Use sysfs file to get average power instead of debugfs

include/rocm_smi/rocm_smi.h

@@ -446,7 +446,7 @@ rsmi_status_t rsmi_dev_gpu_clk_freq_set(uint32_t dv_ind,
  *  @param[in] dv_ind a device index
  *
  *  @param[inout] name a pointer to a caller provided char buffer to which the
- *  speed will be written
+ *  name will be written
  *
  *  @param[in] len the length of the caller provided buffer @p name.
  *
@@ -588,7 +588,7 @@ rsmi_status_t rsmi_dev_fan_speed_set(uint32_t dv_ind, uint32_t sensor_ind,
  *
  *  @details Given a device index @p dv_ind and a pointer to a uint64_t
  *  @p power, this function will write the current average power consumption to
- *  the uint64_t in milliwatts pointed to by @p power. This function requires
+ *  the uint64_t in microwatts pointed to by @p power. This function requires
  *  root privilege.
  *
  *  @param[in] dv_ind a device index

include/rocm_smi/rocm_smi_monitor.h

@@ -63,6 +63,7 @@ enum MonitorTypes {
   kMonPowerCap,
   kMonPowerCapMax,
   kMonPowerCapMin,
+  kMonPowerAve,
   kMonTempMax,
   kMonTempMin,
   kMonTempMaxHyst,

include/rocm_smi/rocm_smi_power_mon.h

@@ -55,7 +55,6 @@ namespace smi {
 
 enum PowerMonTypes {
   kPowerMaxGPUPower,
-  kPowerAveGPUPower,
 };
 
 

src/rocm_smi.cc

@@ -851,11 +851,10 @@ rsmi_dev_power_ave_get(uint32_t dv_ind, uint32_t sensor_ind, uint64_t *power) {
   if (power == nullptr) {
     return RSMI_STATUS_INVALID_ARGS;
   }
-  (void)sensor_ind;   // Not used yet
-  // ++sensor_ind;  // power sysfs files have 1-based indices
+  ++sensor_ind;  // power sysfs files have 1-based indices
 
   rsmi_status_t ret;
-  ret = get_power_mon_value(amd::smi::kPowerAveGPUPower, dv_ind, power);
+  ret = get_dev_mon_value(amd::smi::kMonPowerAve, dv_ind, sensor_ind, power);
 
   return ret;
   CATCH

src/rocm_smi_monitor.cc

@@ -74,7 +74,7 @@ static const char *kMonNameFName = "name";
 static const char *kMonPowerCapName = "power#_cap";
 static const char *kMonPowerCapMaxName = "power#_cap_max";
 static const char *kMonPowerCapMinName = "power#_cap_min";
-
+static const char *kMonPowerAveName = "power#_average";
 static const char *kMonTempMaxName = "temp#_max";
 static const char *kMonTempMinName = "temp#_min";
 static const char *kMonTempMaxHystName = "temp#_max_hyst";
@@ -99,6 +99,7 @@ static const std::map<MonitorTypes, const char *> kMonitorNameMap = {
     {kMonPowerCap, kMonPowerCapName},
     {kMonPowerCapMax, kMonPowerCapMaxName},
     {kMonPowerCapMin, kMonPowerCapMinName},
+    {kMonPowerAve, kMonPowerAveName},
     {kMonTempMax, kMonTempMaxName},
     {kMonTempMin, kMonTempMinName},
     {kMonTempMaxHyst, kMonTempMaxHystName},

src/rocm_smi_power_mon.cc

@@ -66,7 +66,6 @@ static const char *kPowerMonPMName = "amdgpu_pm_info";
 // Using this map in case we add other files from dri directory to parse.
 static const std::map<PowerMonTypes, const char *> kMonitorNameMap = {
     {kPowerMaxGPUPower, kPowerMonPMName},
-    {kPowerAveGPUPower, kPowerMonPMName},
 };
 
 PowerMon::PowerMon(std::string path, RocmSMI_env_vars const *e) :
@@ -87,10 +86,6 @@ static int parse_power_str(std::string s, PowerMonTypes type, uint64_t *val) {
       search_str = "(max GPU)";
       break;
 
-    case kPowerAveGPUPower:
-      search_str = "(average GPU)";
-      break;
-
     default:
       assert(!"Invalid search Power type requested");
       return EINVAL;
@@ -118,7 +113,6 @@ static int parse_power_str(std::string s, PowerMonTypes type, uint64_t *val) {
 
   switch (type) {
     case kPowerMaxGPUPower:
-    case kPowerAveGPUPower:
       l_ss >> num_units;
       l_ss >> sz;
       assert(sz == "W");  // We only expect Watts at this time

tests/rocm_smi_test/functional/rsmi_sanity.cc

@@ -641,8 +641,8 @@ void TestSanity::Run(void) {
         std::cout << "\t**Power Cap Range: " << val2_ui64 << " to " <<
                                                  val_ui64 << " uW" << std::endl;
       }
-#if 0    // Same as above (disable for now)
-      err = rsmi_dev_power_ave_get(i, &val_ui64);
+
+      err = rsmi_dev_power_ave_get(i, 0, &val_ui64);
       IF_VERB(STANDARD) {
         std::cout << "\t**Averge Power Usage: ";
         CHK_RSMI_PERM_ERR(err)
@@ -651,7 +651,6 @@ void TestSanity::Run(void) {
         }
         std::cout << "\t=======" << std::endl;
       }
-#endif
     }
     IF_VERB(STANDARD) {
       std::cout << "***** Testing write api's" << std::endl;

tests/rocm_smi_test/functional/rsmi_sanity.h

@@ -42,8 +42,8 @@
  * DEALINGS WITH THE SOFTWARE.
  *
  */
-#ifndef ROCM_SMI_LIB_TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_
-#define ROCM_SMI_LIB_TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_
+#ifndef TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_
+#define TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_
 
 #include "rocm_smi_test/test_base.h"
 
@@ -70,4 +70,4 @@ class TestSanity : public TestBase {
   virtual void DisplayTestInfo(void);
 };
 
-#endif  // ROCM_SMI_LIB_TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_
+#endif  // TESTS_ROCM_SMI_TEST_FUNCTIONAL_RSMI_SANITY_H_