mheap_.sweepPagesPerByte = 0
		unlock(&mheap_.lock)
		// Flush all mcaches.
		for _, pp := range allp {
			pp.mcache.prepareForSweep()
		}
		// Sweep all spans eagerly.
		for sweepone() != ^uintptr(0) {
		}
		// Free workbufs eagerly.
		prepareFreeWorkbufs()
		for freeSomeWbufs(false) {
		}
		// All "free" events for this mark/sweep cycle have

	intRegs, floatRegs uint8 // registers needed for ABIInternal

	flags bitset8
	alg   AlgKind // valid if Align > 0

	// size of prefix of object that contains all pointers. valid if Align > 0.
	// Note that for pointers, this is always PtrSize even if the element type
	// is NotInHeap. See size.go:PtrDataSize for details.
	ptrBytes int64

// Stack formats a stack trace of the calling goroutine into buf
// and returns the number of bytes written to buf.
// If all is true, Stack formats stack traces of all other goroutines
// into buf after the trace for the current goroutine.
func Stack(buf []byte, all bool) int {
	var stw worldStop
	if all {
		stw = stopTheWorld(stwAllGoroutinesStack)
	}

	n := 0
	if len(buf) > 0 {
		gp := getg()

	{AMOVBU, C_SOREG, C_NONE, C_REG, 8, 4, REGZERO, 0, 0},
	{AMOVWL, C_SOREG, C_NONE, C_REG, 8, 4, REGZERO, 0, 0},
	{AMOVVL, C_SOREG, C_NONE, C_REG, 8, 4, REGZERO, sys.MIPS64, 0},
	{ALL, C_SOREG, C_NONE, C_REG, 8, 4, REGZERO, 0, 0},
	{ALLV, C_SOREG, C_NONE, C_REG, 8, 4, REGZERO, sys.MIPS64, 0},

	{AMOVW, C_REG, C_NONE, C_LEXT, 35, 12, REGSB, sys.MIPS64, 0},
	{AMOVWU, C_REG, C_NONE, C_LEXT, 35, 12, REGSB, sys.MIPS64, 0},

    // correctly.
    // CHECK-NOT: tf.RngReadAndSkip
    %alg = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
    %delta = "tf.Const"() {value = dense<10> : tensor<ui64>} : () -> tensor<ui64>
    %1 = "tf.RngReadAndSkip"(%resource, %alg, %delta) : (tensor<!tf_type.resource<tensor<3xi64>>>, tensor<i32>, tensor<ui64>) -> tensor<3xi64>
    tf_device.return %1 : tensor<3xi64>

	{AMOVWL, C_SOREG, C_NONE, C_NONE, C_REG, C_NONE, 8, 4, REGZERO, 0},
	{AMOVVL, C_SOREG, C_NONE, C_NONE, C_REG, C_NONE, 8, 4, REGZERO, 0},
	{ALL, C_SOREG, C_NONE, C_NONE, C_REG, C_NONE, 8, 4, REGZERO, 0},
	{ALLV, C_SOREG, C_NONE, C_NONE, C_REG, C_NONE, 8, 4, REGZERO, 0},

	{AMOVW, C_REG, C_NONE, C_NONE, C_LEXT, C_NONE, 35, 12, 0, 0},
	{AMOVWU, C_REG, C_NONE, C_NONE, C_LEXT, C_NONE, 35, 12, 0, 0},

	// maxScanWork is a worst-case estimate of the amount of scan work that
	// needs to be performed in this GC cycle. Specifically, it represents
	// the case where *all* scannable memory turns out to be live, and
	// *all* allocated stack space is scannable.
	maxStackScan := c.maxStackScan.Load()
	maxScanWork := int64(scan + maxStackScan + c.globalsScan.Load())
	if work > scanWorkExpected {

				}

				// cpuHog1 called below is the primary source of CPU
				// load, but there may be some background work by the
				// runtime. Since the OS rusage measurement will
				// include all work done by the process, also compare
				// against all samples in our profile.
				var value time.Duration
				for _, sample := range p.Sample {
					value += time.Duration(sample.Value[1]) * time.Nanosecond
				}

	}

	gcw := &getg().m.p.ptr().gcw
	gcw.bytesMarked += uint64(span.elemsize)
}

// gcMarkTinyAllocs greys all active tiny alloc blocks.
//
// The world must be stopped.
func gcMarkTinyAllocs() {
	assertWorldStopped()

	for _, p := range allp {
		c := p.mcache
		if c == nil || c.tiny == 0 {
			continue
		}
		_, span, objIndex := findObject(c.tiny, 0, 0)