// consistency, a P going idle must the idle mask simultaneously with
	// updates to the idle P list under the sched.lock, otherwise a racing
	// pidleget may clear the mask before pidleput sets the mask,
	// corrupting the bitmap.
	//
	// N.B., procresize takes ownership of all Ps in stopTheWorldWithSema.
	idlepMask pMask

	// Bitmask of Ps that may have a timer, one bit per P. Reads and writes

	mtop := int((logM - 2) / _W) // -2 because the top word of N bits is the (N-1)/W'th word.
	mmask := ^Word(0)
	if mbits := (logM - 1) & (_W - 1); mbits != 0 {
		mmask = (1 << mbits) - 1
	}
	if i > mtop {
		i = mtop
	}
	advance := false
	z = z.setWord(1)
	for ; i >= 0; i-- {
		yi := y[i]
		if i == mtop {
			yi &= mmask
		}
		for j := 0; j < _W; j += n {
			if advance {

//go:linkname abigen_sync_atomic_OrInt64 sync/atomic.OrInt64
func abigen_sync_atomic_OrInt64(addr *int64, mask int64) (old int64)

//go:linkname abigen_sync_atomic_OrUint64 sync/atomic.OrUint64
func abigen_sync_atomic_OrUint64(addr *uint64, mask uint64) (old uint64)

//go:linkname abigen_sync_atomic_OrUintptr sync/atomic.OrUintptr
func abigen_sync_atomic_OrUintptr(addr *uintptr, mask uintptr) (old uintptr)

	// ITypeImmMask is a mask including only the immediate portion of
	// I-type instructions.
	ITypeImmMask = 0xfff00000

	// JTypeImmMask is a mask including only the immediate portion of
	// J-type instructions.
	JTypeImmMask = 0xfffff000

	// STypeImmMask is a mask including only the immediate portion of
	// S-type instructions.
	STypeImmMask = 0xfe000f80

	// UTypeImmMask is a mask including only the immediate portion of

the SIGPROF signal in particular.

The non-Go code should not change the signal mask on any threads
created by the Go runtime. If the non-Go code starts new threads
itself, those threads may set the signal mask as they please.

If the non-Go code starts a new thread, changes the signal mask, and
then invokes a Go function in that thread, the Go runtime will

      setPredecessor(succ, pred);
    }
  }

  @Override
  void insertEntry(
      int entryIndex, @ParametricNullness K key, @ParametricNullness V value, int hash, int mask) {
    super.insertEntry(entryIndex, key, value, hash, mask);
    setSucceeds(lastEntry, entryIndex);
    setSucceeds(entryIndex, ENDPOINT);
  }

  @Override
  void accessEntry(int index) {
    if (accessOrder) {

	separator := byte(0x80) // gets reset to 0x00 once used
	for i := byte(0); i < chunk; i++ {
		mask := byte(int8(i-nx) >> 7) // 0x00 after the end of data

		// if we reached the end of the data, replace with 0x80 or 0x00
		d.x[i] = (^mask & separator) | (mask & d.x[i])

		// zero the separator once used
		separator &= mask

		if i >= 56 {
			// we might have to write the length here if all fit in one block

// The filename is generated by taking pattern and adding a random string to the end.
// If pattern includes a "*", the random string replaces the last "*".
// The file is created with mode 0o600 (before umask).
// If dir is the empty string, CreateTemp uses the default directory for temporary files, as returned by [TempDir].
// Multiple programs or goroutines calling CreateTemp simultaneously will not choose the same file.

	ADD	$16, OUTENC, OUTENC
	ADD	$-16, OUTDEC, OUTDEC
	VSPLTISB	$8, KEY                  // vspltisb 3,8
	MOVD	CNT, CTR                         // mtctr 7
	VSUBUBM	MASK, KEY, MASK                  // vsububm 5,5,3

loop192:
	VPERM	IN1, IN1, MASK, KEY // vperm 3,2,2,5
	VSLDOI	$12, ZERO, IN0, TMP // vsldoi 6,0,1,12
	VCIPHERLAST	KEY, RCON, KEY      // vcipherlast 3,3,4

	VXOR	IN0, TMP, IN0       // vxor 1,1,6

    ReducedPrecisionSupport mask = ReducedPrecisionSupport::None;
    if (toco_flags.quantize_to_float16()) {
      mask |= ReducedPrecisionSupport::Float16Inference;
    }
    if (toco_flags.allow_bfloat16()) {
      mask |= ReducedPrecisionSupport::Bfloat16Inference;
    }
    if (toco_flags.accumulation_type() == toco::IODataType::FLOAT16) {
      mask |= ReducedPrecisionSupport::Float16Accumulation;