Arm backend: Fix CPU cycle counters over backend delegate code (#10393)

### Summary

This fixes a bug where we renamed the weak functions but not the
non-weak versions.

### Test plan
Tests by manually looking at the logs when running and detecting that cycles was
not zero.

Signed-off-by: Zingo Andersen <[email protected]>

Author: zingo

backends/arm/runtime/EthosUBackend.cpp

@@ -144,6 +144,18 @@ class EthosUBackend final : public ::executorch::runtime::BackendInterface {
 
     EXECUTORCH_PROF_SCOPE(event_tracer, "EthosUBackend::execute()");
 
+    // CollectArm_CPU_Cycles is just used to save the numbers of CPU cycles
+    // used, If etdump is used the EXECUTORCH_PROF_SCOPE() above will do the
+    // same. If not, this is a cheap way of getting some stats and the
+    // CollectArm_CPU_Cycles object can safely be removed in production code.
+    //
+    // The EthosUBackendExecuteCallbacks class uses the C++
+    // constructor/destructor to make sure that EthosUBackend_execute_begin()
+    // and EthosUBackend_execute_end() is called while CollectArm_CPU_Cycles is
+    // in scope. e.g. We meassure from now until we exit this metod (in any way
+    // we might do it).
+    EthosUBackendExecuteCallbacks CollectArm_CPU_Cycles;
+
     ExecutionHandle* execution_handle = (ExecutionHandle*)input_handle;
     VelaHandles handles;
 

examples/arm/executor_runner/arm_perf_monitor.cpp

@@ -1,4 +1,4 @@
-/* Copyright 2024 Arm Limited and/or its affiliates.
+/* Copyright 2024-2025 Arm Limited and/or its affiliates.
  *
  * This source code is licensed under the BSD-style license found in the
  * LICENSE file in the root directory of this source tree.
@@ -98,14 +98,14 @@ void ethosu_inference_end(struct ethosu_driver* drv, void*) {
 }
 
 // Callback invoked at start of ArmBackend::execute()
-void ArmBackend_execute_begin() {
+void EthosUBackend_execute_begin() {
   // Save Cortex-M cycle clock to calculate total CPU cycles used in
   // ArmBackend_execute_end()
   ethosu_ArmBackendExecuteCycleCountStart = ARM_PMU_Get_CCNTR();
 }
 
 // Callback invoked at end of ArmBackend::execute()
-void ArmBackend_execute_end() {
+void EthosUBackend_execute_end() {
   // Add Cortex-M cycle clock used during this ArmBackend::execute()
   ethosu_ArmBackendExecuteCycleCount +=
       (ARM_PMU_Get_CCNTR() - ethosu_ArmBackendExecuteCycleCountStart);