RGBCube/serenity
1
Fork 0
mirror of https://github.com/RGBCube/serenity synced 2025-05-15 01:54:57 +00:00
Code Issues Activity Actions

Everywhere: Improve CPU usage calculation

Browse source 
As threads come and go, we can't simply account for how many time
slices the threads at any given point may have been using. We need to
also account for threads that have since disappeared. This means we
also need to track how many time slices we have expired globally.

However, because this doesn't account for context switches outside of
the system timer tick values may still be under-reported. To solve this
we will need to track more accurate time information on each context
switch.

This also fixes top's cpu usage calculation which was still based on
the number of context switches.

Fixes #6473
This commit is contained in:
Tom 2021-07-14 12:05:59 -06:00 • committed by Andreas Kling
parent ef85c4f747
commit 7e77a2ec40
17 changed files with 132 additions and 83 deletions
Download patch file Download diff file Expand all files Collapse all files

25
Userland/Applications/SystemMonitor/ProcessModel.cpp
View file

@ -318,17 +318,18 @@ void ProcessModel::update()
auto previous_tid_count = m_tids.size();
auto all_processes = Core::ProcessStatisticsReader::get_all(m_proc_all);
u64 last_sum_ticks_scheduled = 0, last_sum_ticks_scheduled_kernel = 0;
for (auto& it : m_threads) {
auto& current_state = it.value->current_state;
last_sum_ticks_scheduled += current_state.ticks_user + current_state.ticks_kernel;
last_sum_ticks_scheduled_kernel += current_state.ticks_kernel;
}
HashTable<int> live_tids;
u64 sum_ticks_scheduled = 0, sum_ticks_scheduled_kernel = 0;
u64 total_ticks_scheduled_diff = 0;
if (all_processes.has_value()) {
for (auto& process : all_processes.value()) {
if (m_has_total_ticks)
total_ticks_scheduled_diff = all_processes->total_ticks_scheduled - m_total_ticks_scheduled;
m_total_ticks_scheduled = all_processes->total_ticks_scheduled;
m_total_ticks_scheduled_kernel = all_processes->total_ticks_scheduled_kernel;
m_has_total_ticks = true;
for (auto& process : all_processes.value().processes) {
for (auto& thread : process.threads) {
ThreadState state;
state.kernel = process.kernel;
@ -388,6 +389,7 @@ void ProcessModel::update()
c.total_cpu_percent = 0.0;
c.total_cpu_percent_kernel = 0.0;
}
Vector<int, 16> tids_to_remove;
for (auto& it : m_threads) {
if (!live_tids.contains(it.key)) {
@ -398,8 +400,8 @@ void ProcessModel::update()
u32 ticks_scheduled_diff = (thread.current_state.ticks_user + thread.current_state.ticks_kernel)
- (thread.previous_state.ticks_user + thread.previous_state.ticks_kernel);
u32 ticks_scheduled_diff_kernel = thread.current_state.ticks_kernel - thread.previous_state.ticks_kernel;
thread.current_state.cpu_percent = ((float)ticks_scheduled_diff * 100) / (float)(sum_ticks_scheduled - last_sum_ticks_scheduled);
thread.current_state.cpu_percent_kernel = ((float)ticks_scheduled_diff_kernel * 100) / (float)(sum_ticks_scheduled - last_sum_ticks_scheduled);
thread.current_state.cpu_percent = total_ticks_scheduled_diff > 0 ? ((float)ticks_scheduled_diff * 100) / (float)total_ticks_scheduled_diff : 0;
thread.current_state.cpu_percent_kernel = total_ticks_scheduled_diff > 0 ? ((float)ticks_scheduled_diff_kernel * 100) / (float)total_ticks_scheduled_diff : 0;
if (it.value->current_state.pid != 0) {
auto& cpu_info = m_cpus[thread.current_state.cpu];
cpu_info.total_cpu_percent += thread.current_state.cpu_percent;
@ -407,6 +409,7 @@ void ProcessModel::update()
m_tids.append(it.key);
}
}
for (auto tid : tids_to_remove)
m_threads.remove(tid);
@ -414,7 +417,7 @@ void ProcessModel::update()
on_cpu_info_change(m_cpus);
if (on_state_update)
on_state_update(all_processes->size(), m_threads.size());
on_state_update(all_processes.has_value() ? all_processes->processes.size() : 0, m_threads.size());
// FIXME: This is a rather hackish way of invalidating indices.
// It would be good if GUI::Model had a way to orchestrate removal/insertion while preserving indices.