mltop 0.6.0

Resource monitor for ML engineers written in Rust
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148

use nvml_wrapper::{error::NvmlError, Nvml};
use std::collections::HashMap;
use sysinfo::System;
use sysinfo::ThreadKind;

#[derive(Clone)]
pub enum ProcessType {
    GpuGraphic,
    GpuCompute,
    Cpu,
    UserThread,
    KernelThread,
}

impl ProcessType {
    fn is_thread(&self) -> bool {
        matches!(self, ProcessType::UserThread | ProcessType::KernelThread)
    }
}

impl ToString for ProcessType {
    fn to_string(&self) -> String {
        match self {
            ProcessType::GpuGraphic => "GRAPHIC".to_string(),
            ProcessType::GpuCompute => "COMPUTE".to_string(),
            ProcessType::Cpu => "CPU".to_string(),
            // For now i'll display both thread types equally,
            // I'm not sure if we really want to distiguish between the two
            ProcessType::UserThread => "THREAD".to_string(),
            ProcessType::KernelThread => "THREAD".to_string(),
        }
    }
}

#[derive(Clone)]
pub struct Process {
    pub pid: u32,
    pub type_: ProcessType,
    pub command: String,
    // percentage 0-100% (can be higher than 100% if uses more than one core)
    pub cpu_usage: f32,
    // bytes
    pub memory: u64,
    // percentage 0-100%
    pub memory_usage: f32,
}

impl Process {
    pub fn is_thread(&self) -> bool {
        self.type_.is_thread()
    }
}

#[derive(Clone)]
pub struct ProcessesSnapshot {
    pub processes: Vec<Process>,
}

impl ProcessesSnapshot {
    pub fn from_sysinfo_nvml(sys: &System, nvml: Option<&Nvml>) -> ProcessesSnapshot {
        let total_memory = sys.total_memory();

        let mut processes: HashMap<u32, Process> = sys
            .processes()
            .iter()
            .filter_map(|(pid, p)| {
                let pid = pid.as_u32();
                let memory = p.memory();

                // TODO: Investigate why p.cmd() sometimes returns an empty array
                let cmd_list = p.cmd();
                if cmd_list.is_empty() {
                    return None;
                }

                Some((
                    pid,
                    Process {
                        pid,
                        type_: match p.thread_kind() {
                            Some(tk) => match tk {
                                ThreadKind::Kernel => ProcessType::KernelThread,
                                ThreadKind::Userland => ProcessType::UserThread,
                            },
                            None => ProcessType::Cpu,
                        },
                        command: cmd_list
                            .iter()
                            .map(|s| s.to_string())
                            .collect::<Vec<_>>()
                            .join(" "),
                        memory,
                        memory_usage: (memory as f32 / total_memory as f32) * 100.0,
                        cpu_usage: p.cpu_usage(),
                    },
                ))
            })
            .collect();

        // find which ones are GPU and mark them as such
        if nvml.is_some() {
            let nvml = nvml.as_ref().unwrap();

            match _gpu_compute_pids(&nvml) {
                Ok(pids) => _update_process_type(pids, &mut processes, ProcessType::GpuCompute),
                Err(_) => {}
            }
            match _gpu_graphics_pids(&nvml) {
                Ok(pids) => _update_process_type(pids, &mut processes, ProcessType::GpuGraphic),
                Err(_) => {}
            }
        }

        ProcessesSnapshot {
            processes: processes.into_values().collect(),
        }
    }
}

fn _gpu_compute_pids(nvml: &Nvml) -> Result<Vec<u32>, NvmlError> {
    let device = nvml.device_by_index(0)?;
    Ok(device
        .running_compute_processes()?
        .iter()
        .map(|pi| pi.pid)
        .collect())
}

fn _gpu_graphics_pids(nvml: &Nvml) -> Result<Vec<u32>, NvmlError> {
    let device = nvml.device_by_index(0)?;
    Ok(device
        .running_graphics_processes()?
        .iter()
        .map(|pi| pi.pid)
        .collect())
}

fn _update_process_type(
    pids: Vec<u32>,
    processes: &mut HashMap<u32, Process>,
    process_type: ProcessType,
) {
    for pid in pids {
        if let Some(obj) = processes.get_mut(&pid) {
            obj.type_ = process_type.clone();
        }
    }
}