#include "mlir/Pass/Pass.h"  // from @llvm-project
#include "mlir/Pass/PassManager.h"  // from @llvm-project
#include "mlir/Pass/PassRegistry.h"  // from @llvm-project
#include "mlir/Rewrite/FrozenRewritePatternSet.h"  // from @llvm-project
#include "mlir/Support/LLVM.h"  // from @llvm-project
#include "mlir/Support/LogicalResult.h"  // from @llvm-project
#include "mlir/Support/TypeID.h"  // from @llvm-project

        RankedTensorType::get({cur_rank}, rewriter.getIntegerType(64));
    auto new_shape_const_attr =
        DenseElementsAttr::get(shape_spec_type, new_shape.getShape());
    rewriter.setInsertionPointAfter(weight_op);
    auto new_shape_const = rewriter.create<arith::ConstantOp>(
        weight_op->getLoc(), shape_spec_type, new_shape_const_attr);
    auto reshape_op = rewriter.create<TF::ReshapeOp>(

}

// resetWithControl resets b and adds control v.
// It is equivalent to b.Reset(kind); b.AddControl(v),
// except that it is one call instead of two and avoids a bounds check.
// It is intended for use by rewrite rules, where this matters.
func (b *Block) resetWithControl(kind BlockKind, v *Value) {
	b.Kind = kind
	b.ResetControls()
	b.Aux = nil
	b.AuxInt = 0
	b.Controls[0] = v
	v.Uses++
}

        test_name = name + "_test"
        test_file = test_name + ".cc"

        template_file = "//tensorflow/compiler/aot:test"
        template_file += if_oss("", "_google") + ".cc"

        # Rule to rewrite the template_file to produce the test_file.
        native.genrule(
            name = ("gen_" + test_name),
            testonly = 1,
            srcs = [
                template_file,
                header_file,

        // The op is disabled by default. Otherwise, values will be saved
        // during calibration.
        rewriter.getNamedAttr("enabled", rewriter.getBoolAttr(enabled)),
        rewriter.getNamedAttr("func_name", rewriter.getStringAttr(func_name)),
        rewriter.getNamedAttr("node_name", rewriter.getStringAttr(node_name)),
    };
    return dump_attributes;
  }

  StringAttr DuplicateFunction(Operation *op,

	}

	if n.Op() == ir.OCALLFUNC && n.Fun.Op() == ir.OCLOSURE {
		directClosureCall(n)
	}

	if ir.IsFuncPCIntrinsic(n) {
		// For internal/abi.FuncPCABIxxx(fn), if fn is a defined function, rewrite
		// it to the address of the function of the ABI fn is defined.
		name := n.Fun.(*ir.Name).Sym().Name
		arg := n.Args[0]
		var wantABI obj.ABI
		switch name {
		case "FuncPCABI0":
			wantABI = obj.ABI0

go mod tidy
# pin expanded versions
ensure_require_replace_directives_for_all_dependencies
# group require/replace directives
group_directives

# Phase 4: copy root go.mod to staging dirs and rewrite

kube::log::status "go.mod: propagate to staging modules" >&11
for repo in $(kube::util::list_staging_repos); do
  (
    cd "staging/src/k8s.io/${repo}"

    echo "=== propagating to ${repo}"

<p>
Note that if there are no read-write or write-write data races on memory location <i>x</i>,
then any read <i>r</i> on <i>x</i> has only one possible <i>W</i>(<i>r</i>):
the single <i>w</i> that immediately precedes it in the happens before order.
</p>

<p>
More generally, it can be shown that any Go program that is data-race-free,
meaning it has no program executions with read-write or write-write data races,

    const Type quantized_type =
        quant_type.castFromExpressedType(expressed_type);

    rewriter.setInsertionPointAfter(op);
    auto q = rewriter.create<quantfork::QuantizeCastOp>(
        op->getLoc(), quantized_type, op->getResult(0));
    auto dq = rewriter.create<quantfork::DequantizeCastOp>(op->getLoc(),
                                                           expressed_type, q);

    TF_RETURN_IF_ERROR(root.status());

    // We already have a TensorFlow function call into the cluster -- the
    // original node we set out to rewrite.  We just wire in the correct control
    // deps and we're done.
    RemoveAllIncomingControlEdges(g, n);
    Operation inverse_predicate_as_control =
        DataToControl(root, inverse_predicated_compilation_key);