namespace mlir {
namespace quant {

constexpr int k8Bits = 8;
constexpr int k32Bits = 32;
constexpr unsigned kSigned = QuantizationFlags::Signed;

DeviceTarget::DeviceTarget(MLIRContext* ctx) : ctx_(ctx) {
  f32_ = FloatType::getF32(ctx_);
  i8_ = IntegerType::get(ctx_, k8Bits);
  i8_min_ = QuantizedType::getDefaultMinimumForInteger(kSigned, k8Bits);
  i8_max_ = QuantizedType::getDefaultMaximumForInteger(kSigned, k8Bits);

  %dq_weight = "quantfork.dcast"(%q_weight) : (tensor<2x3x3x2x!quant.uniform<i8:f32, 0.074855112561992565:-1>>) -> tensor<2x3x3x2xf32>
  %q_bias = "quantfork.qcast"(%bias) : (tensor<2xf32>) -> tensor<2x!quant.uniform<i32:f32, 0.044022349891595126>>
  %dq_bias = "quantfork.dcast"(%q_bias) : (tensor<2x!quant.uniform<i32:f32, 0.044022349891595126>>) -> tensor<2xf32>

}

// writeLSB writes the code c for "Least Significant Bits first" data.
func (w *Writer) writeLSB(c uint32) error {
	w.bits |= c << w.nBits
	w.nBits += w.width
	for w.nBits >= 8 {
		if err := w.w.WriteByte(uint8(w.bits)); err != nil {
			return err
		}
		w.bits >>= 8
		w.nBits -= 8
	}
	return nil
}

// writeMSB writes the code c for "Most Significant Bits first" data.
func (w *Writer) writeMSB(c uint32) error {

		atomic.And8(v.bytep, ^mask)
	}
}

// pinnerBits is the same type as gcBits but has different methods.
type pinnerBits gcBits

// ofObject returns the pinState of the n'th object.
// nosplit, because it's called by isPinned, which is nosplit
//
//go:nosplit
func (p *pinnerBits) ofObject(n uintptr) pinState {
	bytep, mask := (*gcBits)(p).bitp(n * 2)
	byteVal := atomic.Load8(bytep)
	return pinState{bytep, byteVal, mask}

}

// readLSB returns the next code for "Least Significant Bits first" data.
func (r *Reader) readLSB() (uint16, error) {
	for r.nBits < r.width {
		x, err := r.r.ReadByte()
		if err != nil {
			return 0, err
		}
		r.bits |= uint32(x) << r.nBits
		r.nBits += 8
	}
	code := uint16(r.bits & (1<<r.width - 1))
	r.bits >>= r.width
	r.nBits -= r.width
	return code, nil
}

// bytes, and throws if it finds an unpinned Go pointer. The gcbits mark each
// pointer value. The src pointer is off bytes into the gcbits.
//
//go:nosplit
//go:nowritebarrier
func cgoCheckBits(src unsafe.Pointer, gcbits *byte, off, size uintptr) {
	skipMask := off / goarch.PtrSize / 8
	skipBytes := skipMask * goarch.PtrSize * 8
	ptrmask := addb(gcbits, skipMask)
	src = add(src, skipBytes)
	off -= skipBytes

}

// setCheckmark throws if marking object is a checkmarks violation,
// and otherwise sets obj's checkmark. It returns true if obj was
// already checkmarked.
func setCheckmark(obj, base, off uintptr, mbits markBits) bool {
	if !mbits.isMarked() {
		printlock()
		print("runtime: checkmarks found unexpected unmarked object obj=", hex(obj), "\n")
		print("runtime: found obj at *(", hex(base), "+", hex(off), ")\n")

		if s.state.get() != mSpanInUse {
			continue
		}

		// Find all allocated objects.
		abits := s.allocBitsForIndex(0)
		for i := uintptr(0); i < uintptr(s.nelems); i++ {
			if abits.index < uintptr(s.freeindex) || abits.isMarked() {
				x := s.base() + i*s.elemsize
				trace.HeapObjectExists(x, s.typePointersOfUnchecked(x).typ)
			}
			abits.advance()
		}
	}

	// Write out all the goroutine stacks.
	forEachGRace(func(gp *g) {

// RUN: tf-opt %s -tfl-prepare-quantize="quantize-signed=true post-training-quantize=true activation-number-of-bits=16" -cse | FileCheck %s

// CHECK-LABEL: QuantizeUnidirectionalLstmFullPerTensor
func.func @QuantizeUnidirectionalLstmFullPerTensor(%arg0: tensor<1x2x3xf32>) -> (tensor<1x2x3xf32>) {
  %input = "quantfork.stats"(%arg0) {layerStats = dense<[0.0, 1.0]> : tensor<2xf32>} : (tensor<1x2x3xf32>) -> tensor<1x2x3xf32>

		obj, span, objIndex := findObject(ptr, 0, 0)
		if obj == 0 {
			continue
		}
		// TODO: Consider making two passes where the first
		// just prefetches the mark bits.
		mbits := span.markBitsForIndex(objIndex)
		if mbits.isMarked() {
			continue
		}
		mbits.setMarked()

		// Mark span.
		arena, pageIdx, pageMask := pageIndexOf(span.base())
		if arena.pageMarks[pageIdx]&pageMask == 0 {