Add OAM_ID

Change-Id: I771b2f7f088940838c09ba3521a7955faa64e7ec Signed-off-by: Galantsev, Dmitrii <[email protected]>
ROCm · Sep 10, 2024 · a0f7290 · a0f7290
1 parent 8c56855
commit a0f7290
Show file tree

Hide file tree

Showing 6 changed files with 19 additions and 1 deletion.
diff --git a/common/rdc_field.data b/common/rdc_field.data
@@ -36,6 +36,7 @@ THE SOFTWARE.
 FLD_DESC_ENT(RDC_FI_INVALID,             "Unknown/Invalid field",                       "INVALID",          false)
 FLD_DESC_ENT(RDC_FI_GPU_COUNT,           "GPU count in the system",                     "GPU_COUNT",        true)
 FLD_DESC_ENT(RDC_FI_DEV_NAME,            "Name of the device",                          "DEV_NAME",         true)
+FLD_DESC_ENT(RDC_FI_OAM_ID,              "OAM ID of the device",                        "OAM_ID",           true)
 FLD_DESC_ENT(RDC_FI_GPU_CLOCK,           "Current GPU clock frequencies",               "GPU_CLOCK",        true)
 FLD_DESC_ENT(RDC_FI_MEM_CLOCK,           "Current Memory clock frequencies",            "MEM_CLOCK",        true)
 FLD_DESC_ENT(RDC_FI_MEMORY_TEMP,         "Memory temperature in millidegrees Celsius",  "MEMORY_TEMP",      true)

diff --git a/include/rdc/rdc.h b/include/rdc/rdc.h
@@ -149,6 +149,7 @@ typedef enum {
   //!< @brief Identifier fields
   RDC_FI_GPU_COUNT = 1,  //!< GPU count in the system
   RDC_FI_DEV_NAME,       //!< Name of the device
+  RDC_FI_OAM_ID,         //!< OAM ID of the device
 
   /**
    * @brief Frequency related fields

diff --git a/python_binding/RdcReader.py b/python_binding/RdcReader.py
@@ -5,6 +5,7 @@
 default_field_ids = [
         rdc_field_t.RDC_FI_GPU_MEMORY_USAGE,
         rdc_field_t.RDC_FI_GPU_MEMORY_TOTAL,
+        rdc_field_t.RDC_FI_OAM_ID,
         rdc_field_t.RDC_FI_POWER_USAGE,
         rdc_field_t.RDC_FI_GPU_CLOCK,
         rdc_field_t.RDC_FI_GPU_UTIL,

diff --git a/python_binding/rdc_bootstrap.py b/python_binding/rdc_bootstrap.py
@@ -75,6 +75,7 @@ class rdc_field_t(c_int):
      RDC_FI_INVALID = 0
      RDC_FI_GPU_COUNT = 1
      RDC_FI_DEV_NAME = 2
+     RDC_FI_OAM_ID = 3
      RDC_FI_GPU_CLOCK = 100
      RDC_FI_MEM_CLOCK = 101
      RDC_FI_MEMORY_TEMP = 200

diff --git a/rdc_libs/rdc/src/RdcMetricFetcherImpl.cc b/rdc_libs/rdc/src/RdcMetricFetcherImpl.cc
@@ -623,6 +623,20 @@ rdc_status_t RdcMetricFetcherImpl::fetch_smi_field(uint32_t gpu_index, rdc_field
       }
       break;
     }
+    case RDC_FI_OAM_ID: {
+      amdsmi_asic_info_t asic_info;
+      value->status = amdsmi_get_gpu_asic_info(processor_handle, &asic_info);
+      value->type = INTEGER;
+      if (value->status == AMDSMI_STATUS_SUCCESS) {
+        // 0xFFFF means not supported for OAM ID
+        if (asic_info.oam_id == 0xFFFF) {
+          value->status = AMDSMI_STATUS_NOT_SUPPORTED;
+        } else {
+          value->value.l_int = asic_info.oam_id;
+        }
+      }
+      break;
+    }
     case RDC_FI_ECC_CORRECT_TOTAL:
     case RDC_FI_ECC_UNCORRECT_TOTAL:
       get_ecc_total(gpu_index, field_id, value);

diff --git a/rdc_libs/rdc/src/RdcSmiLib.cc b/rdc_libs/rdc/src/RdcSmiLib.cc
@@ -176,7 +176,7 @@ rdc_status_t RdcSmiLib::rdc_telemetry_fields_query(uint32_t field_ids[MAX_NUM_FI
       RDC_EVNT_XGMI_0_BEATS_TX, RDC_EVNT_XGMI_1_NOP_TX,   RDC_EVNT_XGMI_1_REQ_TX,
       RDC_EVNT_XGMI_1_RESP_TX,  RDC_EVNT_XGMI_1_BEATS_TX, RDC_EVNT_XGMI_0_THRPUT,
       RDC_EVNT_XGMI_1_THRPUT,   RDC_EVNT_XGMI_2_THRPUT,   RDC_EVNT_XGMI_3_THRPUT,
-      RDC_EVNT_XGMI_4_THRPUT,   RDC_EVNT_XGMI_5_THRPUT,
+      RDC_EVNT_XGMI_4_THRPUT,   RDC_EVNT_XGMI_5_THRPUT,   RDC_FI_OAM_ID,
   };
   std::copy(fields.begin(), fields.end(), field_ids);
   *field_count = fields.size();