitor, persist mode.
#
# Keeps a single long-lived `nvidia-smi --loop=5` (or reads amdgpu sysfs
# in a blocking-read loop) instead of forking nvidia-smi/lspci/awk/tr/bc
# every interval. No sleep, no polling loop in bash — nvidia-smi's own
# periodic emitter drives updates and we block on `read`.
#
# Configure with `interval=persist` and `markup=pango` in the i3blocks
# config. In persist mode each newline is a separate status update, so
# we emit exactly ONE line (with inline pango markup for color).
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
set -u
# Nerd Font glyph: display / desktop icon (U+F108).
ICON=$'\uf108'
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
emit() {
local temp=$1 load=$2 color
if [[ $load == 'N/A' ]]; then
color='#FFFFFF'
elif ((load < 50)); then
color='#50FA7B'
elif ((load < 75)); then
color='#F1FA8C'
else
color='#FF5555'
2025-11-01 15:36:22 +01:00
fi
printf '<span color="%s">%s %s°C, %s%%</span>\n' \
"$color" "$ICON" "$temp" "$load"
}
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
# Prefer NVIDIA if present (persist via --loop).
if command -v nvidia-smi > /dev/null 2>&1; then
# One child process for the lifetime of i3blocks; emits CSV every 5s.
nvidia-smi \
--query-gpu=temperature.gpu,utilization.gpu \
--format=csv,noheader,nounits \
--loop=5 2> /dev/null |
while IFS=',' read -r temp load; do
# Strip leading/trailing whitespace using parameter expansion.
temp=${temp## }
temp=${temp%% }
load=${load## }
load=${load%% }
[[ -z $temp || -z $load ]] && continue
emit "$temp" "$load"
done
exit 0
2024-11-15 14:06:42 +01:00
fi
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
# AMD fallback: read sysfs directly; emit once (i3blocks restarts on exit).
amdgpu=''
for d in /sys/class/hwmon/hwmon*/; do
[[ -r ${d}name ]] || continue
read -r n < "${d}name"
[[ $n == amdgpu ]] && {
amdgpu=$d
break
}
done
if [[ -n $amdgpu ]]; then
temp='N/A'
if [[ -r ${amdgpu}temp1_input ]]; then
read -r milli < "${amdgpu}temp1_input"
temp=$((milli / 1000))
2025-11-01 15:36:22 +01:00
fi
i3blocks: eliminate fork-storm with persist mode + zero-fork sysfs reads Resource-usage report showed ~29 cores of average load coming from i3blocks helper scripts forking awk/tr/grep/bc/sensors/nvidia-smi every tick. Rewrite all five hot-path scripts to eliminate forks: - volume.sh: persist mode, blocks on 'pactl subscribe' event stream. No polling, no sleep, no fork per tick. - gpu_monitor.sh: persist mode, single long-lived 'nvidia-smi --loop=5' feeds a bash 'while read' loop. Falls back to /sys for amdgpu. - battery_status.sh: reads /sys/class/power_supply/BAT*/ directly. Zero forks; replaces 'acpi | awk' pipeline. - cpu_monitor.sh: reads /proc/loadavg and k10temp/coretemp /sys/class/hwmon. Zero forks; replaces 'sensors | awk | tr' + bc arithmetic. - motherboard_temp.sh: reads nct*/it*/f71* Super-I/O hwmon node directly. Zero forks. Configure volume + gpu_monitor with interval=persist so i3blocks keeps one long-lived producer each instead of forking per tick. Also add: - kill_stale_recorders.sh -- kill stray ffmpeg x11grab / dotnet-trace / dotnet-monitor processes left running after sessions. - monitors.slice -- resource-capped user slice (CPUQuota=50%, MemoryMax=512M, MemorySwapMax=0 for zram safety, TasksMax=256) to bound future monitoring regressions. - efficient-polling-scripts SKILL -- rules for writing status-bar and polling scripts without forks; fork-pipeline to bash-builtin translation table; verification checklist. Verified live: strace -c on cpu_monitor.sh shows 1 execve / 0 clones; persist producers (pactl subscribe, nvidia-smi --loop) show 0 CPU ticks over a 3s idle sample. Per-invocation timing 1.6-1.9 ms (was 30-80 ms).
2026-04-20 21:54:11 +02:00
load='N/A'
# drm card matching the amdgpu hwmon exposes gpu_busy_percent.
for card in /sys/class/drm/card*/device/gpu_busy_percent; do
[[ -r $card ]] && {
read -r load < "$card"
break
}
done
emit "$temp" "$load"
exit 0
2024-11-15 14:06:42 +01:00
fi
printf '<span color="#FF5555">%s No supported GPU</span>\n' "$ICON"