[SWDEV-450553] Added gpu memory overdrive to metric function

Signed-off-by: gabrpham <Gabriel.Pham@amd.com>
Change-Id: If7bd6865d641a5a83c594a4d3c57938b1b6dc18e


[ROCm/amdsmi commit: 7d8e54d0e1]

projects/amdsmi/CHANGELOG.md

@@ -13,6 +13,9 @@ Full documentation for amd_smi_lib is available at [https://rocm.docs.amd.com/pr
 - **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.
 
+- **Added GPU memory overdrive percentage to `amd-smi metric -o`**.
+Added `amdsmi_get_gpu_mem_overdrive_level()` function to amd-smi C and Python Libraries.
+
 ### Removals
 
 - N/A

projects/amdsmi/amdsmi_cli/amdsmi_commands.py

@@ -1851,18 +1851,19 @@ class AMDSMICommands():
             if args.overdrive:
                 try:
                     overdrive_level = amdsmi_interface.amdsmi_get_gpu_overdrive_level(args.gpu)
-
                     od_unit = '%'
-                    if self.logger.is_human_readable_format():
-                        overdrive_level = f"{overdrive_level} {od_unit}"
-                    if self.logger.is_json_format():
-                        overdrive_level = {"value" : overdrive_level,
-                                           "unit" : od_unit}
-
-                    values_dict['overdrive'] = overdrive_level
+                    values_dict['overdrive'] = self.helpers.unit_format(self.logger, overdrive_level, od_unit)
                 except amdsmi_exception.AmdSmiLibraryException as e:
                     values_dict['overdrive'] = "N/A"
-                    logging.debug("Failed to get overdrive level for gpu %s | %s", gpu_id, e.get_error_info())
+                    logging.debug("Failed to get gpu overdrive level for gpu %s | %s", gpu_id, e.get_error_info())
+
+                try:
+                    mem_overdrive_level = amdsmi_interface.amdsmi_get_gpu_mem_overdrive_level(args.gpu)
+                    od_unit = '%'
+                    values_dict['mem_overdrive'] = self.helpers.unit_format(self.logger, mem_overdrive_level, od_unit)
+                except amdsmi_exception.AmdSmiLibraryException as e:
+                    values_dict['mem_overdrive'] = "N/A"
+                    logging.debug("Failed to get mem overdrive level for gpu %s | %s", gpu_id, e.get_error_info())
         if "perf_level" in current_platform_args:
             if args.perf_level:
                 try:

projects/amdsmi/amdsmi_cli/amdsmi_parser.py

@@ -740,7 +740,7 @@ class AMDSMIParser(argparse.ArgumentParser):
         # Help text for Arguments only on Linux Baremetal platforms
         fan_help = "Current fan speed"
         vc_help = "Display voltage curve"
-        overdrive_help = "Current GPU clock overdrive level"
+        overdrive_help = "Current GPU clock overdrive and GPU memory clock overdrive level"
         perf_level_help = "Current DPM performance level"
         xgmi_err_help = "XGMI error information since last read"
         energy_help = "Amount of energy consumed"

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

@@ -1610,7 +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:
@@ -2156,6 +2156,38 @@ except AmdSmiException as e:
     print(e)
 ```
 
+### amdsmi_get_gpu_mem_overdrive_level
+
+Description: Get the GPU memory clock overdrive percent associated with the device with provided
+device handle. It is not supported on virtual machine guest
+
+Input parameters:
+
+* `processor_handle` handle for the given device
+
+Output: Overdrive percentage as integer
+
+Exceptions that can be thrown by `amdsmi_get_gpu_mem_overdrive_level` function:
+
+* `AmdSmiLibraryException`
+* `AmdSmiRetryException`
+* `AmdSmiParameterException`
+
+Example:
+
+```python
+try:
+    devices = amdsmi_get_processor_handles()
+    if len(devices) == 0:
+        print("No GPUs on machine")
+    else:
+        for device in devices:
+            od_level = amdsmi_get_gpu_mem_overdrive_level(dev)
+            print(od_level)
+except AmdSmiException as e:
+    print(e)
+```
+
 ### amdsmi_get_clk_freq
 
 Description: Get the list of possible system clock speeds of device for a

projects/amdsmi/include/amd_smi/amdsmi.h

@@ -2957,6 +2957,30 @@ amdsmi_set_gpu_perf_determinism_mode(amdsmi_processor_handle processor_handle, u
 amdsmi_status_t amdsmi_get_gpu_overdrive_level(amdsmi_processor_handle processor_handle,
                                                 uint32_t *od);
 
+/**
+ *  @brief Get the GPU memory clock overdrive percent associated with the device with provided
+ *  processor handle. It is not supported on virtual machine guest
+ *
+ *  @platform{gpu_bm_linux}
+ *
+ *  @details Given a processor handle @p processor_handle and a pointer to a uint32_t @p od,
+ *  this function will write the overdrive percentage to the uint32_t pointed
+ *  to by @p od
+ *
+ *  @param[in] processor_handle a processor handle
+ *
+ *  @param[in,out] od a pointer to uint32_t to which the GPU memory clock overdrive percentage
+ *  will be written
+ *  If this parameter is nullptr, this function will return
+ *  ::AMDSMI_STATUS_INVAL if the function is supported with the provided,
+ *  arguments and ::AMDSMI_STATUS_NOT_SUPPORTED if it is not supported with the
+ *  provided arguments.
+ *
+ *  @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
+ */
+amdsmi_status_t amdsmi_get_gpu_mem_overdrive_level(amdsmi_processor_handle processor_handle,
+                                                uint32_t *od);
+
 /**
  *  @brief Get the list of possible system clock speeds of device for a
  *  specified clock type. It is not supported on virtual machine guest

projects/amdsmi/py-interface/README.md

@@ -2156,6 +2156,38 @@ except AmdSmiException as e:
     print(e)
 ```
 
+### amdsmi_get_gpu_mem_overdrive_level
+
+Description: Get the GPU memory clock overdrive percent associated with the device with provided
+device handle. It is not supported on virtual machine guest
+
+Input parameters:
+
+* `processor_handle` handle for the given device
+
+Output: Overdrive percentage as integer
+
+Exceptions that can be thrown by `amdsmi_get_gpu_mem_overdrive_level` function:
+
+* `AmdSmiLibraryException`
+* `AmdSmiRetryException`
+* `AmdSmiParameterException`
+
+Example:
+
+```python
+try:
+    devices = amdsmi_get_processor_handles()
+    if len(devices) == 0:
+        print("No GPUs on machine")
+    else:
+        for device in devices:
+            od_level = amdsmi_get_gpu_mem_overdrive_level(dev)
+            print(od_level)
+except AmdSmiException as e:
+    print(e)
+```
+
 ### amdsmi_get_clk_freq
 
 Description: Get the list of possible system clock speeds of device for a

projects/amdsmi/py-interface/__init__.py

@@ -150,6 +150,7 @@ from .amdsmi_interface import amdsmi_get_gpu_volt_metric
 from .amdsmi_interface import amdsmi_get_utilization_count
 from .amdsmi_interface import amdsmi_get_gpu_perf_level
 from .amdsmi_interface import amdsmi_get_gpu_overdrive_level
+from .amdsmi_interface import amdsmi_get_gpu_mem_overdrive_level
 from .amdsmi_interface import amdsmi_get_clk_freq
 from .amdsmi_interface import amdsmi_get_gpu_od_volt_info
 from .amdsmi_interface import amdsmi_get_gpu_metrics_info

projects/amdsmi/py-interface/amdsmi_interface.py

@@ -3458,6 +3458,24 @@ def amdsmi_get_gpu_overdrive_level(
     return od_level.value
 
 
+def amdsmi_get_gpu_mem_overdrive_level(
+    processor_handle: amdsmi_wrapper.amdsmi_processor_handle,
+) -> int:
+    if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
+        raise AmdSmiParameterException(
+            processor_handle, amdsmi_wrapper.amdsmi_processor_handle
+        )
+
+    mem_od_level = ctypes.c_uint32()
+    _check_res(
+        amdsmi_wrapper.amdsmi_get_gpu_mem_overdrive_level(
+            processor_handle, ctypes.byref(mem_od_level)
+        )
+    )
+
+    return mem_od_level.value
+
+
 def amdsmi_get_clk_freq(
     processor_handle: amdsmi_wrapper.amdsmi_processor_handle, clk_type: AmdSmiClkType
 ) -> Dict[str, Any]:

projects/amdsmi/py-interface/amdsmi_wrapper.py

@@ -2060,6 +2060,9 @@ amdsmi_set_gpu_perf_determinism_mode.argtypes = [amdsmi_processor_handle, uint64
 amdsmi_get_gpu_overdrive_level = _libraries['libamd_smi.so'].amdsmi_get_gpu_overdrive_level
 amdsmi_get_gpu_overdrive_level.restype = amdsmi_status_t
 amdsmi_get_gpu_overdrive_level.argtypes = [amdsmi_processor_handle, ctypes.POINTER(ctypes.c_uint32)]
+amdsmi_get_gpu_mem_overdrive_level = _libraries['libamd_smi.so'].amdsmi_get_gpu_mem_overdrive_level
+amdsmi_get_gpu_mem_overdrive_level.restype = amdsmi_status_t
+amdsmi_get_gpu_mem_overdrive_level.argtypes = [amdsmi_processor_handle, ctypes.POINTER(ctypes.c_uint32)]
 amdsmi_get_clk_freq = _libraries['libamd_smi.so'].amdsmi_get_clk_freq
 amdsmi_get_clk_freq.restype = amdsmi_status_t
 amdsmi_get_clk_freq.argtypes = [amdsmi_processor_handle, amdsmi_clk_type_t, ctypes.POINTER(struct_amdsmi_frequencies_t)]
@@ -2671,7 +2674,8 @@ __all__ = \
     'amdsmi_get_gpu_ecc_enabled', 'amdsmi_get_gpu_ecc_status',
     'amdsmi_get_gpu_event_notification', 'amdsmi_get_gpu_fan_rpms',
     'amdsmi_get_gpu_fan_speed', 'amdsmi_get_gpu_fan_speed_max',
-    'amdsmi_get_gpu_id', 'amdsmi_get_gpu_memory_partition',
+    'amdsmi_get_gpu_id', 'amdsmi_get_gpu_mem_overdrive_level',
+    'amdsmi_get_gpu_memory_partition',
     'amdsmi_get_gpu_memory_reserved_pages',
     'amdsmi_get_gpu_memory_total', 'amdsmi_get_gpu_memory_usage',
     'amdsmi_get_gpu_metrics_header_info',

projects/amdsmi/src/amd_smi/amd_smi.cc

@@ -1506,6 +1506,12 @@ amdsmi_status_t amdsmi_get_gpu_overdrive_level(
     return rsmi_wrapper(rsmi_dev_overdrive_level_get, processor_handle, od);
 }
 
+amdsmi_status_t amdsmi_get_gpu_mem_overdrive_level(
+            amdsmi_processor_handle processor_handle,
+            uint32_t *od) {
+    return rsmi_wrapper(rsmi_dev_mem_overdrive_level_get, processor_handle, od);
+}
+
 amdsmi_status_t  amdsmi_set_gpu_overdrive_level(
             amdsmi_processor_handle processor_handle, uint32_t od) {
     return rsmi_wrapper(rsmi_dev_overdrive_level_set_v1, processor_handle, od);