samueleresca

END
{
  printf("\nThread state analysis:\n");
  printf("%-16s %-6s %6s %6s %6s %6s %6s %6s %6s \n", "COMM", "PID", "CPU", "RUNQ", "SLP", "USL", "SUS", "LCK", "DEA");

  // Convert nanoseconds to milliseconds
  $ms = (uint64) 1000000;

  for ($kv : @cpu)
  {

samueleresca

tracepoint:lock:contention_begin
{
  @lock_ts[tid] = nsecs;
}

// Track futex wake operations (lock release)
tracepoint:lock:contention_end
/has_key(@lock_ts, tid)/
{
    @lck[comm, tid] = sum(nsecs - @lock_ts[tid]);

samueleresca

// The futex states as defined in uapi/linux/futex.h
#define FUTEX_WAIT		0
#define FUTEX_WAIT_BITSET	9
#define FUTEX_PRIVATE_FLAG	128
#define FUTEX_WAIT_PRIVATE	(FUTEX_WAIT | FUTEX_PRIVATE_FLAG)
#define FUTEX_WAIT_BITSET_PRIVATE	(FUTEX_WAIT_BITSET | FUTEX_PRIVATE_FLAG)

// Track futex wait operations (lock attempts)
tracepoint:syscalls:sys_enter_futex
/args->op == FUTEX_WAIT || args->op == FUTEX_WAIT_BITSET || args->op == FUTEX_WAIT_PRIVATE || args->op == FUTEX_WAIT_BITSET_PRIVATE /

samueleresca

#!/usr/bin/bpftrace

#include <linux/sched.h>

tracepoint:sched:sched_switch
{
  $prev_pid = args->prev_pid;
  $next_pid = args->next_pid;
  $prev_state = args->prev_state;

samueleresca

tracepoint:sched:sched_switch
{
  $prev_pid = args->prev_pid;
  $next_pid = args->next_pid;
  $prev_state = args->prev_state;

  // Record on-CPU time for previous task
  if (has_key(@start, $prev_pid))
  {
    @cpu[comm, $prev_pid] = sum(nsecs - @start[$prev_pid]);

samueleresca

// On-CPU time tracking tracking
tracepoint:sched:sched_switch
{
  $prev_pid = args->prev_pid;
  $next_pid = args->next_pid;
  $prev_state = args->prev_state;

  // Record on-CPU time for previous task
  if (has_key(@start, $prev_pid))
  {

samueleresca

use cache_size::l1_cache_line_size;

/// Returns the result of a matrix multiplication.
/// Uses loop tiling to optimize the matrix multiplication.
/// The function relies on cache_size crate to get the cache line size.
///
/// # Arguments
///
/// * `n` - The dimension of the matrix.
pub fn optimized_tiled(n: usize) -> Vec<Vec<f64>> {

samueleresca

/// # Arguments
///
/// * `n` - The dimension of the matrix.
/// 
/// # Returns a matrix with values from 0 to n - 1.
/// Initializes a matrix with values from 0 to n - 1.
pub fn generate_matrix(n: usize) -> Vec<Vec<f64>> {
    let mut matrix = vec![vec![0.0; n]; n];
    let mut value = 0.0;
    for r in 0..n {

samueleresca

/// Returns the result of the matrix multiplication.
/// 
/// # Arguments
/// 
/// * `n` - The dimension of the matrix.
pub fn non_optimized(n: usize) -> Vec<Vec<f64>> {
    let m1 = generate_matrix(n);
    let m2 = m1.clone();

    let mut res = vec![vec![0.0; n]; n];

samueleresca

/// Returns an initialized matrix. The matrix is initialized in a standard way.
///
/// # Arguments
///
/// * `n` - The dimension of the matrix.
pub fn standard_initialize(n: usize) -> Vec<Vec<i32>> {
    let mut data = vec![vec![0i32; n]; n];

    for r in 0..n {
        for c in 0..n {