// This method attempts to match a pattern and rewrite it. The rewriter
  // argument is the orchestrator of the sequence of rewrites. The pattern is
  // expected to interact with it to perform any changes to the IR from here.
  mlir::LogicalResult matchAndRewrite(
      mlir::tfd::ConvertTftToDhtOp op,
      mlir::PatternRewriter& rewriter) const override {
    // Look through the inputs of the ConvertTftToDhtOp.

};

// Quantization rewrite pattern using DQ as the root op.
struct StableHloQuantization
    : public StableHloQuantizationBase<StableHloQuantization> {
  explicit StableHloQuantization(MLIRContext* ctx)
      : StableHloQuantizationBase<StableHloQuantization>(ctx) {}
};

// Quantization rewrite pattern using Q as the root op. This is for the

  region.front().eraseArguments(0, region.getNumArguments());
}

// Rewrites an `mhlo.if` op or its region. If `region_idx` is not set, the op
// operands and results are rewritten. If `region_idx` is set, region
// `region_idx` is rewritten to take in and return an additional token. Returns
// true if the op or its region was rewritten.
bool ProcessRegionIfOp(OpBuilder& builder, IfOp region_if,

				a = append(a, s[:i])
				s = s[i+1:]
				i = 0
				continue
			}
		}
		i++
	}
	return append(a, s)
}

// rewrite rewrites a subname to having only printable characters and no white
// space.
// From go/src/testing/match.go
func rewrite(s string) string {
	b := []byte{}
	for _, r := range s {
		switch {
		case isSpace(r):
			b = append(b, '_')
		case !strconv.IsPrint(r):

// some generic types that are legal in MHLO. This pass legalizes TF types into
// types that are legal in MHLO. For example, TF::Qint8Type is converted to i8.
// Rewrites here should run before TF to MHLO op legalizations are run.

#include <memory>

#include "absl/log/log.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/Support/Casting.h"

 See the License for the specific language governing permissions and
 limitations under the License.
 ==============================================================================*/
// Rewrites computations generated by the xla.compile() Python code into
// XlaLaunch nodes.
//
// xla.compile() does two main things:
// a) marks operators that make up an XLA computation with the attribute

		t.Errorf("line number missing in assembly:\n%s", out)
	}
}

var issue16214src = `
package main

func Mod32(x uint32) uint32 {
	return x % 3 // frontend rewrites it as HMUL with 2863311531, the LITERAL node has unknown Pos
}

    const std::map<string, int>& host_compute_core, FunctionLibraryRuntime* flr,
    FunctionLibraryDefinition* fld, std::vector<string>* shape_inference_graphs,
    bool* has_outside_compilation);

// Rewrites XLA computation in `clusters` to replace outside compilation nodes
// with XlaHostCompute, and moves those outside compilations into `g`. If shapes
// of outside compilation outputs cannot be determined now, we will store shape

    auto filter_transpose_op = rewriter.create<mlir::stablehlo::TransposeOp>(
        op.getLoc(), /*resultType0=*/new_filter_tensor_type, /*operand=*/filter,
        rewriter.getDenseI64ArrayAttr(kOihwToHwioPermutation));

    // [b, 0, 1, f]x[0, 1, i, o]->[b, 0, 1, f]
    const auto new_dimension_nums = rewriter.getAttr<ConvDimensionNumbersAttr>(

  // graph rewrites. Returns false if the inference_type is DT_FLOAT or
  // `weight_quantization` flag is set.
  bool RunPropagationAndRewriteQuantizationPasses() const {
    return inference_type != tensorflow::DT_FLOAT && !weight_quantization;
  }

  // TODO: b/202075505 - make implicit weight type clearer
  // Whether run the passes and graph rewrites for dynamic range quantization.