refactor: Another small optimization pass (#350)

Making some small changes that would hopefully improve performance a bit.

- Remove redundant string generations for CPU data conversion
- Switch to fnv for PID hashmap and hashsets
- Use buffered reading to avoid having to store too many lines as strings
This commit is contained in:
Clement Tsang 2020-12-11 19:54:02 -05:00 committed by GitHub
parent 86135e466c
commit 8c4ad90e67
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 123 additions and 86 deletions

1
.gitignore vendored
View file

@ -12,6 +12,7 @@
rust-unmangle
*.svg
*.data
*.data.old
# IntelliJ
.idea/

7
Cargo.lock generated
View file

@ -121,6 +121,7 @@ dependencies = [
"ctrlc",
"dirs-next",
"fern",
"fnv",
"futures",
"heim",
"indexmap",
@ -389,6 +390,12 @@ dependencies = [
"num-traits",
]
[[package]]
name = "fnv"
version = "1.0.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
[[package]]
name = "futures"
version = "0.3.8"

View file

@ -19,9 +19,10 @@ doc = false
[profile.release]
debug = 1
lto = true
# debug = true
# lto = false
opt-level = 3
lto = "fat"
codegen-units = 1
[dependencies]
@ -33,6 +34,7 @@ crossterm = "0.18.2"
ctrlc = {version = "3.1", features = ["termination"]}
clap = "2.33"
dirs-next = "2.0.0"
fnv = "1.0.7"
futures = "0.3.8"
indexmap = "1.6.0"
itertools = "0.9.0"

View file

@ -3,7 +3,7 @@
use std::time::Instant;
#[cfg(target_os = "linux")]
use std::collections::HashMap;
use fnv::FnvHashMap;
use sysinfo::{System, SystemExt};
@ -73,7 +73,7 @@ pub struct DataCollector {
pub data: Data,
sys: System,
#[cfg(target_os = "linux")]
pid_mapping: HashMap<crate::Pid, processes::PrevProcDetails>,
pid_mapping: FnvHashMap<crate::Pid, processes::PrevProcDetails>,
#[cfg(target_os = "linux")]
prev_idle: f64,
#[cfg(target_os = "linux")]
@ -99,7 +99,7 @@ impl Default for DataCollector {
data: Data::default(),
sys: System::new_with_specifics(sysinfo::RefreshKind::new()),
#[cfg(target_os = "linux")]
pid_mapping: HashMap::new(),
pid_mapping: FnvHashMap::default(),
#[cfg(target_os = "linux")]
prev_idle: 0_f64,
#[cfg(target_os = "linux")]

View file

@ -6,7 +6,7 @@ use sysinfo::ProcessStatus;
use crate::utils::error::{self, BottomError};
#[cfg(target_os = "linux")]
use std::collections::{hash_map::RandomState, HashMap};
use fnv::{FnvHashMap, FnvHashSet};
#[cfg(not(target_os = "linux"))]
use sysinfo::{ProcessExt, ProcessorExt, System, SystemExt};
@ -88,7 +88,7 @@ pub struct PrevProcDetails {
pub cpu_time: f64,
pub proc_stat_path: PathBuf,
// pub proc_statm_path: PathBuf,
pub proc_exe_path: PathBuf,
// pub proc_exe_path: PathBuf,
pub proc_io_path: PathBuf,
pub proc_cmdline_path: PathBuf,
pub just_read: bool,
@ -98,7 +98,7 @@ impl PrevProcDetails {
pub fn new(pid: Pid) -> Self {
PrevProcDetails {
proc_io_path: PathBuf::from(format!("/proc/{}/io", pid)),
proc_exe_path: PathBuf::from(format!("/proc/{}/exe", pid)),
// proc_exe_path: PathBuf::from(format!("/proc/{}/exe", pid)),
proc_stat_path: PathBuf::from(format!("/proc/{}/stat", pid)),
// proc_statm_path: PathBuf::from(format!("/proc/{}/statm", pid)),
proc_cmdline_path: PathBuf::from(format!("/proc/{}/cmdline", pid)),
@ -111,23 +111,17 @@ impl PrevProcDetails {
fn cpu_usage_calculation(
prev_idle: &mut f64, prev_non_idle: &mut f64,
) -> error::Result<(f64, f64)> {
use std::io::prelude::*;
use std::io::BufReader;
// From SO answer: https://stackoverflow.com/a/23376195
let mut path = std::path::PathBuf::new();
path.push("/proc");
path.push("stat");
let stat_results = std::fs::read_to_string(path)?;
let first_line: &str;
let split_results = stat_results.split('\n').collect::<Vec<&str>>();
if split_results.is_empty() {
return Err(error::BottomError::InvalidIO(format!(
"Unable to properly split the stat results; saw {} values, expected at least 1 value.",
split_results.len()
)));
} else {
first_line = split_results[0];
}
let mut reader = BufReader::new(std::fs::File::open(path)?);
let mut first_line = String::new();
reader.read_line(&mut first_line)?;
let val = first_line.split_whitespace().collect::<Vec<&str>>();
@ -176,20 +170,6 @@ fn cpu_usage_calculation(
Ok((result, cpu_percentage))
}
#[cfg(target_os = "linux")]
fn get_process_io(path: &PathBuf) -> std::io::Result<String> {
Ok(std::fs::read_to_string(path)?)
}
#[cfg(target_os = "linux")]
fn get_linux_process_io_usage(stat: &[&str]) -> (u64, u64) {
// Represents read_bytes and write_bytes
(
stat[9].parse::<u64>().unwrap_or(0),
stat[11].parse::<u64>().unwrap_or(0),
)
}
#[cfg(target_os = "linux")]
fn get_linux_process_vsize_rss(stat: &[&str]) -> (u64, u64) {
// Represents vsize and rss (bytes and page numbers respectively)
@ -200,6 +180,7 @@ fn get_linux_process_vsize_rss(stat: &[&str]) -> (u64, u64) {
}
#[cfg(target_os = "linux")]
/// Preferably use this only on small files.
fn read_path_contents(path: &PathBuf) -> std::io::Result<String> {
Ok(std::fs::read_to_string(path)?)
}
@ -246,11 +227,14 @@ fn get_linux_cpu_usage(
#[allow(clippy::too_many_arguments)]
#[cfg(target_os = "linux")]
fn read_proc<S: core::hash::BuildHasher>(
fn read_proc(
pid: Pid, cpu_usage: f64, cpu_fraction: f64,
pid_mapping: &mut HashMap<Pid, PrevProcDetails, S>, use_current_cpu_total: bool,
pid_mapping: &mut FnvHashMap<Pid, PrevProcDetails>, use_current_cpu_total: bool,
time_difference_in_secs: u64, mem_total_kb: u64, page_file_kb: u64,
) -> error::Result<ProcessHarvest> {
use std::io::prelude::*;
use std::io::BufReader;
let pid_stat = pid_mapping
.entry(pid)
.or_insert_with(|| PrevProcDetails::new(pid));
@ -321,11 +305,33 @@ fn read_proc<S: core::hash::BuildHasher>(
let mem_usage_bytes = mem_usage_kb * 1024;
// This can fail if permission is denied!
let (total_read_bytes, total_write_bytes, read_bytes_per_sec, write_bytes_per_sec) =
if let Ok(io_results) = get_process_io(&pid_stat.proc_io_path) {
let io_stats = io_results.split_whitespace().collect::<Vec<&str>>();
let (total_read_bytes, total_write_bytes) = get_linux_process_io_usage(&io_stats);
let (total_read_bytes, total_write_bytes, read_bytes_per_sec, write_bytes_per_sec) =
if let Ok(file) = std::fs::File::open(&pid_stat.proc_io_path) {
let reader = BufReader::new(file);
let mut lines = reader.lines().skip(4);
// Represents read_bytes and write_bytes, at the 5th and 6th lines (1-index, not 0-index)
let total_read_bytes = if let Some(Ok(read_bytes_line)) = lines.next() {
if let Some(read_bytes) = read_bytes_line.split_whitespace().last() {
read_bytes.parse::<u64>().unwrap_or(0)
} else {
0
}
} else {
0
};
let total_write_bytes = if let Some(Ok(write_bytes_line)) = lines.next() {
if let Some(write_bytes) = write_bytes_line.split_whitespace().last() {
write_bytes.parse::<u64>().unwrap_or(0)
} else {
0
}
} else {
0
};
let read_bytes_per_sec = if time_difference_in_secs == 0 {
0
} else {
@ -371,14 +377,13 @@ fn read_proc<S: core::hash::BuildHasher>(
#[cfg(target_os = "linux")]
pub fn get_process_data(
prev_idle: &mut f64, prev_non_idle: &mut f64,
pid_mapping: &mut HashMap<Pid, PrevProcDetails, RandomState>, use_current_cpu_total: bool,
pid_mapping: &mut FnvHashMap<Pid, PrevProcDetails>, use_current_cpu_total: bool,
time_difference_in_secs: u64, mem_total_kb: u64, page_file_kb: u64,
) -> crate::utils::error::Result<Vec<ProcessHarvest>> {
// TODO: [PROC THREADS] Add threads
use std::collections::HashSet;
if let Ok((cpu_usage, cpu_fraction)) = cpu_usage_calculation(prev_idle, prev_non_idle) {
let mut pids_to_clear: HashSet<Pid> = pid_mapping.keys().cloned().collect();
let mut pids_to_clear: FnvHashSet<Pid> = pid_mapping.keys().cloned().collect();
let process_vector: Vec<ProcessHarvest> = std::fs::read_dir("/proc")?
.filter_map(|dir| {
if let Ok(dir) = dir {

View file

@ -215,8 +215,12 @@ fn main() -> Result<()> {
if app.used_widgets.use_cpu {
// CPU
app.canvas_data.cpu_data =
convert_cpu_data_points(&app.data_collection, false);
convert_cpu_data_points(
&app.data_collection,
&mut app.canvas_data.cpu_data,
false,
);
}
// Processes

View file

@ -178,9 +178,9 @@ pub fn convert_disk_row(
}
pub fn convert_cpu_data_points(
current_data: &data_farmer::DataCollection, is_frozen: bool,
) -> Vec<ConvertedCpuData> {
let mut cpu_data_vector: Vec<ConvertedCpuData> = Vec::new();
current_data: &data_farmer::DataCollection, existing_cpu_data: &mut Vec<ConvertedCpuData>,
is_frozen: bool,
) {
let current_time = if is_frozen {
if let Some(frozen_instant) = current_data.frozen_instant {
frozen_instant
@ -191,39 +191,62 @@ pub fn convert_cpu_data_points(
current_data.current_instant
};
// Initialize cpu_data_vector if the lengths don't match...
if let Some((_time, data)) = &current_data.timed_data_vec.last() {
if data.cpu_data.len() + 1 != existing_cpu_data.len() {
*existing_cpu_data = vec![ConvertedCpuData {
cpu_name: "All".to_string(),
short_cpu_name: "All".to_string(),
cpu_data: vec![],
legend_value: String::new(),
}];
existing_cpu_data.extend(
data.cpu_data
.iter()
.enumerate()
.map(|(itx, cpu_usage)| ConvertedCpuData {
cpu_name: if let Some(cpu_harvest) = current_data.cpu_harvest.get(itx) {
if let Some(cpu_count) = cpu_harvest.cpu_count {
format!("{}{}", cpu_harvest.cpu_prefix, cpu_count)
} else {
cpu_harvest.cpu_prefix.to_string()
}
} else {
String::default()
},
short_cpu_name: if let Some(cpu_harvest) = current_data.cpu_harvest.get(itx)
{
if let Some(cpu_count) = cpu_harvest.cpu_count {
cpu_count.to_string()
} else {
cpu_harvest.cpu_prefix.to_string()
}
} else {
String::default()
},
legend_value: format!("{:.0}%", cpu_usage.round()),
cpu_data: vec![],
})
.collect::<Vec<ConvertedCpuData>>(),
);
} else {
existing_cpu_data
.iter_mut()
.skip(1)
.zip(&data.cpu_data)
.for_each(|(cpu, cpu_usage)| {
cpu.cpu_data = vec![];
cpu.legend_value = format!("{:.0}%", cpu_usage.round());
});
}
}
for (time, data) in &current_data.timed_data_vec {
let time_from_start: f64 = (current_time.duration_since(*time).as_millis() as f64).floor();
for (itx, cpu) in data.cpu_data.iter().enumerate() {
// Check if the vector exists yet
if cpu_data_vector.len() <= itx {
let new_cpu_data = ConvertedCpuData {
cpu_name: if let Some(cpu_harvest) = current_data.cpu_harvest.get(itx) {
if let Some(cpu_count) = cpu_harvest.cpu_count {
format!("{}{}", cpu_harvest.cpu_prefix, cpu_count)
} else {
cpu_harvest.cpu_prefix.to_string()
}
} else {
String::default()
},
short_cpu_name: if let Some(cpu_harvest) = current_data.cpu_harvest.get(itx) {
if let Some(cpu_count) = cpu_harvest.cpu_count {
cpu_count.to_string()
} else {
cpu_harvest.cpu_prefix.to_string()
}
} else {
String::default()
},
..ConvertedCpuData::default()
};
cpu_data_vector.push(new_cpu_data);
}
if let Some(cpu_data) = cpu_data_vector.get_mut(itx) {
cpu_data.legend_value = format!("{:.0}%", cpu.round());
if let Some(cpu_data) = existing_cpu_data.get_mut(itx + 1) {
cpu_data.cpu_data.push((-time_from_start, *cpu));
}
}
@ -232,15 +255,6 @@ pub fn convert_cpu_data_points(
break;
}
}
let mut extended_vec = vec![ConvertedCpuData {
cpu_name: "All".to_string(),
short_cpu_name: "All".to_string(),
cpu_data: vec![],
legend_value: String::new(),
}];
extended_vec.extend(cpu_data_vector);
extended_vec
}
pub fn convert_mem_data_points(

View file

@ -309,7 +309,11 @@ pub fn handle_force_redraws(app: &mut App) {
}
if app.cpu_state.force_update.is_some() {
app.canvas_data.cpu_data = convert_cpu_data_points(&app.data_collection, app.is_frozen);
convert_cpu_data_points(
&app.data_collection,
&mut app.canvas_data.cpu_data,
app.is_frozen,
);
app.cpu_state.force_update = None;
}