Changed power parameter in amdsmi_get_energy_count() to energy_accumulator

Issue linked here: https://github.com/ROCm/amdsmi/issues/38

Signed-off-by: gabrpham <Gabriel.Pham@amd.com>
Change-Id: I622236eb3f0144aefeb6c82d2713b4822bfeeb11

CHANGELOG.md

@@ -8,7 +8,10 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/pr
 
 ### Changes
 
-- **Added Pytest functionality to test amdsmi API calls in Python**.
+- **Added Pytest functionality to test amdsmi API calls in Python**.  
+
+- **Changed the `power` parameter in `amdsmi_get_energy_count()` to `energy_accumulator`**.  
+Changes propagate forwards into the python interface as well, however we are maintaing backwards compatibility and keeping the `power` field in the python API until ROCm 6.4.
 
 ### Removals
 
@@ -28,7 +31,7 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/pr
 
 ### Upcoming changes
 
-- N/A
+- **Python API for `amdsmi_get_energy_count()` will deprecate the `power` field in ROCm 6.4 and use `energy_accumulator` field instead**.  
 
 ## amd_smi_lib for ROCm 6.2.1
 
@@ -46,9 +49,9 @@ Guest VMs can view enabled/disabled ras features that are on Host cards.
 
 ### Fixes
 
-- **Fixed TypeError in `amd-smi process -G`**.
+- **Fixed TypeError in `amd-smi process -G`**.  
 
-- **Updated CLI error strings to handle empty and invalid GPU/CPU inputs**.
+- **Updated CLI error strings to handle empty and invalid GPU/CPU inputs**.  
 
 - **Fixed Guest VM showing passthrough options**.  
 

amdsmi_cli/amdsmi_commands.py

@@ -1884,7 +1884,7 @@ class AMDSMICommands():
                 try:
                     energy_dict = amdsmi_interface.amdsmi_get_energy_count(args.gpu)
 
-                    energy = energy_dict['power'] * round(energy_dict['counter_resolution'], 1)
+                    energy = round(energy_dict["energy_accumulator"] * energy_dict["counter_resolution"], 3)
                     energy /= 1000000
                     energy = round(energy, 3)
 

docs/how-to/using-amdsmi-for-python.md

@@ -1534,7 +1534,8 @@ except AmdSmiException as e:
 
 ### amdsmi_get_energy_count
 
-Description: Get the energy accumulator counter of the device.
+Description: Get the energy accumulator counter information of the device.
+energy_accumulator * counter_resolution = total_energy_consumption in micro-Joules
 It is not supported on virtual machine guest
 
 Input parameters:
@@ -1545,7 +1546,8 @@ Output: Dictionary with fields
 
 Field | Content
 ---|---
-`power` |  power
+`power` |  counter for energy accumulation since last restart/gpu rest (Deprecating in 6.4)
+`energy_accumulator` |  counter for energy accumulation since last restart/gpu rest
 `counter_resolution` |  counter resolution
 `timestamp` |  timestamp
 
@@ -1564,8 +1566,8 @@ try:
         print("No GPUs on machine")
     else:
         for device in devices:
-            power = amdsmi_get_energy_count(device)
-            print(power)
+            energy_dict = amdsmi_get_energy_count(device)
+            print(energy_dict)
 except AmdSmiException as e:
     print(e)
 ```

include/amd_smi/amdsmi.h

@@ -2382,14 +2382,13 @@ amdsmi_status_t amdsmi_set_gpu_pci_bandwidth(amdsmi_processor_handle processor_h
  *  @p power, and a pointer to a uint64_t @p timestamp, this function will write
  *  amount of energy consumed to the uint64_t pointed to by @p power,
  *  and the timestamp to the uint64_t pointed to by @p timestamp.
- *  The amdsmi_get_power_ave() is an average of a short time. This function
- *  accumulates all energy consumed.
+ *  This function accumulates all energy consumed.
  *
  *  @param[in] processor_handle a processor handle
  *  @param[in,out] counter_resolution resolution of the counter @p power in
  *  micro Joules
  *
- *  @param[in,out] power a pointer to uint64_t to which the energy
+ *  @param[in,out] energy_accumulator a pointer to uint64_t to which the energy
  *  counter will be written
  *  If this parameter is nullptr, this function will return
  *  ::AMDSMI_STATUS_INVAL if the function is supported with the provided,
@@ -2402,7 +2401,7 @@ amdsmi_status_t amdsmi_set_gpu_pci_bandwidth(amdsmi_processor_handle processor_h
  *  @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
  */
 amdsmi_status_t
-amdsmi_get_energy_count(amdsmi_processor_handle processor_handle, uint64_t *power,
+amdsmi_get_energy_count(amdsmi_processor_handle processor_handle, uint64_t *energy_accumulator,
                           float *counter_resolution, uint64_t *timestamp);
 
 /** @} End PowerQuer */

py-interface/README.md

@@ -1534,7 +1534,8 @@ except AmdSmiException as e:
 
 ### amdsmi_get_energy_count
 
-Description: Get the energy accumulator counter of the device.
+Description: Get the energy accumulator counter information of the device.
+energy_accumulator * counter_resolution = total_energy_consumption in micro-Joules
 It is not supported on virtual machine guest
 
 Input parameters:
@@ -1545,7 +1546,8 @@ Output: Dictionary with fields
 
 Field | Content
 ---|---
-`power` |  power
+`power` |  counter for energy accumulation since last restart/gpu rest (Deprecating in 6.4)
+`energy_accumulator` |  counter for energy accumulation since last restart/gpu rest
 `counter_resolution` |  counter resolution
 `timestamp` |  timestamp
 
@@ -1564,8 +1566,8 @@ try:
         print("No GPUs on machine")
     else:
         for device in devices:
-            power = amdsmi_get_energy_count(device)
-            print(power)
+            energy_dict = amdsmi_get_energy_count(device)
+            print(energy_dict)
 except AmdSmiException as e:
     print(e)
 ```
@@ -1608,7 +1610,7 @@ except AmdSmiException as e:
 
 ### amdsmi_set_gpu_od_clk_info
 
-Description: This function sets the clock frequency information
+Description: This function sets the clock frequency information.
 It is not supported on virtual machine guest
 
 Input parameters:

py-interface/amdsmi_interface.py

@@ -3105,17 +3105,18 @@ def amdsmi_get_energy_count(processor_handle: amdsmi_wrapper.amdsmi_processor_ha
             processor_handle, amdsmi_wrapper.amdsmi_processor_handle
         )
 
-    power = ctypes.c_uint64()
+    energy_accumulator= ctypes.c_uint64()
     counter_resolution = ctypes.c_float()
     timestamp = ctypes.c_uint64()
 
     _check_res(
         amdsmi_wrapper.amdsmi_get_energy_count(processor_handle, ctypes.byref(
-            power), ctypes.byref(counter_resolution), ctypes.byref(timestamp))
+            energy_accumulator), ctypes.byref(counter_resolution), ctypes.byref(timestamp))
     )
 
     return {
-        'power': power.value,
+        'power': energy_accumulator.value, # deprecating in 6.4
+        'energy_accumulator': energy_accumulator.value,
         'counter_resolution': counter_resolution.value,
         'timestamp': timestamp.value,
     }

src/amd_smi/amd_smi.cc

@@ -1591,9 +1591,9 @@ amdsmi_status_t amdsmi_get_utilization_count(amdsmi_processor_handle processor_h
 }
 
 amdsmi_status_t amdsmi_get_energy_count(amdsmi_processor_handle processor_handle,
-            uint64_t *power, float *counter_resolution, uint64_t *timestamp) {
+            uint64_t *energy_accumulator, float *counter_resolution, uint64_t *timestamp) {
     return rsmi_wrapper(rsmi_dev_energy_count_get, processor_handle,
-            power, counter_resolution, timestamp);
+            energy_accumulator, counter_resolution, timestamp);
 }
 
 amdsmi_status_t amdsmi_get_gpu_bdf_id(

tests/amd_smi_test/functional/metrics_counter_read.cc

@@ -104,10 +104,10 @@ void TestMetricsCounterRead::Run(void) {
         std::cout << "\t**GPU METRICS ENERGY COUNTER:\n";
     }
 
-    uint64_t power;
+    uint64_t energy_accumulator;
     uint64_t timestamp;
     float counter_resolution;
-    err = amdsmi_get_energy_count(processor_handles_[i], &power, &counter_resolution, &timestamp);
+    err = amdsmi_get_energy_count(processor_handles_[i], &energy_accumulator, &counter_resolution, &timestamp);
     if (err != AMDSMI_STATUS_SUCCESS) {
       if (err == AMDSMI_STATUS_NOT_SUPPORTED) {
         IF_VERB(STANDARD) {
@@ -119,10 +119,10 @@ void TestMetricsCounterRead::Run(void) {
     } else {
       CHK_ERR_ASRT(err);
       IF_VERB(STANDARD) {
-          std::cout << std::dec << "power counter="
-          << power << '\n';
-          std::cout << "power in uJ="
-          << (double)(power * counter_resolution) << '\n';
+          std::cout << std::dec << "energy_accumulator counter="
+          << energy_accumulator << '\n';
+          std::cout << "energy_accumulator in uJ="
+          << (double)(energy_accumulator * counter_resolution) << '\n';
           std::cout << std::dec << "timestamp="
           << timestamp << '\n';
       }