* the capacity, it is doubled in size unless some other thread
   * holds the lock. If a hashed slot is empty, and lock is
   * available, a new Cell is created. Otherwise, if the slot
   * exists, a CAS is tried.  Retries proceed by "double hashing",
   * using a secondary hash (Marsaglia XorShift) to try to find a
   * free slot.
   *
   * The table size is capped because, when there are more threads

			b.Run("idle", runOne)

			b.Run("loaded", func(b *testing.B) {
				stop := applyGCLoad(b)
				runOne(b)
				// Make sure to stop the timer before we wait! The load created above
				// is very heavy-weight and not easy to stop, so we could end up
				// confusing the benchmarking framework for small b.N.
				b.StopTimer()
				stop()
			})
		}
	}

	// Measure the cost of counting goroutines

	buf := *bufp

	boff := 0 // starting read position in buf
	nbuf := 0 // end valid data in buf

	for {
		if boff >= nbuf {
			boff = 0
			stop := globalOSMetrics.time(osMetricReadDirent)
			nbuf, err = syscall.ReadDirent(fd, buf)
			stop()
			if err != nil {
				if isSysErrNotDir(err) {
					return nil
				}
				err = osErrToFileErr(err)
				if err == errFileNotFound {
					return nil

	//
	// Writes protected by trace.lock.
	shutdown atomic.Bool

	// Number of goroutines in syscall exiting slow path.
	exitingSyscall atomic.Int32

	// seqGC is the sequence counter for GC begin/end.
	//
	// Mutated only during stop-the-world.
	seqGC uint64

	// minPageHeapAddr is the minimum address of the page heap when tracing started.
	minPageHeapAddr uint64

   * the capacity, it is doubled in size unless some other thread
   * holds the lock. If a hashed slot is empty, and lock is
   * available, a new Cell is created. Otherwise, if the slot
   * exists, a CAS is tried.  Retries proceed by "double hashing",
   * using a secondary hash (Marsaglia XorShift) to try to find a
   * free slot.
   *
   * The table size is capped because, when there are more threads

 * <li>The daemon closes the connection once it has received the {@link Finished} message.</li>
 * </ul>
 *
 * <p>To stop a daemon:</p>
 *
 * <ul>
 * <li>The client creates a connection to daemon.</li>
 * <li>The client sends exactly one {@link Stop} message.</li>
 * <li>The daemon sends exactly one {@link Result} message. It may no longer send any messages.</li>

	// Fast path: all ASCII, no upper case.
	// Most paths look like this already.
	for i := 0; i < len(s); i++ {
		c := s[i]
		if c >= utf8.RuneSelf || 'A' <= c && c <= 'Z' {
			goto Slow
		}
	}
	return s

Slow:
	var b strings.Builder
	for _, r := range s {
		// SimpleFold(x) cycles to the next equivalent rune > x
		// or wraps around to smaller values. Iterate until it wraps,

			}

			var sloc []*Location
			for _, addr := range addrs {
				loc := locs[addr]
				if locs[addr] == nil {
					loc = &Location{
						Address: addr,
					}
					p.Location = append(p.Location, loc)
					locs[addr] = loc
				}
				sloc = append(sloc, loc)
			}
			s := &Sample{
				Value:    make([]int64, 2),
				Location: sloc,
			}

// Do not remove or change the type signature.
// See go.dev/issue/67401.
//
//go:linkname atomicwb
//go:nosplit
func atomicwb(ptr *unsafe.Pointer, new unsafe.Pointer) {
	slot := (*uintptr)(unsafe.Pointer(ptr))
	buf := getg().m.p.ptr().wbBuf.get2()
	buf[0] = *slot
	buf[1] = uintptr(new)
}

// atomicstorep performs *ptr = new atomically and invokes a write barrier.
//
//go:nosplit

	// call f, and the second would return immediately, without
	// waiting for the first's call to f to complete.
	// This is why the slow path falls back to a mutex, and why
	// the o.done.Store must be delayed until after f returns.

	if o.done.Load() == 0 {
		// Outlined slow-path to allow inlining of the fast-path.
		o.doSlow(f)
	}
}

func (o *Once) doSlow(f func()) {
	o.m.Lock()
	defer o.m.Unlock()