RGBCube/superfreq
1
Fork 0
mirror of https://github.com/RGBCube/superfreq synced 2025-07-27 17:07:44 +00:00
Code Issues Activity

daemon: cleaner idle time handling and progressive back-off adjustments

Browse source
This commit is contained in:
NotAShelf 2025-05-17 04:31:05 +03:00
parent 035d7b8848
commit 70a59cec82
No known key found for this signature in database
GPG key ID: 29D95B64378DB4BF
1 changed files with 24 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

34
src/daemon.rs
View file

@ -70,20 +70,34 @@ fn compute_new(params: &IntervalParams) -> u64 {
// Adjust for system idleness // Adjust for system idleness
if params.is_system_idle { if params.is_system_idle {
// Progressive back-off based on idle time duration let idle_time_seconds = params.last_user_activity.as_secs();
let idle_time_mins = params.last_user_activity.as_secs() / 60;
if idle_time_mins >= 1 { // Apply adjustment only if the system has been idle for a non-zero duration.
// Logarithmic back-off starting after 1 minute of idleness // The factor starts at 1.0 for 0 seconds idle time and increases.
// Use log base 2 to double the interval for each power of 2 minutes of idle time if idle_time_seconds > 0 {
// Example: 1min->1.5x, 2min->2x, 4min->3x, 8min->4x, 16min->5x, etc. let idle_factor = if idle_time_seconds < 120 {
let idle_factor = 1.0 + (idle_time_mins as f32).log2().max(0.5); // Less than 2 minutes (0 to 119 seconds)
// Linear interpolation from 1.0 (at 0s) to 2.0 (at 120s).
// Value at 60s (1 min) = 1.0 + 60.0/120.0 = 1.5.
// This should provide a smooth transition from no multiplier (or 1.0x)
// up to the point where the logarithmic scale takes over at 2 minutes.
1.0 + (idle_time_seconds as f32) / 120.0
} else {
// 2 minutes (120 seconds) or more
let idle_time_minutes = idle_time_seconds / 60;
// At 2 minutes (120s), (2_f32).log2() = 1.0. So, factor = 1.0 + 1.0 = 2.0.
1.0 + (idle_time_minutes as f32).log2().max(0.5)
};
// Cap the multiplier to avoid excessive intervals // Cap the multiplier to avoid excessive intervals
let capped_factor = idle_factor.min(5.0); let capped_factor = idle_factor.min(5.0); // max factor of 5x
debug!( debug!(
"System idle for {idle_time_mins} minutes, applying idle factor: {capped_factor:.1}x" "System idle for {} seconds (approx. {} minutes), applying idle factor: {:.2}x (raw: {:.2}x)",
idle_time_seconds,
(idle_time_seconds as f32 / 60.0).round(),
capped_factor,
idle_factor
); );
let multiplied = adjusted_interval as f64 * f64::from(capped_factor); let multiplied = adjusted_interval as f64 * f64::from(capped_factor);
@ -93,10 +107,10 @@ fn compute_new(params: &IntervalParams) -> u64 {
multiplied.round() as u64 multiplied.round() as u64
}; };
} }
// If idle_time_seconds is 0, no factor is applied by this block, effectively 1.0x.
} }
// Adjust for CPU/temperature volatility // Adjust for CPU/temperature volatility
// If either CPU usage or temperature is changing rapidly, decrease interval
if params.cpu_volatility > 10.0 || params.temp_volatility > 2.0 { if params.cpu_volatility > 10.0 || params.temp_volatility > 2.0 {
// XXX: This operation reduces the interval, so overflow is not an issue. // XXX: This operation reduces the interval, so overflow is not an issue.
// Using f64 for precision in multiplication before rounding. // Using f64 for precision in multiplication before rounding.