count *= shape.dimensions(dim);
    }
  }
  rewrites->push_back({"{{I}}", absl::StrCat(i)});
  rewrites->push_back({"{{TYPE}}", type});
  rewrites->push_back({"{{DIM_VARS}}", absl::StrJoin(dim_vars, ", ")});
  rewrites->push_back({"{{DIM_SIZES}}", dim_sizes});
  rewrites->push_back({"{{INDICES}}", indices});
  rewrites->push_back({"{{COUNT}}", absl::StrCat(count)});
  return absl::OkStatus();
}

  // TODO: b/230572023 - Consider improving shape inference for While op instead
  // of dropping the attribute. This need not be correct for models not trained
  // on TPU.

  // Optimizes TF graph via cleanups, merges, rewrites, constant folding,
  // and edge case handling where possible.
  pm.addNestedPass<func::FuncOp>(TF::CreateDropWhileShapeInvariantPass());
  pm.addNestedPass<func::FuncOp>(
      tf_executor::CreateTFExecutorGraphPruningPass());

                              PatternRewriter& rewriter) {
  return rewriter.create<TransposeOp>(
      input.getLoc(), input, rewriter.getDenseI64ArrayAttr(permutation));
}

// Defers the transpose of the left-hand side (LHS) to the right-hand side and
// the result of a binary operation. In detail, this rewrites the
// `op(transpose(%rhs), %lhs)` to `transpose(op(%rhs, transpose(%lhs)))`. The

#include "tensorflow/compiler/mlir/tensorflow/transforms/tf_passes.h.inc"

// Rewrites RecvTPUEmbeddingActivationsOp and SendTPUEmbeddingGradients ops to
// internal variants by introducing XlaRecvTPUEmbeddingDeduplicationData op.
struct RewriteTPUEmbeddingOps
    : public impl::RewriteTPUEmbeddingOpsPassBase<RewriteTPUEmbeddingOps> {
  void runOnOperation() override;
};

// Rewrites the given op to `OpT` op after adding the given operand at the end.

  LogicalResult matchAndRewrite(FakeQuantOp op,
                                PatternRewriter &rewriter) const override {
    // TODO: If this pattern comes up more frequently, consider adding core
    // support for failable rewrites.
    if (failableRewrite(op, rewriter)) {
      *hadFailure = true;
      return failure();
    }
    return success();
  }

 private:
  bool *hadFailure;

 */
package okhttp3

import java.io.IOException
import okhttp3.Interceptor.Chain
import okio.Buffer
import okio.BufferedSource
import okio.ForwardingSource
import okio.buffer

/** Rewrites the response body returned from the server to be all uppercase.  */
class UppercaseResponseInterceptor : Interceptor {
  @Throws(IOException::class)
  override fun intercept(chain: Chain): Response {

 */
package okhttp3

import java.io.IOException
import okhttp3.Interceptor.Chain
import okio.Buffer
import okio.BufferedSink
import okio.ForwardingSink
import okio.Sink
import okio.buffer

/** Rewrites the request body sent to the server to be all uppercase.  */
class UppercaseRequestInterceptor : Interceptor {
  @Throws(IOException::class)
  override fun intercept(chain: Chain): Response {

	"os"
	"path/filepath"
	"sort"
	"strings"

	"cmd/internal/telemetry"
)

var (
	fset     = token.NewFileSet()
	exitCode = 0
)

var allowedRewrites = flag.String("r", "",
	"restrict the rewrites to this comma-separated list")

var forceRewrites = flag.String("force", "",
	"force these fixes to run even if the code looks updated")

var allowed, force map[string]bool

var (

  let dependentDialects = ["tf_device::TensorFlowDeviceDialect"];
}

def XlaRewritePass : Pass<"tf-xla-rewrite", "mlir::ModuleOp"> {
  let summary = "Rewrites partition calls into Xla launch ops to make the attached function run on XLA.";

  let description = [{
    This pass rewrites `tf.PartitionedCall` and `tf.StatefulPartitionedCall`
    operations with `_xla_compile_device_type` attribute in a

#include <memory>

#include "llvm/ADT/StringRef.h"
#include "mlir/IR/BuiltinOps.h"  // from @llvm-project
#include "mlir/Pass/Pass.h"  // from @llvm-project

namespace mlir {
namespace TFTPU {

// Creates a pass that rewrites `tf_device.launch_func` on TPUs into TPU runtime
// ops.
std::unique_ptr<mlir::OperationPass<mlir::ModuleOp>> CreateTPURewritePass(
    llvm::StringRef module_name = llvm::StringRef());