//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) {

	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.

  %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>

		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}

// 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>

      bias_shape: Optional[Sequence[int]] = None,
  ) -> module.Module:
    class EinsumModel(module.Module):
      """Einsum class."""

      def __init__(self):
        self._bias = None
        if bias_shape is not None:
          self._bias = array_ops.constant(
              np.random.uniform(size=bias_shape), dtype=dtypes.float32
          )

        self._kernel = np.random.uniform(size=y_shape).astype('f4')

            result *= n;
            result /= i;
          }
          return result;
        } else {
          int nBits = LongMath.log2(n, RoundingMode.CEILING);

          long result = 1;
          long numerator = n--;
          long denominator = 1;

          int numeratorBits = nBits;
          // This is an upper bound on log2(numerator, ceiling).

          /*

	// n must always be a multiple of 8, since gcBits is
	// always rounded up 8 bytes.
	for _, n := range []int{8, 16, 32, 64, 128} {
		b.Run(fmt.Sprintf("bits=%d", n*8), func(b *testing.B) {
			// Initialize a new byte slice with pseduo-random data.
			bits := make([]byte, n)
			rand.Read(bits)

			b.ResetTimer()
			for i := 0; i < b.N; i++ {
				runtime.MSpanCountAlloc(s, bits)
			}
		})
	}
}

		lock(&mheap_.lock)
		mheap_.spanalloc.free(unsafe.Pointer(s))
		unlock(&mheap_.lock)
	})
}

func MSpanCountAlloc(ms *MSpan, bits []byte) int {
	s := (*mspan)(ms)
	s.nelems = uint16(len(bits) * 8)
	s.gcmarkBits = (*gcBits)(unsafe.Pointer(&bits[0]))
	result := s.countAlloc()
	s.gcmarkBits = nil
	return result
}

const (
	TimeHistSubBucketBits = timeHistSubBucketBits