return
	}

	if debug {
		// obj must be one of lhs
		found := false
		for _, lhs := range lhs {
			if obj == lhs {
				found = true
				break
			}
		}
		if !found {
			panic("inconsistent lhs")
		}
	}

	// We have multiple variables on the lhs and one init expr.
	// Make sure all variables have been given the same type if

				check.declare(check.scope, name, lhs[i], scopePos)
			}

		case varDecl:
			top := len(check.delayed)

			lhs0 := make([]*Var, len(d.spec.Names))
			for i, name := range d.spec.Names {
				lhs0[i] = NewVar(name.Pos(), pkg, name.Name, nil)
			}

			// initialize all variables
			for i, obj := range lhs0 {
				var lhs []*Var
				var init ast.Expr
				switch len(d.spec.Values) {

	// If the lhs doesn't have a type yet, use the type of x.
	if lhs.typ == nil {
		lhs.typ = x.typ
	}

	check.assignment(x, lhs.typ, "constant declaration")
	if x.mode == invalid {
		return
	}

	lhs.val = x.val
}

// initVar checks the initialization lhs = x in a variable declaration.
// If lhs doesn't have a type yet, it is given the type of x,
// or Typ[Invalid] in case of an error.

	// If the lhs doesn't have a type yet, use the type of x.
	if lhs.typ == nil {
		lhs.typ = x.typ
	}

	check.assignment(x, lhs.typ, "constant declaration")
	if x.mode == invalid {
		return
	}

	lhs.val = x.val
}

// initVar checks the initialization lhs = x in a variable declaration.
// If lhs doesn't have a type yet, it is given the type of x,
// or Typ[Invalid] in case of an error.

def LiftDotGeneralWithBiasSameShape : Pat<
  (StableHLO_AddOp:$res
    (StableHLO_DotGeneralOp
        $lhs, $rhs, $dot_dimension_numbers, $precision_config),
    $bias),
  (LiftAsTFXlaCallModule<"composite_dot_general_with_bias_same_shape_fn">
    (ArgumentList $lhs, $rhs, $bias),
    (ResultList $res),
    (NamedAttributeList
      (NamedAttr<"dot_dimension_numbers"> $dot_dimension_numbers),

// RUN: stablehlo-quant-opt -optimize-int-graph -split-input-file %s -verify-diagnostics | FileCheck %s

// CHECK-LABEL: func @convolution_add_add
func.func @convolution_add_add(
    %lhs: tensor<?x3x2x1xi8>, %rhs: tensor<2x1x1x1xi8>,
    %zp_offset: tensor<?x2x2x1xi32>, %bias: tensor<1xi32>
  ) -> tensor<?x2x2x1xi32> {
  // CHECK-DAG: %[[conv:.*]] = mhlo.convolution
  // CHECK-DAG: %[[combined:.*]] = chlo.broadcast_add %[[zp_offset:.*]], %[[bias:.*]]

		case token.ASSIGN, token.DEFINE:
			if len(s.Lhs) == 0 {
				check.error(s, InvalidSyntaxTree, "missing lhs in assignment")
				return
			}
			if s.Tok == token.DEFINE {
				check.shortVarDecl(inNode(s, s.TokPos), s.Lhs, s.Rhs)
			} else {
				// regular assignment
				check.assignVars(s.Lhs, s.Rhs)
			}

		default:
			// assignment operations
			if len(s.Lhs) != 1 || len(s.Rhs) != 1 {

		SchemaProps: spec.SchemaProps{Type: []string{"object"}},
	}
)

func TestEquality(t *testing.T) {
	cases := []struct {
		name  string
		lhs   ref.Val
		rhs   ref.Val
		equal bool
	}{
		{
			name: "map lists are equal regardless of order",
			lhs: UnstructuredToVal([]interface{}{
				map[string]interface{}{
					"key": "a",
					"val": 1,
				},
				map[string]interface{}{
					"key": "b",

        op, "unary einsum equation does not require transpose");
  } else if (needs_reduce_sum && !needs_transpose) {
    rewriter.replaceOp(op, lhs);
    return success();
  }

  lhs = createTransposeOp(lhs, lhs.getLoc(), lhs_transpose, &rewriter);
  rewriter.replaceOp(op, lhs);
  return success();
}

std::vector<int64_t> inverseTransposeVector(
    llvm::ArrayRef<int64_t> input, llvm::ArrayRef<int32_t> permutation) {

	// so that the conversion below happens).

	checkLHS := func(i int, typ *types.Type) {
		if n := lhs[i]; typ != nil && ir.DeclaredBy(n, stmt) && n.Type() == nil {
			base.Assertf(typ.Kind() == types.TNIL, "unexpected untyped nil")
			n.SetType(defaultType(typ))
		}
		if lhs[i].Typecheck() == 0 {
			lhs[i] = AssignExpr(lhs[i])
		}
		checkassign(lhs[i])
	}

	assignType := func(i int, typ *types.Type) {
		checkLHS(i, typ)