// either 16, 24, or 32 bytes to select
// AES-128, AES-192, or AES-256.
func NewCipher(key []byte) (cipher.Block, error) {
	k := len(key)
	switch k {
	default:
		return nil, KeySizeError(k)
	case 16, 24, 32:
		break
	}
	if boring.Enabled {
		return boring.NewAESCipher(key)
	}
	return newCipher(key)
}

// newCipherGeneric creates and returns a new cipher.Block
// implemented in pure Go.

}

// CHECK-LABEL: QuantizeStridedSlice
func.func @QuantizeStridedSlice(tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>, tensor<1xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<1x2x2x5xf32> {
^bb0(%arg0: tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>, %arg1: tensor<1xi32>, %arg2: tensor<1xi32>, %arg3: tensor<1xi32>):
  %0 = "tfl.dequantize"(%arg0) : (tensor<12x2x2x5x!quant.uniform<u8:f32, 0.1>>) -> tensor<12x2x2x5xf32>

    %8 = stablehlo.convert %7 : (tensor<1x2xi32>) -> tensor<1x2xf32>
    %9 = stablehlo.convert %2 : (tensor<2x3xi8>) -> tensor<2x3xf32>
    %10 = stablehlo.dot_general %8, %9, contracting_dims = [1] x [0] : (tensor<1x2xf32>, tensor<2x3xf32>) -> tensor<1x3xf32>
    %11 = stablehlo.convert %3 : (tensor<1x3xi32>) -> tensor<1x3xf32>

    ```mlir
      %2 = "tf.A"(%arg0) : (tensor<?xi32>) -> tensor<?xi32>
      %3 = "tf.B"(%2) {device = "tpu0"} : (tensor<?xi32>) -> tensor<?xi32>
      %4 = "tf.C"(%2, %3) {device = "tpu0"} : (tensor<?xi32>, tensor<?xi32>) -> tensor<?xi32>
      %5 = "tf.D"(%4) : (tensor<?xi32>) -> tensor<?xi32>
    ```

    After the pass, we will have:

    ```mlir
      %0 = "tf.A"(%arg0) : (tensor<?xi32>) -> tensor<?xi32>

    %0 = "tfl.pseudo_const"() {value = dense<0.962260901> : tensor<1xf32>} : () -> tensor<1xf32>
    %1 = func.call @func_0_GPU_FLOAT(%arg0, %0) {tac.device = "GPU", tac.inference_type = "FLOAT", tac.interface_name = "func_0"} : (tensor<1x200x200x200xf32>, tensor<1xf32>) -> tensor<1x200x200x200xf32>
    %2 = "tfl.pseudo_const"() {value = dense<0.895973444> : tensor<1xf32>} : () -> tensor<1xf32>

// CHECK-DAG: %[[BROADCAST_DIM:.+]] = arith.constant dense<{{\[3, 2, 1, 1\]}}> : tensor<4xi32>
// CHECK-DAG: %[[EXPAND_DIM1:.+]] = arith.constant dense<3> : tensor<1xi32>
// CHECK-DAG: %[[EXPAND_DIM0:.+]] = arith.constant dense<2> : tensor<1xi32>
// CHECK: %[[EXPAND0:.+]] = "tfl.expand_dims"(%[[ARG0]], %[[EXPAND_DIM0]]) : (tensor<1x2x!quant.uniform<i8:f32, 2.000000e+00:3>>, tensor<1xi32>) -> tensor<1x2x1x!quant.uniform<i8:f32, 2.000000e+00:3>>

		// bits after the first one bit.
		top := uint(x[i])
		top |= top >> 1
		top |= top >> 2
		top |= top >> 4
		top |= top >> 8
		top |= top >> 16
		top |= top >> 16 >> 16 // ">> 32" doesn't compile on 32-bit architectures
		return i*_W + bits.Len(top)
	}
	return 0
}

// trailingZeroBits returns the number of consecutive least significant zero
// bits of x.

DATA ·IEEEConst+1544(SB)/8,$0x0000000090db8c44

	/* x^161856 mod p(x), x^161792 mod p(x) */
DATA ·IEEEConst+1552(SB)/8,$0x0000000067a2c786
DATA ·IEEEConst+1560(SB)/8,$0x000000010010a4ce

	/* x^160832 mod p(x), x^160768 mod p(x) */
DATA ·IEEEConst+1568(SB)/8,$0x0000000048b9496c
DATA ·IEEEConst+1576(SB)/8,$0x00000001c8f4c72c

	/* x^159808 mod p(x), x^159744 mod p(x) */
DATA ·IEEEConst+1584(SB)/8,$0x000000015a422de6

  // The first Operand is assumed to be a TensorType around a variant with a
  // single subtype (e.g. tensor<!tf_type.variant<tensor<2xi32>>>). We will
  // copy this type to the first result, and copy the singular variant subtype
  // to the second result (tensor<2xi32>).
  DCOMMENT_OP(op, "Inferring shape for TensorListPopBackOp.");

  auto src_list_handle_t =

from tensorflow.compiler.mlir.tensorflow.gen_mlir_passthrough_op import mlir_passthrough_op

mlir_module = '''python
func @main(%arg0 : tensor<10xf32>, %arg1 : tensor<10xf32>) -> tensor<10x10xf32> {
   %add = "magic.op"(%arg0, %arg1) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10x10xf32>
   return %ret : tensor<10x10xf32>
}
'''

@tf.function
def foo(x, y):