#!/usr/bin/env bash # Written by Draco (tytydraco @ GitHub) # The name of the current branch for logging purposes BRANCH="balance" # Maximum unsigned integer size in C UINT_MAX="4294967295" # Duration in nanoseconds of one scheduling period SCHED_PERIOD="$((4 * 1000 * 1000))" # How many tasks should we have at a maximum in one scheduling period SCHED_TASKS="8" write() { # Bail out if file does not exist [[ ! -f "$1" ]] && return 1 # Make file writable in case it is not already chmod +w "$1" 2> /dev/null # Write the new value and bail if there's an error if ! echo "$2" > "$1" 2> /dev/null then echo "Failed: $1 → $2" return 1 fi # Log the success echo "$1 → $2" } # Check for root permissions and bail if not granted if [[ "$(id -u)" -ne 0 ]] then echo "No root permissions. Exiting." exit 1 fi # Detect if we are running on Android grep -q android /proc/cmdline && ANDROID=true # Log the date and time for records sake echo "Time of execution: $(date)" echo "Branch: $BRANCH" # Sync to data in the rare case a device crashes sync # Limit max perf event processing time to this much CPU usage write /proc/sys/kernel/perf_cpu_time_max_percent 5 # Group tasks for less stutter but less throughput write /proc/sys/kernel/sched_autogroup_enabled 1 # Execute child process before parent after fork write /proc/sys/kernel/sched_child_runs_first 1 # Preliminary requirement for the following values write /proc/sys/kernel/sched_tunable_scaling 0 # Reduce the maximum scheduling period for lower latency write /proc/sys/kernel/sched_latency_ns "$SCHED_PERIOD" # Schedule this ratio of tasks in the guarenteed sched period write /proc/sys/kernel/sched_min_granularity_ns "$((SCHED_PERIOD / SCHED_TASKS))" # Require preeptive tasks to surpass half of a sched period in vmruntime write /proc/sys/kernel/sched_wakeup_granularity_ns "$((SCHED_PERIOD / 2))" # Reduce the frequency of task migrations write /proc/sys/kernel/sched_migration_cost_ns 5000000 # Always allow sched boosting on top-app tasks [[ "$ANDROID" == true ]] && write /proc/sys/kernel/sched_min_task_util_for_colocation 0 # Improve real time latencies by reducing the scheduler migration time write /proc/sys/kernel/sched_nr_migrate 32 # Disable scheduler statistics to reduce overhead write /proc/sys/kernel/sched_schedstats 0 # Disable unnecessary printk logging write /proc/sys/kernel/printk_devkmsg off # Start non-blocking writeback later write /proc/sys/vm/dirty_background_ratio 10 # Start blocking writeback later write /proc/sys/vm/dirty_ratio 30 # Require dirty memory to stay in memory for longer write /proc/sys/vm/dirty_expire_centisecs 3000 # Run the dirty memory flusher threads less often write /proc/sys/vm/dirty_writeback_centisecs 3000 # Disable read-ahead for swap devices write /proc/sys/vm/page-cluster 0 # Update /proc/stat less often to reduce jitter write /proc/sys/vm/stat_interval 10 # Swap to the swap device at a fair rate write /proc/sys/vm/swappiness 100 # Fairly prioritize page cache and file structures write /proc/sys/vm/vfs_cache_pressure 100 # Enable Explicit Congestion Control write /proc/sys/net/ipv4/tcp_ecn 1 # Enable fast socket open for receiver and sender write /proc/sys/net/ipv4/tcp_fastopen 3 # Disable SYN cookies write /proc/sys/net/ipv4/tcp_syncookies 0 if [[ -f "/sys/kernel/debug/sched_features" ]] then # Consider scheduling tasks that are eager to run write /sys/kernel/debug/sched_features NEXT_BUDDY # Schedule tasks on their origin CPU if possible write /sys/kernel/debug/sched_features TTWU_QUEUE fi [[ "$ANDROID" == true ]] && if [[ -d "/dev/stune/" ]] then # We are not concerned with prioritizing latency write /dev/stune/top-app/schedtune.prefer_idle 0 # Mark top-app as boosted, find high-performing CPUs write /dev/stune/top-app/schedtune.boost 1 fi # Loop over each CPU in the system for cpu in /sys/devices/system/cpu/cpu*/cpufreq do # Fetch the available governors from the CPU avail_govs="$(cat "$cpu/scaling_available_governors")" # Attempt to set the governor in this order for governor in schedutil interactive do # Once a matching governor is found, set it and break for this CPU if [[ "$avail_govs" == *"$governor"* ]] then write "$cpu/scaling_governor" "$governor" break fi done done # Apply governor specific tunables for schedutil find /sys/devices/system/cpu/ -name schedutil -type d | while IFS= read -r governor do # Consider changing frequencies once per scheduling period write "$governor/up_rate_limit_us" "$((SCHED_PERIOD / 1000))" write "$governor/down_rate_limit_us" "$((4 * SCHED_PERIOD / 1000))" write "$governor/rate_limit_us" "$((SCHED_PERIOD / 1000))" # Jump to hispeed frequency at this load percentage write "$governor/hispeed_load" 90 write "$governor/hispeed_freq" "$UINT_MAX" done # Apply governor specific tunables for interactive find /sys/devices/system/cpu/ -name interactive -type d | while IFS= read -r governor do # Consider changing frequencies once per scheduling period write "$governor/timer_rate" "$((SCHED_PERIOD / 1000))" write "$governor/min_sample_time" "$((SCHED_PERIOD / 1000))" # Jump to hispeed frequency at this load percentage write "$governor/go_hispeed_load" 90 write "$governor/hispeed_freq" "$UINT_MAX" done for queue in /sys/block/*/queue do # Choose the first governor available avail_scheds="$(cat "$queue/scheduler")" for sched in cfq noop kyber bfq mq-deadline none do if [[ "$avail_scheds" == *"$sched"* ]] then write "$queue/scheduler" "$sched" break fi done # Do not use I/O as a source of randomness write "$queue/add_random" 0 # Disable I/O statistics accounting write "$queue/iostats" 0 # Reduce heuristic read-ahead in exchange for I/O latency write "$queue/read_ahead_kb" 128 # Reduce the maximum number of I/O requests in exchange for latency write "$queue/nr_requests" 64 done # Always return success, even if the last write fails exit 0