}

	return sm, nil
}

func getLatency(promAPI promv1.API, workloadName, workloadNamespace string, duration time.Duration, quantile float64) (time.Duration, error) {
	latencyQuery := fmt.Sprintf(`histogram_quantile(%f, sum(rate(%s_bucket{%s=~"%s.*", %s=~"%s.*",reporter="destination"}[%s])) by (le))`,
		quantile, reqDur, destWorkloadLabel, workloadName, destWorkloadNamespaceLabel, workloadNamespace, duration)

                     "intervalFactor": 2,
                     "legendFormat": "{{cluster}} - {{namespace}} - {{name}} - mean\n"
                  }
               ],
               "title": "Workqueue Waiting Duration Quantile",
               "type": "timeseries"
            },
            {
               "datasource": {
                  "type": "datasource",
                  "uid": "-- Mixed --"
               },

	// distribution quantile is less than the next worst-case mean
	// mutator utilization. At this point, all further
	// contributions to the distribution must be beyond the
	// desired quantile and hence cannot affect it.
	//
	// First, find the highest desired distribution quantile.
	maxQ := quantiles[0]
	for _, q := range quantiles {
		if q > maxQ {
			maxQ = q
		}
	}

// RUN: tf-opt %s -tfl-prepare-quantize -tfl-quantize  | FileCheck %s
// RUN: tf-opt %s -tfl-quantize="legacy-quantize=true" | FileCheck --check-prefix=LEGACY %s
// RUN: tf-opt %s -tfl-prepare-quantize -tfl-quantize="ops-blocklist=tfl.fully_connected,tfl.softmax locs-blocklist=Block,NullBlock" | FileCheck --check-prefix=BLOCK %s

// CHECK-LABEL: QuantizeFloatConst
func.func @QuantizeFloatConst() -> tensor<2x2x!quant.uniform<u8:f32, 7.8431372549019615E-4:128>> {

// RUN: tf-opt %s -tfl-prepare-quantize="quantize-allowlist=quantize_float_placeholder_only,not_reset_input" | FileCheck %s
// RUN: tf-opt %s -tfl-prepare-quantize="disable-set-input-nodes-quantization-params=true" | FileCheck --check-prefix=MixedPrecision %s
// RUN: tf-opt %s -tfl-prepare-quantize="is-qdq-conversion=true" | FileCheck --check-prefix=QDQ %s

// CHECK-LABEL: main
// Uses `main` function to match the default target function of QuantSpecs and

// RUN: stablehlo-quant-opt %s -split-input-file -stablehlo-quantize -verify-each=false | FileCheck %s

// Tests for PopulateFusedGemmStylePatterns are handled in
// quantize_composite_functions for module-level evaluation of functions.

module attributes {tf_saved_model.semantics} {
// CHECK: quantize_simple_xla_call_module(%[[ARG_0:.+]]: tensor<1x4xf32>)
  func.func private @quantize_simple_xla_call_module(%arg0: tensor<1x4xf32>) -> tensor<1x3xf32> {

// RUN: tf-opt %s -tfl-quantize-variables | FileCheck %s
// RUN: tf-opt %s -tfl-prepare-quantize -tfl-quantize -tfl-post-quantize -tfl-quantize-variables -tfl-quantize -tfl-post-quantize | FileCheck --check-prefix=WHOLE-PASSES %s

// CHECK-LABEL: QuantizeReadVariable
func.func @QuantizeReadVariable() -> (tensor<1x2x1x3x!quant.uniform<i8:f32, 1.0>>) {
  %1 = "tfl.var_handle"() : () -> tensor<!tf_type.resource>

// RUN: tf-opt %s -tfl-post-quantize | FileCheck %s
// RUN: tf-opt %s -tfl-post-quantize-remove-qdq | FileCheck --check-prefix=QDQ %s

// CHECK-LABEL: RemoveUnused
// QDQ-LABEL: RemoveUnused
func.func @RemoveUnused(%arg0: tensor<4xf32>, %arg1: tensor<i32>) -> (tensor<2xf32>,tensor<2xf32>) {
  %0 = "tfl.quantize"(%arg0) {qtype = tensor<4x!quant.uniform<u8:f32, 1.0>>} : (tensor<4xf32>) -> tensor<4x!quant.uniform<u8:f32, 1.0>>

#include "tensorflow/compiler/mlir/quantization/common/quantization_lib/quantization_utils.h"

namespace mlir {
namespace TFL {

//===----------------------------------------------------------------------===//
// The actual Quantize Pass.
//===----------------------------------------------------------------------===//
namespace {
#define GEN_PASS_DEF_QUANTIZEPASS
#include "tensorflow/compiler/mlir/lite/transforms/passes.h.inc"

// RUN: tf-opt %s -tfl-prepare-quantize="quantize-signed=true" | FileCheck %s
// RUN: tf-opt %s -tfl-prepare-quantize="quantize-signed=true disable-per-channel=true" | FileCheck --check-prefix=PerTensor %s

// CHECK-LABEL: uint8_to_int8
func.func @uint8_to_int8(%arg0: tensor<2x2xf32>) -> tensor<2x2xf32> {
  %1 = "tfl.quantize"(%arg0) {qtype = tensor<2x2x!quant.uniform<u8:f32, 1.0:128>>} : (tensor<2x2xf32>) -> tensor<2x2x!quant.uniform<u8:f32, 1.0:128>>