//go:nosplit
func setSignalstackSP(s *stackt, sp uintptr) {
	s.ss_sp = sp
}

//go:nosplit
//go:nowritebarrierrec
func sigaddset(mask *sigset, i int) {
	mask.__bits[(i-1)/32] |= 1 << ((uint32(i) - 1) & 31)
}

func sigdelset(mask *sigset, i int) {
	mask.__bits[(i-1)/32] &^= 1 << ((uint32(i) - 1) & 31)
}

//go:nosplit
func (c *sigctxt) fixsigcode(sig uint32) {
}

func setProcessCPUProfiler(hz int32) {

//go:nosplit
func setSignalstackSP(s *stackt, sp uintptr) {
	s.ss_sp = sp
}

//go:nosplit
//go:nowritebarrierrec
func sigaddset(mask *sigset, i int) {
	mask.__bits[(i-1)/32] |= 1 << ((uint32(i) - 1) & 31)
}

func sigdelset(mask *sigset, i int) {
	mask.__bits[(i-1)/32] &^= 1 << ((uint32(i) - 1) & 31)
}

//go:nosplit
func (c *sigctxt) fixsigcode(sig uint32) {
}

func setProcessCPUProfiler(hz int32) {

		f := &instFormats[i]
		if xNoCond&(f.mask|condMask) != f.value || f.priority <= priority {
			continue
		}
		delta := uint32(0)
		deltaShift := uint(0)
		for opBits := f.opBits; opBits != 0; opBits >>= 16 {
			n := uint(opBits & 0xFF)
			off := uint((opBits >> 8) & 0xFF)
			delta |= (x >> off) & (1<<n - 1) << deltaShift
			deltaShift += n
		}
		op := f.op + Op(delta)

	mgf1XOR(db, hash, h)

	// 11. Set the leftmost 8 * emLen - emBits bits of the leftmost octet in
	//     maskedDB to zero.

	db[0] &= 0xff >> (8*emLen - emBits)

	// 12. Let EM = maskedDB || H || 0xbc.
	em[emLen-1] = 0xbc

	// 13. Output EM.
	return em, nil
}

func emsaPSSVerify(mHash, em []byte, emBits, sLen int, hash hash.Hash) error {
	// See RFC 8017, Section 9.1.2.

}

// emit emits the least significant nBits bits of bits to the bit-stream.
// The precondition is bits < 1<<nBits && nBits <= 16.
func (e *encoder) emit(bits, nBits uint32) {
	nBits += e.nBits
	bits <<= 32 - nBits
	bits |= e.bits
	for nBits >= 8 {
		b := uint8(bits >> 24)
		e.writeByte(b)
		if b == 0xff {
			e.writeByte(0x00)
		}
		bits <<= 8
		nBits -= 8
	}
	e.bits, e.nBits = bits, nBits
}

// To rotate x right by k bits, call RotateLeft8(x, -k).
//
// This function's execution time does not depend on the inputs.
func RotateLeft8(x uint8, k int) uint8 {
	const n = 8
	s := uint(k) & (n - 1)
	return x<<s | x>>(n-s)
}

// RotateLeft16 returns the value of x rotated left by (k mod 16) bits.
// To rotate x right by k bits, call RotateLeft16(x, -k).
//

  %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>

	if w.err != nil {
		return
	}
	w.bits |= uint64(b) << w.nbits
	w.nbits += nb
	if w.nbits >= 48 {
		bits := w.bits
		w.bits >>= 48
		w.nbits -= 48
		n := w.nbytes
		bytes := w.bytes[n : n+6]
		bytes[0] = byte(bits)
		bytes[1] = byte(bits >> 8)
		bytes[2] = byte(bits >> 16)
		bytes[3] = byte(bits >> 24)
		bytes[4] = byte(bits >> 32)
		bytes[5] = byte(bits >> 40)
		n += 6
		if n >= bufferFlushSize {

		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}

				nbits -= shift
			}

			// convert any partial leading digit and advance to next word
			if nbits == 0 {
				// no partial digit remaining, just advance
				w = x[k]
				nbits = _W
			} else {
				// partial digit in current word w (== x[k-1]) and next word x[k]
				w |= x[k] << nbits
				i--
				s[i] = digits[w&mask]

				// advance
				w = x[k] >> (shift - nbits)