Normalize FLOPs with DVFS in op_profile analysis of XProf

PiperOrigin-RevId: 911448273

Author: Profiler Team

xprof/convert/BUILD

@@ -707,6 +707,16 @@ cc_library(
     ],
 )
 
+cc_test(
+    name = "op_metrics_to_record_test",
+    srcs = ["op_metrics_to_record_test.cc"],
+    deps = [
+        ":op_metrics_to_record",
+        "@com_google_googletest//:gtest_main",
+        "@org_xprof//plugin/xprof/protobuf:op_metrics_proto_cc",
+    ],
+)
+
 cc_library(
     name = "op_stack",
     hdrs = ["op_stack.h"],

xprof/convert/op_metrics_to_record.h

@@ -51,8 +51,8 @@ inline double GigaFlopsPerSecondPerCoreNormalizedOnDvfs(
     return GigaFlopsPerSecondPerCore(metrics);
   }
   return GigaFlopsPerSecondPerCore(metrics) *
-         (tsl::profiler::SafeDivide(metrics.normalized_time_ps(),
-                                    metrics.time_ps()));
+         (tsl::profiler::SafeDivide(metrics.time_ps(),
+                                    metrics.normalized_time_ps()));
 }
 
 inline double GigaModelFlopsPerSecondPerCore(const OpMetrics& metrics) {

xprof/convert/op_metrics_to_record_test.cc

@@ -0,0 +1,56 @@
+/* Copyright 2026 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+
+#include "xprof/convert/op_metrics_to_record.h"
+
+#include "<gtest/gtest.h>"
+#include "plugin/xprof/protobuf/op_metrics.pb.h"
+
+namespace tensorflow {
+namespace profiler {
+namespace {
+
+constexpr double kMaxError = 1E-10;
+
+TEST(OpMetricsToRecordTest, GigaFlopsPerSecondPerCoreNormalizedOnDvfs) {
+  OpMetrics metrics;
+  metrics.set_time_ps(100);
+  metrics.set_normalized_time_ps(200);
+  metrics.set_flops_v2(1000);
+  metrics.set_occurrences(1);
+  metrics.set_num_cores(1);
+
+  // GigaFlopsPerSecondPerCore = (flops_v2 / (time_ps / 1000.0)) = 1000 / 0.1 =
+  // 10000. Multiplier = time_ps / normalized_time_ps = 100 / 200 = 0.5.
+  // Expected normalized GFLOPS = 10000 * 0.5 = 5000.
+  EXPECT_NEAR(5000.0, GigaFlopsPerSecondPerCoreNormalizedOnDvfs(metrics),
+              kMaxError);
+}
+
+TEST(OpMetricsToRecordTest, GigaFlopsPerSecondPerCoreNormalizedOnDvfsFallback) {
+  OpMetrics metrics;
+  metrics.set_time_ps(100);
+  metrics.set_normalized_time_ps(0);
+  metrics.set_flops_v2(1000);
+  metrics.set_occurrences(1);
+  metrics.set_num_cores(1);
+
+  EXPECT_NEAR(10000.0, GigaFlopsPerSecondPerCoreNormalizedOnDvfs(metrics),
+              kMaxError);
+}
+
+}  // namespace
+}  // namespace profiler
+}  // namespace tensorflow

xprof/convert/op_profile_builder.cc

@@ -203,9 +203,9 @@ void PopulateOpMetricsNode(
   metrics->set_avg_time_ps(tsl::profiler::SafeDivide(op_metrics.time_ps(),
                                                      op_metrics.occurrences()));
 
-  double uncapped_flops_utilization =
-      tsl::profiler::SafeDivide(GigaFlopsPerSecondPerCore(op_metrics),
-                                peak_gigaflops_per_second_per_core);
+  double uncapped_flops_utilization = tsl::profiler::SafeDivide(
+      GigaFlopsPerSecondPerCoreNormalizedOnDvfs(op_metrics),
+      peak_gigaflops_per_second_per_core);
 
   double flops_utilization = CapUtilization(uncapped_flops_utilization);
   // The UI expects flops_utilization = flop_util / time_fraction. See: