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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252
//! Statistics for work packets
use super::work_counter::{WorkCounter, WorkCounterBase, WorkDuration};
#[cfg(feature = "perf_counter")]
use crate::scheduler::work_counter::WorkPerfEvent;
use crate::vm::VMBinding;
use crate::MMTK;
use std::any::TypeId;
use std::collections::HashMap;
use std::marker::PhantomData;
use std::sync::atomic::{AtomicBool, Ordering};
/// Merge and print the work-packet level statistics from all worker threads
#[derive(Default)]
pub struct SchedulerStat {
/// Map work packet type IDs to work packet names
work_id_name_map: HashMap<TypeId, &'static str>,
/// Count the number of work packets executed for different types
work_counts: HashMap<TypeId, usize>,
/// Collect work counters from work threads.
/// Two dimensional vectors are used, e.g.
/// `[[foo_0, ..., foo_n], ..., [bar_0, ..., bar_n]]`.
/// The first dimension is for different types of work counters,
/// (`foo` and `bar` in the above example).
/// The second dimension if for work counters of the same type but from
/// different threads (`foo_0` and `bar_0` are from the same thread).
/// The order of insertion is determined by when [`SchedulerStat::merge`] is
/// called for each [`WorkerLocalStat`].
/// We assume different threads have the same set of work counters
/// (in the same order).
work_counters: HashMap<TypeId, Vec<Vec<Box<dyn WorkCounter>>>>,
}
impl SchedulerStat {
/// Extract the work-packet name from the full type name.
/// i.e. simplifies `crate::scheduler::gc_work::SomeWorkPacket<Semispace>` to `SomeWorkPacket`.
fn work_name(&self, name: &str) -> String {
let end_index = name.find('<').unwrap_or(name.len());
let name = name[..end_index].to_owned();
match name.rfind(':') {
Some(start_index) => name[(start_index + 1)..end_index].to_owned(),
_ => name,
}
}
/// Used during statistics printing at [`crate::memory_manager::harness_end`]
pub fn harness_stat(&self) -> HashMap<String, String> {
let mut stat = HashMap::new();
let mut counts = HashMap::<String, usize>::new();
let mut times = HashMap::<String, f64>::new();
// Work counts
let mut total_count = 0;
for (t, c) in &self.work_counts {
total_count += c;
let n = self.work_id_name_map[t];
// We can have the same work names for different TypeIDs since work names strip
// type parameters away, while the same work packet with different type parameters
// are given different TypeIDs. Hence, we check if the key exists and update instead of
// overwrite it
let pkt = format!("work.{}.count", self.work_name(n));
let val = counts.entry(pkt).or_default();
*val += c;
}
stat.insert("total-work.count".to_owned(), format!("{}", total_count));
// Work execution times
let mut duration_overall: WorkCounterBase = Default::default();
for (t, vs) in &self.work_counters {
// Name of the work packet type
let n = self.work_id_name_map[t];
// Iterate through different types of work counters
for v in vs.iter() {
// Aggregate work counters of the same type but from different
// worker threads
let fold = v
.iter()
.fold(Default::default(), |acc: WorkCounterBase, x| {
acc.merge(x.get_base())
});
// Update the overall execution time
duration_overall.merge_inplace(&fold);
let name = v.first().unwrap().name();
let pkt_total = format!("work.{}.{}.total", self.work_name(n), name);
let pkt_min = format!("work.{}.{}.min", self.work_name(n), name);
let pkt_max = format!("work.{}.{}.max", self.work_name(n), name);
// We can have the same work names for different TypeIDs since work names strip
// type parameters away, while the same work packet with different type parameters
// are given different TypeIDs. Hence, we check if the key exists and update
// instead of overwrite it
let val = times.entry(pkt_total).or_default();
*val += fold.total;
let val = times.entry(pkt_min).or_default();
*val = f64::min(*val, fold.min);
let val = times.entry(pkt_max).or_default();
*val = f64::max(*val, fold.max);
}
}
// Convert to ms and print out overall execution time
stat.insert(
"total-work.time.total".to_owned(),
format!("{:.3}", duration_overall.total / 1e6),
);
stat.insert(
"total-work.time.min".to_owned(),
format!("{:.3}", duration_overall.min / 1e6),
);
stat.insert(
"total-work.time.max".to_owned(),
format!("{:.3}", duration_overall.max / 1e6),
);
for (pkt, count) in counts {
stat.insert(pkt, format!("{}", count));
}
for (pkt, time) in times {
stat.insert(pkt, format!("{:.3}", time / 1e6));
}
stat
}
/// Merge work counters from different worker threads
pub fn merge<C>(&mut self, stat: &WorkerLocalStat<C>) {
// Merge work packet type ID to work packet name mapping
for (id, name) in &stat.work_id_name_map {
self.work_id_name_map.insert(*id, *name);
}
// Merge work count for different work packet types
for (id, count) in &stat.work_counts {
if self.work_counts.contains_key(id) {
*self.work_counts.get_mut(id).unwrap() += *count;
} else {
self.work_counts.insert(*id, *count);
}
}
// Merge work counter for different work packet types
for (id, counters) in &stat.work_counters {
// Initialize the two dimensional vector
// [
// [], // foo counter
// [], // bar counter
// ]
let vs = self
.work_counters
.entry(*id)
.or_insert_with(|| vec![vec![]; counters.len()]);
// [
// [counters[0] of type foo],
// [counters[1] of type bar]
// ]
for (v, c) in vs.iter_mut().zip(counters.iter()) {
v.push(c.clone());
}
}
}
}
/// Describing a single work packet
pub struct WorkStat {
type_id: TypeId,
type_name: &'static str,
}
impl WorkStat {
/// Stop all work counters for the work packet type of the just executed
/// work packet
pub fn end_of_work<VM: VMBinding>(&self, worker_stat: &mut WorkerLocalStat<VM>) {
if !worker_stat.is_enabled() {
return;
};
// Insert type ID, name pair
worker_stat
.work_id_name_map
.insert(self.type_id, self.type_name);
// Increment work count
*worker_stat.work_counts.entry(self.type_id).or_insert(0) += 1;
// Stop counters
worker_stat
.work_counters
.entry(self.type_id)
.and_modify(|v| {
v.iter_mut().for_each(|c| c.stop());
});
}
}
/// Worker thread local counterpart of [`SchedulerStat`]
pub struct WorkerLocalStat<C> {
work_id_name_map: HashMap<TypeId, &'static str>,
work_counts: HashMap<TypeId, usize>,
work_counters: HashMap<TypeId, Vec<Box<dyn WorkCounter>>>,
enabled: AtomicBool,
_phantom: PhantomData<C>,
}
unsafe impl<C> Send for WorkerLocalStat<C> {}
impl<C> Default for WorkerLocalStat<C> {
fn default() -> Self {
WorkerLocalStat {
work_id_name_map: Default::default(),
work_counts: Default::default(),
work_counters: Default::default(),
enabled: AtomicBool::new(false),
_phantom: Default::default(),
}
}
}
impl<VM: VMBinding> WorkerLocalStat<VM> {
pub fn is_enabled(&self) -> bool {
self.enabled.load(Ordering::SeqCst)
}
pub fn enable(&self) {
self.enabled.store(true, Ordering::SeqCst);
}
/// Measure the execution of a work packet by starting all counters for that
/// type
pub fn measure_work(
&mut self,
work_id: TypeId,
work_name: &'static str,
mmtk: &'static MMTK<VM>,
) -> WorkStat {
let stat = WorkStat {
type_id: work_id,
type_name: work_name,
};
if self.is_enabled() {
self.work_counters
.entry(work_id)
.or_insert_with(|| Self::counter_set(mmtk))
.iter_mut()
.for_each(|c| c.start());
}
stat
}
#[allow(unused_variables, unused_mut)]
fn counter_set(mmtk: &'static MMTK<VM>) -> Vec<Box<dyn WorkCounter>> {
let mut counters: Vec<Box<dyn WorkCounter>> = vec![Box::new(WorkDuration::new())];
#[cfg(feature = "perf_counter")]
for e in &mmtk.options.work_perf_events.events {
counters.push(Box::new(WorkPerfEvent::new(
&e.0,
e.1,
e.2,
*mmtk.options.perf_exclude_kernel,
)));
}
counters
}
}