Merge amd-dev into amd-master 20240215

Signed-off-by: Maisam Arif <maisarif@amd.com>
Change-Id: I089fcea8184d323f5e65311bb63c231cb94890e3


[ROCm/amdsmi commit: fad7339c39]

projects/amdsmi/CMakeLists.txt

@@ -28,7 +28,7 @@ find_program(GIT NAMES git)
 
 ## Setup the package version based on git tags.
 set(PKG_VERSION_GIT_TAG_PREFIX "amdsmi_pkg_ver")
-get_package_version_number("24.2.0" ${PKG_VERSION_GIT_TAG_PREFIX} GIT)
+get_package_version_number("24.3.0" ${PKG_VERSION_GIT_TAG_PREFIX} GIT)
 message("Package version: ${PKG_VERSION_STR}")
 set(${AMD_SMI_LIBS_TARGET}_VERSION_MAJOR "${CPACK_PACKAGE_VERSION_MAJOR}")
 set(${AMD_SMI_LIBS_TARGET}_VERSION_MINOR "${CPACK_PACKAGE_VERSION_MINOR}")

projects/amdsmi/amdsmi_cli/README.md

@@ -78,7 +78,7 @@ amd-smi will report the version and current platform detected when running the c
 ~$ amd-smi
 usage: amd-smi [-h]  ...
 
-AMD System Management Interface | Version: 24.2.0.0 | ROCm version: 6.1.0 | Platform: Linux
+AMD System Management Interface | Version: 24.3.0.0 | ROCm version: 6.1.0 | Platform: Linux
 Baremetal
 
 options:

projects/amdsmi/amdsmi_cli/amdsmi_commands.py

@@ -152,10 +152,10 @@ class AMDSMICommands():
 
         # Store values based on format
         if self.logger.is_human_readable_format():
-            self.logger.store_output(args.gpu, 'AMDSMI_SPACING_REMOVAL', {'bdf':bdf, 'uuid':uuid})
+            self.logger.store_output(args.gpu, 'AMDSMI_SPACING_REMOVAL', {'gpu_bdf':bdf, 'gpu_uuid':uuid})
         else:
-            self.logger.store_output(args.gpu, 'bdf', bdf)
-            self.logger.store_output(args.gpu, 'uuid', uuid)
+            self.logger.store_output(args.gpu, 'gpu_bdf', bdf)
+            self.logger.store_output(args.gpu, 'gpu_uuid', uuid)
 
         if multiple_devices:
             self.logger.store_multiple_device_output()
@@ -323,16 +323,6 @@ class AMDSMICommands():
         if args.asic:
             try:
                 asic_info = amdsmi_interface.amdsmi_get_gpu_asic_info(args.gpu)
-                asic_info['vendor_id'] = hex(asic_info['vendor_id'])
-                asic_info['vendor_name'] = asic_info['vendor_name'].replace(',', '')
-                asic_info['device_id'] = hex(asic_info['device_id'])
-                asic_info['rev_id'] = hex(asic_info['rev_id'])
-                if asic_info['asic_serial'] != '':
-                    asic_info['asic_serial'] = hex(int(asic_info['asic_serial'], base=16))
-                else:
-                    asic_info['asic_serial'] = "N/A"
-                if asic_info['oam_id'] == 0xFFFF: # uint 16 max
-                    asic_info['oam_id'] = "N/A"
                 static_dict['asic'] = asic_info
             except amdsmi_exception.AmdSmiLibraryException as e:
                 static_dict['asic'] = "N/A"
@@ -560,21 +550,32 @@ class AMDSMICommands():
             static_dict['vram'] = vram_info
         if args.cache:
             try:
-                cache_info = amdsmi_interface.amdsmi_get_gpu_cache_info(args.gpu)
-                logging.debug(f"cache_info dictionary = {cache_info}")
-
+                cach_info_list = amdsmi_interface.amdsmi_get_gpu_cache_info(args.gpu)
+                logging.debug(f"cache_info dictionary = {cach_info_list}")
                 if self.logger.is_human_readable_format():
-                    for key, cache_values in cache_info.items():
-                        cache_values['cache_size'] = f"{cache_values['cache_size']} KB"
+                    cache_info_dict_format = {}
+                    for cache_dict in cach_info_list:
+                        cache_index = "cache_" + str(cache_dict["cache"])
+                        cache_info_dict_format[cache_index] = cache_dict
+
+                        # Remove cache index from new dictionary
+                        cache_info_dict_format[cache_index].pop("cache")
+
+                        # Add cache_size unit
+                        cache_info_dict_format[cache_index]["cache_size"] = f"{cache_info_dict_format[cache_index]['cache_size']} KB"
+
                         # take cache_properties out of list -> display as string, removing brackets
-                        cache_values['cache_properties'] = ", ".join(cache_values['cache_properties'])
-                logging.debug(f"After human_readable | cache_info = {cache_info}")
+                        cache_info_dict_format[cache_index]["cache_properties"] = ", ".join(cache_info_dict_format[cache_index]["cache_properties"])
+
+                    cach_info_list = cache_info_dict_format
+
+                logging.debug(f"After human_readable | cache_info = {cach_info_list}")
 
             except amdsmi_exception.AmdSmiLibraryException as e:
-                cache_info = "N/A"
+                cach_info_list = "N/A"
                 logging.debug("Failed to get cache info for gpu %s | %s", gpu_id, e.get_error_info())
 
-            static_dict['cache'] = cache_info
+            static_dict['cache_info'] = cach_info_list
         if 'ras' in current_platform_args:
             if args.ras:
                 ras_dict = {"eeprom_version": "N/A",
@@ -1961,7 +1962,7 @@ class AMDSMICommands():
             if (args.cpu_metrics_ver):
                 static_dict["metric_version"] = {}
                 try:
-                    version = amdsmi_interface.amdsmi_get_metrics_table_version(args.cpu)
+                    version = amdsmi_interface.amdsmi_get_hsmp_metrics_table_version(args.cpu)
                     static_dict["metric_version"]["version"] = version
                 except amdsmi_exception.AmdSmiLibraryException as e:
                     static_dict["metric_version"]["version"] = "N/A"
@@ -1982,7 +1983,7 @@ class AMDSMICommands():
                     static_dict["metrics_table"]["cpu_model"] = "N/A"
                     logging.debug("Failed to get cpu model | %s", e.get_error_info())
                 try:
-                    metrics_table = amdsmi_interface.amdsmi_get_metrics_table(args.cpu)
+                    metrics_table = amdsmi_interface.amdsmi_get_hsmp_metrics_table(args.cpu)
                     static_dict["metrics_table"]["response"] = metrics_table
                 except amdsmi_exception.AmdSmiLibraryException as e:
                     static_dict["metrics_table"]["response"] = "N/A"

projects/amdsmi/amdsmi_cli/amdsmi_parser.py

@@ -994,8 +994,8 @@ core limit value"
 
 
     def _add_monitor_parser(self, subparsers, func):
-        if not(self.helpers.is_baremetal() and self.helpers.is_linux()):
-            # This subparser is only applicable to Baremetal Linux
+        if not(self.helpers.is_linux()):
+            # This subparser is only applicable to Linux
             return
 
         # Subparser help text

projects/amdsmi/docs/doxygen/Doxyfile

@@ -48,7 +48,7 @@ PROJECT_NAME           = AMD SMI
 # could be handy for archiving the generated documentation or if some version
 # control system is used.
 
-PROJECT_NUMBER         = "24.2.0.0"
+PROJECT_NUMBER         = "24.3.0.0"
 
 # Using the PROJECT_BRIEF tag one can provide an optional one line description
 # for a project that appears at the top of each page and should give viewer a
@@ -2276,7 +2276,7 @@ INCLUDE_FILE_PATTERNS  =
 # recursively expanded use the := operator instead of the = operator.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-PREDEFINED             =
+PREDEFINED             = ENABLE_ESMI_LIB
 
 # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this
 # tag can be used to specify a list of macro names that should be expanded. The

projects/amdsmi/example/amd_smi_drm_example.cc

@@ -51,6 +51,7 @@
 #include <cstring>
 #include <iostream>
 #include <vector>
+#include <sstream>
 
 #include "amd_smi/amdsmi.h"
 
@@ -212,6 +213,13 @@ void getFWNameFromId(int id, char *name)
 	}
 }
 
+template <typename T>
+std::string print_unsigned_int(T value) {
+  std::stringstream ss;
+  ss << static_cast<uint64_t>(value | 0);
+
+  return ss.str();
+}
 
 int main() {
     amdsmi_status_t ret;
@@ -315,7 +323,7 @@ int main() {
                 printf("\tCache Level: %d, Cache Size: %d KB, Cache type: 0x%x\n",
                     cache_info.cache[i].cache_level,
                     cache_info.cache[i].cache_size,
-                    cache_info.cache[i].properties);
+                    cache_info.cache[i].cache_properties);
                 printf("\tMax number CU shared: %d, Number of instances: %d\n",
                     cache_info.cache[i].max_num_cu_shared,
                     cache_info.cache[i].num_cache_instance);
@@ -655,6 +663,177 @@ int main() {
                       << "\n\n";
             std::cout << "\t\t Max Power Cap: " << cap_info.max_power_cap
                       << "\n\n";
+
+            /// Get GPU Metrics info
+            std::cout << "\n\n";
+            amdsmi_gpu_metrics_t gpu_metrics;
+            ret = amdsmi_get_gpu_metrics_info(processor_handles[j], &gpu_metrics);
+            CHK_AMDSMI_RET(ret)
+            printf("    Output of amdsmi_get_gpu_metrics_info:\n");
+            printf("\tDevice[%d] BDF %04lx:%02x:%02x.%d\n\n", i,
+                   bdf.fields.domain_number,
+                   bdf.fields.bus_number,
+                   bdf.fields.device_number,
+                   bdf.fields.function_number);
+
+            std::cout << "\t**.common_header.format_revision : "
+                      << print_unsigned_int(gpu_metrics.common_header.format_revision) << "\n";
+            std::cout << "\t**.common_header.content_revision : "
+                      << print_unsigned_int(gpu_metrics.common_header.content_revision) << "\n";
+            std::cout << "\t**.temperature_edge : " << std::dec
+                      << gpu_metrics.temperature_edge << "\n";
+            std::cout << "\t**.temperature_hotspot : " << std::dec
+                      << gpu_metrics.temperature_hotspot << "\n";
+            std::cout << "\t**.temperature_mem : " << std::dec
+                      << gpu_metrics.temperature_mem << "\n";
+            std::cout << "\t**.temperature_vrgfx : " << std::dec
+                      << gpu_metrics.temperature_vrgfx << "\n";
+            std::cout << "\t**.temperature_vrsoc : " << std::dec
+                      << gpu_metrics.temperature_vrsoc << "\n";
+            std::cout << "\t**.temperature_vrmem : " << std::dec
+                      << gpu_metrics.temperature_vrmem << "\n";
+            std::cout << "\t**.average_gfx_activity : " << std::dec
+                      << gpu_metrics.average_gfx_activity << "\n";
+            std::cout << "\t**.average_umc_activity : " << std::dec
+                      << gpu_metrics.average_umc_activity << "\n";
+            std::cout << "\t**.average_mm_activity : " << std::dec
+                      << gpu_metrics.average_mm_activity << "\n";
+            std::cout << "\t**.average_socket_power : " << std::dec
+                      << gpu_metrics.average_socket_power << "\n";
+            std::cout << "\t**.energy_accumulator : " << std::dec
+                      << gpu_metrics.energy_accumulator << "\n";
+            std::cout << "\t**.system_clock_counter : " << std::dec
+                      << gpu_metrics.system_clock_counter << "\n";
+            std::cout << "\t**.average_gfxclk_frequency : " << std::dec
+                      << gpu_metrics.average_gfxclk_frequency << "\n";
+            std::cout << "\t**.average_socclk_frequency : " << std::dec
+                      << gpu_metrics.average_socclk_frequency << "\n";
+            std::cout << "\t**.average_uclk_frequency : " << std::dec
+                      << gpu_metrics.average_uclk_frequency << "\n";
+            std::cout << "\t**.average_vclk0_frequency : " << std::dec
+                      << gpu_metrics.average_vclk0_frequency<< "\n";
+            std::cout << "\t**.average_dclk0_frequency : " << std::dec
+                      << gpu_metrics.average_dclk0_frequency << "\n";
+            std::cout << "\t**.average_vclk1_frequency : " << std::dec
+                      << gpu_metrics.average_vclk1_frequency << "\n";
+            std::cout << "\t**.average_dclk1_frequency : " << std::dec
+                      << gpu_metrics.average_dclk1_frequency << "\n";
+            std::cout << "\t**.current_gfxclk : " << std::dec
+                      << gpu_metrics.current_gfxclk << "\n";
+            std::cout << "\t**.current_socclk : " << std::dec
+                      << gpu_metrics.current_socclk << "\n";
+            std::cout << "\t**.current_uclk : " << std::dec
+                      << gpu_metrics.current_uclk << "\n";
+            std::cout << "\t**.current_vclk0 : " << std::dec
+                      << gpu_metrics.current_vclk0 << "\n";
+            std::cout << "\t**.current_dclk0 : " << std::dec
+                      << gpu_metrics.current_dclk0 << "\n";
+            std::cout << "\t**.current_vclk1 : " << std::dec
+                      << gpu_metrics.current_vclk1 << "\n";
+            std::cout << "\t**.current_dclk1 : " << std::dec
+                      << gpu_metrics.current_dclk1 << "\n";
+            std::cout << "\t**.throttle_status : " << std::dec
+                      << gpu_metrics.throttle_status << "\n";
+            std::cout << "\t**.current_fan_speed : " << std::dec
+                      << gpu_metrics.current_fan_speed << "\n";
+            std::cout << "\t**.pcie_link_width : " << std::dec
+                      << gpu_metrics.pcie_link_width << "\n";
+            std::cout << "\t**.pcie_link_speed : " << std::dec
+                      << gpu_metrics.pcie_link_speed << "\n";
+            std::cout << "\t**.gfx_activity_acc : " << std::dec
+                      << gpu_metrics.gfx_activity_acc << "\n";
+            std::cout << "\t**.mem_activity_acc : " << std::dec
+                      << gpu_metrics.mem_activity_acc << "\n";
+            std::cout << "\t**.firmware_timestamp : " << std::dec
+                      << gpu_metrics.firmware_timestamp << "\n";
+            std::cout << "\t**.voltage_soc : " << std::dec
+                      << gpu_metrics.voltage_soc << "\n";
+            std::cout << "\t**.voltage_gfx : " << std::dec
+                      << gpu_metrics.voltage_gfx << "\n";
+            std::cout << "\t**.voltage_mem : " << std::dec
+                      << gpu_metrics.voltage_mem << "\n";
+            std::cout << "\t**.indep_throttle_status : " << std::dec
+                      << gpu_metrics.indep_throttle_status << "\n";
+            std::cout << "\t**.current_socket_power : " << std::dec
+                      << gpu_metrics.current_socket_power << "\n";
+            std::cout << "\t**.gfxclk_lock_status : " << std::dec
+                      << gpu_metrics.gfxclk_lock_status << "\n";
+            std::cout << "\t**.xgmi_link_width : " << std::dec
+                      << gpu_metrics.xgmi_link_width << "\n";
+            std::cout << "\t**.xgmi_link_speed : " << std::dec
+                      << gpu_metrics.xgmi_link_speed << "\n";
+            std::cout << "\t**.pcie_bandwidth_acc : " << std::dec
+                      << gpu_metrics.pcie_bandwidth_acc << "\n";
+            std::cout << "\t**.pcie_bandwidth_inst : " << std::dec
+                      << gpu_metrics.pcie_bandwidth_inst << "\n";
+            std::cout << "\t**.pcie_l0_to_recov_count_acc : " << std::dec
+                      << gpu_metrics.pcie_l0_to_recov_count_acc << "\n";
+            std::cout << "\t**.pcie_replay_count_acc : " << std::dec
+                      << gpu_metrics.pcie_replay_count_acc << "\n";
+            std::cout << "\t**.pcie_replay_rover_count_acc : " << std::dec
+                      << gpu_metrics.pcie_replay_rover_count_acc << "\n";
+
+            std::cout << "\t**.temperature_hbm[] : " << std::dec << "\n";
+            for (const auto& temp : gpu_metrics.temperature_hbm) {
+              std::cout << "\t  -> " << std::dec << temp << "\n";
+            }
+
+            std::cout << "\t**.vcn_activity[] : " << std::dec << "\n";
+            for (const auto& vcn : gpu_metrics.vcn_activity) {
+              std::cout << "\t  -> " << std::dec << vcn << "\n";
+            }
+
+            std::cout << "\t**.xgmi_read_data_acc[] : " << std::dec << "\n";
+            for (const auto& read_data : gpu_metrics.xgmi_read_data_acc) {
+              std::cout << "\t  -> " << std::dec << read_data << "\n";
+            }
+
+            std::cout << "\t**.xgmi_write_data_acc[] : " << std::dec << "\n";
+            for (const auto& write_data : gpu_metrics.xgmi_write_data_acc) {
+              std::cout << "\t  -> " << std::dec << write_data << "\n";
+            }
+
+            std::cout << "\t**.current_gfxclks[] : " << std::dec << "\n";
+            for (const auto& gfxclk : gpu_metrics.current_gfxclks) {
+              std::cout << "\t  -> " << std::dec << gfxclk << "\n";
+            }
+
+            std::cout << "\t**.current_socclks[] : " << std::dec << "\n";
+            for (const auto& socclk : gpu_metrics.current_socclks) {
+              std::cout << "\t  -> " << std::dec << socclk << "\n";
+            }
+
+            std::cout << "\t**.current_vclk0s[] : " << std::dec << "\n";
+            for (const auto& vclk : gpu_metrics.current_vclk0s) {
+              std::cout << "\t  -> " << std::dec << vclk << "\n";
+            }
+
+            std::cout << "\t**.current_dclk0s[] : " << std::dec << "\n";
+            for (const auto& dclk : gpu_metrics.current_dclk0s) {
+              std::cout << "\t  -> " << std::dec << dclk << "\n";
+            }
+
+            std::cout << "\n";
+            std::cout << "\t ** -> Checking metrics with constant changes ** " << "\n";
+            constexpr uint16_t kMAX_ITER_TEST = 10;
+            amdsmi_gpu_metrics_t gpu_metrics_check;
+            for (auto idx = uint16_t(1); idx <= kMAX_ITER_TEST; ++idx) {
+              amdsmi_get_gpu_metrics_info(processor_handles[j], &gpu_metrics_check);
+              std::cout << "\t\t -> firmware_timestamp [" << idx << "/" << kMAX_ITER_TEST << "]: " << gpu_metrics_check.firmware_timestamp << "\n";
+            }
+
+            std::cout << "\n";
+            for (auto idx = uint16_t(1); idx <= kMAX_ITER_TEST; ++idx) {
+              amdsmi_get_gpu_metrics_info(processor_handles[j], &gpu_metrics_check);
+              std::cout << "\t\t -> system_clock_counter [" << idx << "/" << kMAX_ITER_TEST << "]: " << gpu_metrics_check.system_clock_counter << "\n";
+            }
+            std::cout << "\n";
+
+            std::cout << "\n";
+            std::cout << "\t ** Note: Values MAX'ed out (UINTX MAX are unsupported for the version in question) ** " << "\n";
+            std::cout << "\n";
+            std::cout << "+=======+==================+============+=============="
+                      << "+=============+=============+=============+============+\n";
         }
     }
 

projects/amdsmi/example/amdsmi_esmi_intg_example.cc

@@ -287,8 +287,8 @@ int main(int argc, char **argv) {
       double fraction_uq10 = fraction_q10;
       const char* err_str1;
 
-      amdsmi_hsmp_metric_table_t mtbl = {};
-      ret = amdsmi_get_metrics_table(plist[index], &mtbl);
+      amdsmi_hsmp_metrics_table_t mtbl = {};
+      ret = amdsmi_get_hsmp_metrics_table(plist[index], &mtbl);
 
       if (ret != AMDSMI_STATUS_SUCCESS) {
           cout<<"Failed to get Metrics Table for CPU["<<index<<"], Err["<<ret<<"]" << endl;

projects/amdsmi/include/amd_smi/amdsmi.h

@@ -151,7 +151,7 @@ typedef enum {
 #define AMDSMI_LIB_VERSION_YEAR 24
 
 //! Major version should be changed for every header change (adding/deleting APIs, changing names, fields of structures, etc.)
-#define AMDSMI_LIB_VERSION_MAJOR 2
+#define AMDSMI_LIB_VERSION_MAJOR 3
 
 //! Minor version should be updated for each API change, but without changing headers
 #define AMDSMI_LIB_VERSION_MINOR 0
@@ -541,19 +541,19 @@ typedef struct {
  * @brief cache properties
  */
 typedef enum {
-  CACHE_PROPERTIES_ENABLED = 0x00000001,
-  CACHE_PROPERTIES_DATA_CACHE = 0x00000002,
-  CACHE_PROPERTIES_INST_CACHE = 0x00000004,
-  CACHE_PROPERTIES_CPU_CACHE = 0x00000008,
-  CACHE_PROPERTIES_SIMD_CACHE = 0x00000010,
+  AMDSMI_CACHE_PROPERTIES_ENABLED = 0x00000001,
+  AMDSMI_CACHE_PROPERTIES_DATA_CACHE = 0x00000002,
+  AMDSMI_CACHE_PROPERTIES_INST_CACHE = 0x00000004,
+  AMDSMI_CACHE_PROPERTIES_CPU_CACHE = 0x00000008,
+  AMDSMI_CACHE_PROPERTIES_SIMD_CACHE = 0x00000010,
 } amdsmi_cache_properties_type_t;
 
 typedef struct {
   uint32_t num_cache_types;
   struct cache_ {
+    uint32_t cache_properties;  // amdsmi_cache_properties_type_t which is a bitmask
     uint32_t cache_size; /* In KB */
     uint32_t cache_level;
-    uint32_t properties;  // amdsmi_cache_properties_type_t which is a bitmask
     uint32_t max_num_cu_shared;  /* Indicates how many Compute Units share this cache instance */
     uint32_t num_cache_instance;  /* total number of instance of this cache type */
     uint32_t reserved[3];
@@ -1527,7 +1527,7 @@ typedef struct {
 } amdsmi_dpm_level_t;
 
 /**
- * @brief Metrics table (supported only with hsmp proto version 6).
+ * @brief HSMP Metrics table (supported only with hsmp proto version 6).
  */
 typedef struct __attribute__((__packed__)){
     uint32_t accumulation_counter;
@@ -1606,7 +1606,7 @@ typedef struct __attribute__((__packed__)){
 
     /* New items at the end to maintain driver compatibility */
     uint32_t gfxclk_frequency[8];
-} amdsmi_hsmp_metric_table_t;
+} amdsmi_hsmp_metrics_table_t;
 
 #endif
 
@@ -1619,7 +1619,7 @@ typedef struct __attribute__((__packed__)){
 /**
  *  @brief Initialize the AMD SMI library
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details This function initializes the library and the internal data structures,
@@ -1642,7 +1642,7 @@ amdsmi_status_t amdsmi_init(uint64_t init_flags);
 /**
  *  @brief Shutdown the AMD SMI library
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details This function shuts down the library and internal data structures and
@@ -1663,7 +1663,7 @@ amdsmi_status_t amdsmi_shut_down(void);
 /**
  *  @brief Get the list of socket handles in the system.
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details Depends on what flag is passed to ::amdsmi_init.  AMDSMI_INIT_AMD_GPUS
@@ -1726,7 +1726,7 @@ amdsmi_status_t amdsmi_get_cpusocket_handles(uint32_t *socket_count,
 /**
  *  @brief Get information about the given socket
  *
- *  @platform{gpu_bm_linux}  @platform{host}  @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host}  @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details This function retrieves socket information. The @p socket_handle must
@@ -1818,7 +1818,7 @@ amdsmi_status_t amdsmi_get_processor_handles_by_type(amdsmi_socket_handle socket
 /**
  *  @brief Get the list of the processor handles associated to a socket.
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details This function retrieves the processor handles of a socket. The
@@ -1886,7 +1886,7 @@ amdsmi_status_t amdsmi_get_cpucore_handles(amdsmi_cpusocket_handle socket_handle
 /**
  *  @brief Get the processor type of the processor_handle
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{cpu_bm}  @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details This function retrieves the processor type. A processor_handle must be provided
@@ -1906,7 +1906,7 @@ amdsmi_status_t amdsmi_get_processor_type(amdsmi_processor_handle processor_hand
 /**
  *  @brief Get processor handle with the matching bdf.
  *
- *  @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}  
+ *  @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}
  *  @platform{guest_mvf} @platform{guest_windows}
  *
  *  @details Given bdf info @p bdf, this function will get
@@ -2462,7 +2462,7 @@ amdsmi_get_gpu_bad_page_info(amdsmi_processor_handle processor_handle, uint32_t
  *  @brief Returns RAS features info.
  *
  *  @platform{gpu_bm_linux}  @platform{host}
- *  
+ *
  *  @param[in] processor_handle Device handle which to query
  *
  *  @param[out] ras_feature RAS features that are currently enabled and supported on
@@ -2635,7 +2635,7 @@ amdsmi_status_t amdsmi_get_gpu_fan_speed_max(amdsmi_processor_handle processor_h
  *  specified temperature sensor on the specified device. It is not supported on
  *  virtual machine guest
  *
- *  @platform{gpu_bm_linux} @platform{host} 
+ *  @platform{gpu_bm_linux} @platform{host}
  *
  *  @details Given a processor handle @p processor_handle, a sensor type @p sensor_type, a
  *  ::amdsmi_temperature_metric_t @p metric and a pointer to an int64_t @p
@@ -2666,7 +2666,7 @@ amdsmi_status_t amdsmi_get_temp_metric(amdsmi_processor_handle processor_handle,
 /**
  *  @brief Returns gpu cache info.
  *
- *  @platform{gpu_bm_linux}  @platform{host} 
+ *  @platform{gpu_bm_linux}  @platform{host}
  *
  *  @param[in] processor_handle PF of a processor for which to query
  *
@@ -2935,6 +2935,27 @@ amdsmi_status_t amdsmi_reset_gpu(amdsmi_processor_handle processor_handle);
 amdsmi_status_t amdsmi_get_gpu_od_volt_info(amdsmi_processor_handle processor_handle,
                                                amdsmi_od_volt_freq_data_t *odv);
 
+/**
+ *  @brief Get the 'metrics_header_info' from the GPU metrics associated with the device
+ *
+ *  @platform{gpu_bm_linux}  @platform{guest_1vf}
+ *
+ *  @details Given a processor handle @p processor_handle and a pointer to a amd_metrics_table_header_t in which
+ *  the 'metrics_header_info' will stored
+ *
+ *  @param[in] processor_handle Device which to query
+ *
+ *  @param[inout] header_value a pointer to amd_metrics_table_header_t to which the device gpu
+ *  metric unit will be stored
+ *
+ *  @retval ::AMDSMI_STATUS_SUCCESS is returned upon successful call.
+ *          ::AMDSMI_STATUS_NOT_SUPPORTED is returned in case the metric unit
+ *            does not exist for the given device
+ *
+ */
+amdsmi_status_t
+amdsmi_get_gpu_metrics_header_info(amdsmi_processor_handle processor_handle, amd_metrics_table_header_t* header_value);
+
 /**
  *  @brief This function retrieves the gpu metrics information. It is not supported
  *  on virtual machine guest
@@ -4375,7 +4396,7 @@ amdsmi_status_t amdsmi_get_gpu_vram_info(
 /**
  *  @brief          Returns the board part number and board information for the requested device
  *
- *   @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}  @platform{guest_mvf} 
+ *   @platform{gpu_bm_linux}  @platform{host} @platform{guest_1vf}  @platform{guest_mvf}
  *
  *  @param[in]      processor_handle Device which to query
  *
@@ -4648,10 +4669,10 @@ amdsmi_get_gpu_total_ecc_count(amdsmi_processor_handle processor_handle, amdsmi_
 
 
 #ifdef ENABLE_ESMI_LIB
-/*---------------------------------------------------------------------------*/
-/**  @defgroup energyinfo     Energy information (RAPL MSR)                  */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup energyinfo     Energy information (RAPL MSR)
+ *  @{
+ */
 
 /**
  *  @brief Get the core energy for a given core.
@@ -4681,12 +4702,12 @@ amdsmi_status_t amdsmi_get_cpu_core_energy(amdsmi_processor_handle processor_han
 amdsmi_status_t amdsmi_get_cpu_socket_energy(amdsmi_processor_handle processor_handle,
                                              uint64_t *penergy);
 
-/** @}  */
+/** @} End energyinfo */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup systemstatistics     HSMP system statistics                   */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup systemstatistics     HSMP system statistics
+ *  @{
+ */
 
 /**
  *  @brief Get SMU Firmware Version.
@@ -4804,12 +4825,12 @@ amdsmi_status_t amdsmi_get_cpu_socket_freq_range(amdsmi_processor_handle process
 amdsmi_status_t amdsmi_get_cpu_core_current_freq_limit(amdsmi_processor_handle processor_handle,
                                                        uint32_t *freq);
 
-/** @} */
+/** @} systemstatistics */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup powercont    Power Control                                    */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup powercont    Power Control
+ *  @{
+ */
 
 /**
  *  @brief Get the socket power.
@@ -4895,12 +4916,12 @@ amdsmi_status_t amdsmi_set_cpu_socket_power_cap(amdsmi_processor_handle processo
 amdsmi_status_t amdsmi_set_cpu_pwr_efficiency_mode(amdsmi_processor_handle processor_handle,
                                                    uint8_t mode);
 
-/** @} */
+/** @} powercont */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup   perfcont   Performance (Boost limit) Control                */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup perfcont   Performance (Boost limit) Control
+ *  @{
+ */
 
 /**
  *  @brief Get the core boost limit.
@@ -4958,12 +4979,12 @@ amdsmi_status_t amdsmi_set_cpu_core_boostlimit(amdsmi_processor_handle processor
 amdsmi_status_t amdsmi_set_cpu_socket_boostlimit(amdsmi_processor_handle processor_handle,
                                                  uint32_t boostlimit);
 
-/** @} */
+/** @} perfcont */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup   ddrquer    DDR bandwidth monitor                            */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup ddrquer    DDR bandwidth monitor
+ *  @{
+ */
 
 /**
  *  @brief Get the DDR bandwidth data.
@@ -4978,12 +4999,12 @@ amdsmi_status_t amdsmi_set_cpu_socket_boostlimit(amdsmi_processor_handle process
 amdsmi_status_t amdsmi_get_cpu_ddr_bw(amdsmi_processor_handle processor_handle,
                                       amdsmi_ddr_bw_metrics_t *ddr_bw);
 
-/** @} */
+/** @} ddrquer */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup  tempquer   Temperature Query                                 */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup  tempquer   Temperature Query
+ *  @{
+ */
 
 /**
  *  @brief Get socket temperature.
@@ -4999,12 +5020,12 @@ amdsmi_status_t amdsmi_get_cpu_ddr_bw(amdsmi_processor_handle processor_handle,
 amdsmi_status_t amdsmi_get_cpu_socket_temperature(amdsmi_processor_handle processor_handle,
                                                   uint32_t *ptmon);
 
-/** @} */
+/** @} tempquer */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup  dimmstatistics   Dimm statistics                             */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup  dimmstatistics   Dimm statistics
+ *  @{
+ */
 
 /**
  *  @brief Get DIMM temperature range and refresh rate.
@@ -5051,12 +5072,12 @@ amdsmi_status_t amdsmi_get_cpu_dimm_thermal_sensor(amdsmi_processor_handle proce
                                                    uint8_t dimm_addr,
                                                    amdsmi_dimm_thermal_t *dimm_temp);
 
-/** @} */
+/** @} dimmstatistics */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup xgmibwcont     xGMI bandwidth control                                 */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup xgmibwcont     xGMI bandwidth control
+ *  @{
+ */
 
 /**
  *  @brief Set xgmi width.
@@ -5071,12 +5092,12 @@ amdsmi_status_t amdsmi_get_cpu_dimm_thermal_sensor(amdsmi_processor_handle proce
 amdsmi_status_t amdsmi_set_cpu_xgmi_width(amdsmi_processor_handle processor_handle,
                                           uint8_t min, uint8_t max);
 
-/** @} */
+/** @} xgmibwcont */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup gmi3widthcont     GMI3 width control                          */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup gmi3widthcont     GMI3 width control
+ *  @{
+ */
 
 /**
  *  @brief Set gmi3 link width range.
@@ -5092,12 +5113,12 @@ amdsmi_status_t amdsmi_set_cpu_xgmi_width(amdsmi_processor_handle processor_hand
 amdsmi_status_t amdsmi_set_cpu_gmi3_link_width_range(amdsmi_processor_handle processor_handle,
                                                      uint8_t min_link_width, uint8_t max_link_width);
 
-/** @} */
+/** @} gmi3widthcont */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup pstatecnt     Pstate selection                                 */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup pstatecnt     Pstate selection
+ *  @{
+ */
 
 /**
  *  @brief Enable APB.
@@ -5181,12 +5202,12 @@ amdsmi_status_t amdsmi_set_cpu_pcie_link_rate(amdsmi_processor_handle processor_
 amdsmi_status_t amdsmi_set_cpu_df_pstate_range(amdsmi_processor_handle processor_handle,
                                                uint8_t max_pstate, uint8_t min_pstate);
 
-/** @} */
+/** @} pstatecnt */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup bwquer     Bandwidth monitor                                    */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup bwquer     Bandwidth monitor
+ *  @{
+ */
 
 /**
  *  @brief Get current input output bandwidth.
@@ -5216,45 +5237,45 @@ amdsmi_status_t amdsmi_get_cpu_current_io_bandwidth(amdsmi_processor_handle proc
 amdsmi_status_t amdsmi_get_cpu_current_xgmi_bw(amdsmi_processor_handle processor_handle,
                                                amdsmi_link_id_bw_type_t link, uint32_t *xgmi_bw);
 
-/** @} */
+/** @} bwquer */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup MetQuer Metrics Table                                         */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup MetQuer HSMP Metrics Table
+ *  @{
+ */
 
 /**
- *  @brief Get metrics table version
+ *  @brief Get HSMP metrics table version
  *
  *  @platform{cpu_bm}
  *
  *  @param[in]      processor_handle Cpu socket which to query
- *  @param[in,out]  metrics_version input buffer to return the metrics table version.
+ *  @param[in,out]  metrics_version input buffer to return the HSMP metrics table version.
  *
  *  @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
  */
-amdsmi_status_t amdsmi_get_metrics_table_version(amdsmi_processor_handle processor_handle,
+amdsmi_status_t amdsmi_get_hsmp_metrics_table_version(amdsmi_processor_handle processor_handle,
                                                  uint32_t *metrics_version);
 
 /**
- *  @brief Get metrics table
+ *  @brief Get HSMP metrics table
  *
  *  @platform{cpu_bm}
  *
  *  @param[in]      processor_handle Cpu socket which to query
- *  @param[in,out]  metrics_table input buffer to return the metrics table.
+ *  @param[in,out]  metrics_table input buffer to return the HSMP metrics table.
  *
  *  @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
  */
-amdsmi_status_t amdsmi_get_metrics_table(amdsmi_processor_handle processor_handle,
-                                         amdsmi_hsmp_metric_table_t *metrics_table);
+amdsmi_status_t amdsmi_get_hsmp_metrics_table(amdsmi_processor_handle processor_handle,
+                                         amdsmi_hsmp_metrics_table_t *metrics_table);
 
-/** @} */
+/** @} MetQuer */
 
-/*---------------------------------------------------------------------------*/
-/**  @defgroup auxiquer     Auxillary functions                              */
-/*---------------------------------------------------------------------------*/
-/** @{  */
+/*****************************************************************************/
+/**  @defgroup auxiquer     Auxillary functions
+ *  @{
+ */
 
 /**
  *  @brief Get first online core on socket.
@@ -5305,7 +5326,7 @@ amdsmi_status_t amdsmi_get_cpu_model(uint32_t *cpu_model);
  */
 amdsmi_status_t amdsmi_get_esmi_err_msg(amdsmi_status_t status, const char **status_string);
 #endif
-/** @} */
+/** @} auxiquer */
 #ifdef __cplusplus
 }
 #endif  // __cplusplus

projects/amdsmi/py-interface/README.md

@@ -465,32 +465,33 @@ except AmdSmiException as e:
 
 ### amdsmi_get_gpu_cache_info
 
-Description: Returns dictionary of cache information for the given GPU.
+Description: Returns a list of dictionaries containing cache information for the given GPU.
 
 Input parameters:
 
 * `processor_handle` device which to query
 
-Output: Dictionary of Dictionaries containing cache information
-Schema: { cache_index:
-            {
-            cache_properties:
-                {
-                    "type": "array",
-                    "items": {
-                         "type": "string"
-                    }
-                },
-            cache_size: {"type" : "number"},
-            cache_level: {"type" : "number"},
-            max_num_cu_shared: {"type" : "number"},
-            num_cache_instance: {"type" : "number"}
-            }
-        }
+Output: List of Dictionaries containing cache information following the schema below:
+Schema:
+
+```JSON
+{
+    cache: {"type" : "number"},
+    cache_properties:  
+        {
+            "type" : "array",
+            "items" : {"type" : "string"}
+        },
+    cache_size: {"type" : "number"},
+    cache_level: {"type" : "number"},
+    max_num_cu_shared: {"type" : "number"},
+    num_cache_instance: {"type" : "number"}
+}
+```
 
 Field | Description
 ---|---
-`cache_index` | cache index is a string of format "cache_#" up to 10 caches will be available
+`cache` | cache index from 0-9
 `cache_properties` | list of up to 4 cache property type strings. Ex. data ("DATA_CACHE"), instruction ("INST_CACHE"), CPU ("CPU_CACHE"), or SIMD ("SIMD_CACHE").
 `cache_size` | size of cache in KB
 `cache_level` | level of cache
@@ -5958,13 +5959,13 @@ except AmdSmiException as e:
     print(e)
 ```
 
-### amdsmi_get_metrics_table_version
+### amdsmi_get_hsmp_metrics_table_version
 
-Description: Get metrics table version.
+Description: Get HSMP metrics table version.
 
-Output: amdsmi metrics table version
+Output: amdsmi HSMP metrics table version
 
-Exceptions that can be thrown by `amdsmi_get_metrics_table_version` function:
+Exceptions that can be thrown by `amdsmi_get_hsmp_metrics_table_version` function:
 
 * `AmdSmiLibraryException`
 
@@ -5977,19 +5978,19 @@ try:
         print("No CPU sockets on machine")
     else:
         for processor in processor_handles:
-            met_ver = amdsmi_get_metrics_table_version(processor)
+            met_ver = amdsmi_get_hsmp_metrics_table_version(processor)
             print(met_ver)
 except AmdSmiException as e:
     print(e)
 ```
 
-### amdsmi_get_metrics_table
+### amdsmi_get_hsmp_metrics_table
 
-Description: Get metrics table
+Description: Get HSMP metrics table
 
-Output: metric table data
+Output: HSMP metric table data
 
-Exceptions that can be thrown by `amdsmi_get_metrics_table` function:
+Exceptions that can be thrown by `amdsmi_get_hsmp_metrics_table` function:
 
 * `AmdSmiLibraryException`
 
@@ -6002,7 +6003,7 @@ try:
         print("No CPU sockets on machine")
     else:
         for processor in processor_handles:
-            mtbl = amdsmi_get_metrics_table(processor)
+            mtbl = amdsmi_get_hsmp_metrics_table(processor)
             print(mtbl['accumulation_counter'])
             print(mtbl['max_socket_temperature'])
             print(mtbl['max_vr_temperature'])

projects/amdsmi/py-interface/__init__.py

@@ -69,8 +69,8 @@ try:
     from .amdsmi_interface import amdsmi_set_cpu_df_pstate_range
     from .amdsmi_interface import amdsmi_get_cpu_current_io_bandwidth
     from .amdsmi_interface import amdsmi_get_cpu_current_xgmi_bw
-    from .amdsmi_interface import amdsmi_get_metrics_table_version
-    from .amdsmi_interface import amdsmi_get_metrics_table
+    from .amdsmi_interface import amdsmi_get_hsmp_metrics_table_version
+    from .amdsmi_interface import amdsmi_get_hsmp_metrics_table
     from .amdsmi_interface import amdsmi_first_online_core_on_cpu_socket
     from .amdsmi_interface import amdsmi_get_cpu_family
     from .amdsmi_interface import amdsmi_get_cpu_model

projects/amdsmi/py-interface/amdsmi_interface.py

@@ -1349,7 +1349,7 @@ def amdsmi_get_cpu_current_xgmi_bw(
 
     return f"{xgmi_bw.value} Mbps"
 
-def amdsmi_get_metrics_table_version(
+def amdsmi_get_hsmp_metrics_table_version(
     processor_handle: amdsmi_wrapper.amdsmi_processor_handle
 ):
     if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
@@ -1360,7 +1360,7 @@ def amdsmi_get_metrics_table_version(
     metric_tbl_version = ctypes.c_uint32()
 
     _check_res(
-        amdsmi_wrapper.amdsmi_get_metrics_table_version(
+        amdsmi_wrapper.amdsmi_get_hsmp_metrics_table_version(
             processor_handle, ctypes.byref(metric_tbl_version))
     )
 
@@ -1393,7 +1393,7 @@ def check_msb_64(num):
     else:
         return num
 
-def amdsmi_get_metrics_table(
+def amdsmi_get_hsmp_metrics_table(
     processor_handle: amdsmi_wrapper.amdsmi_processor_handle
 ):
     if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
@@ -1401,7 +1401,7 @@ def amdsmi_get_metrics_table(
             processor_handle, amdsmi_wrapper.amdsmi_processor_handle
         )
 
-    mtbl = amdsmi_wrapper.amdsmi_hsmp_metric_table_t()
+    mtbl = amdsmi_wrapper.amdsmi_hsmp_metrics_table_t()
 
     '''Encodings for the metric table defined for hsmp'''
     fraction_q10 = 1 / math.pow(2, 10)
@@ -1409,7 +1409,7 @@ def amdsmi_get_metrics_table(
     fraction_uq16 = 1 / math.pow(2, 16)
 
     _check_res(
-            amdsmi_wrapper.amdsmi_get_metrics_table(
+            amdsmi_wrapper.amdsmi_get_hsmp_metrics_table(
                    processor_handle, mtbl
             )
     )
@@ -1562,23 +1562,50 @@ def amdsmi_get_gpu_asic_info(
             processor_handle, amdsmi_wrapper.amdsmi_processor_handle
         )
 
-    asic_info = amdsmi_wrapper.amdsmi_asic_info_t()
+    asic_info_struct = amdsmi_wrapper.amdsmi_asic_info_t()
     _check_res(
         amdsmi_wrapper.amdsmi_get_gpu_asic_info(
-            processor_handle, ctypes.byref(asic_info))
+            processor_handle, ctypes.byref(asic_info_struct))
     )
 
-    return {
-        "market_name": asic_info.market_name.decode("utf-8"),
-        "vendor_id": asic_info.vendor_id,
-        "vendor_name": asic_info.vendor_name.decode("utf-8"),
-        "subvendor_id": asic_info.subvendor_id if asic_info.subvendor_id == '' else hex(asic_info.subvendor_id),
-        "device_id": asic_info.device_id,
-        "rev_id": asic_info.rev_id,
-        "asic_serial": asic_info.asic_serial.decode("utf-8"),
-        "oam_id": asic_info.oam_id
+    asic_info = {
+        "market_name": asic_info_struct.market_name.decode("utf-8"),
+        "vendor_id": asic_info_struct.vendor_id,
+        "vendor_name": asic_info_struct.vendor_name.decode("utf-8"),
+        "subvendor_id": asic_info_struct.subvendor_id,
+        "device_id": asic_info_struct.device_id,
+        "rev_id": asic_info_struct.rev_id,
+        "asic_serial": asic_info_struct.asic_serial.decode("utf-8"),
+        "oam_id": asic_info_struct.oam_id
     }
 
+    string_values = ["market_name", "vendor_name"]
+    for value in string_values:
+        if not asic_info[value]:
+            asic_info[value] = "N/A"
+
+    hex_values = ["vendor_id", "subvendor_id", "device_id", "rev_id"]
+    for value in hex_values:
+        if asic_info[value]:
+            asic_info[value] = hex(asic_info[value])
+        else:
+            asic_info[value] = "N/A"
+
+    # Ensure hex output for asic_serial
+    if asic_info["asic_serial"]:
+        asic_info["asic_serial"] = str.format("0x{:016X}", int(asic_info["asic_serial"], base=16))
+    else:
+        asic_info["asic_serial"] = "N/A"
+
+    # Check for max value as a sign for not applicable
+    if asic_info["oam_id"] == 0xFFFF: # uint 16 max
+        asic_info["oam_id"] = "N/A"
+
+    # Remove commas from vendor name for clean output
+    asic_info["vendor_name"] = asic_info["vendor_name"].replace(',', '')
+
+    return asic_info
+
 
 def amdsmi_get_power_cap_info(
     processor_handle: amdsmi_wrapper.amdsmi_processor_handle,
@@ -1625,51 +1652,52 @@ def amdsmi_get_gpu_vram_info(
 
 def amdsmi_get_gpu_cache_info(
     processor_handle: amdsmi_wrapper.amdsmi_processor_handle,
-) -> Dict[str, Dict[str, Any]]:
+) -> List[Dict[str, Any]]:
     if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
         raise AmdSmiParameterException(
             processor_handle, amdsmi_wrapper.amdsmi_processor_handle
         )
 
-    cache_info = amdsmi_wrapper.amdsmi_gpu_cache_info_t()
+    cache_info_struct = amdsmi_wrapper.amdsmi_gpu_cache_info_t()
     _check_res(
         amdsmi_wrapper.amdsmi_get_gpu_cache_info(
-            processor_handle, ctypes.byref(cache_info))
+            processor_handle, ctypes.byref(cache_info_struct))
     )
 
-    cache_info_dict = {}
-    for cache_index in range(cache_info.num_cache_types):
+    cache_info_list = []
+    for cache_index in range(cache_info_struct.num_cache_types):
         # Put cache_properties at the start of the dictionary for readability
         cache_dict = {
-            "cache_properties": [],
-            "cache_size": cache_info.cache[cache_index].cache_size,
-            "cache_level": cache_info.cache[cache_index].cache_level,
-            "max_num_cu_shared": cache_info.cache[cache_index].max_num_cu_shared,
-            "num_cache_instance": cache_info.cache[cache_index].num_cache_instance
+            "cache": cache_index,
+            "cache_properties": [], # This will be a list of strings
+            "cache_size": cache_info_struct.cache[cache_index].cache_size,
+            "cache_level": cache_info_struct.cache[cache_index].cache_level,
+            "max_num_cu_shared": cache_info_struct.cache[cache_index].max_num_cu_shared,
+            "num_cache_instance": cache_info_struct.cache[cache_index].num_cache_instance
         }
 
         # Check against cache properties bitmask
-        cache_properties = cache_info.cache[cache_index].properties
-        data_cache = cache_properties & amdsmi_wrapper.CACHE_PROPERTIES_DATA_CACHE
-        inst_cache = cache_properties & amdsmi_wrapper.CACHE_PROPERTIES_INST_CACHE
-        cpu_cache = cache_properties & amdsmi_wrapper.CACHE_PROPERTIES_CPU_CACHE
-        simd_cache = cache_properties & amdsmi_wrapper.CACHE_PROPERTIES_SIMD_CACHE
+        cache_properties = cache_info_struct.cache[cache_index].cache_properties
+        data_cache = cache_properties & amdsmi_wrapper.AMDSMI_CACHE_PROPERTIES_DATA_CACHE
+        inst_cache = cache_properties & amdsmi_wrapper.AMDSMI_CACHE_PROPERTIES_INST_CACHE
+        cpu_cache = cache_properties & amdsmi_wrapper.AMDSMI_CACHE_PROPERTIES_CPU_CACHE
+        simd_cache = cache_properties & amdsmi_wrapper.AMDSMI_CACHE_PROPERTIES_SIMD_CACHE
 
         cache_properties_status = [data_cache, inst_cache, cpu_cache, simd_cache]
         cache_property_list = []
         for cache_property in cache_properties_status:
             if cache_property:
                 property_name = amdsmi_wrapper.amdsmi_cache_properties_type_t__enumvalues[cache_property]
-                property_name = property_name.replace("CACHE_PROPERTIES_", "")
+                property_name = property_name.replace("AMDSMI_CACHE_PROPERTIES_", "")
                 cache_property_list.append(property_name)
 
         cache_dict["cache_properties"] = cache_property_list
-        cache_info_dict[f"cache_{cache_index}"] = cache_dict
+        cache_info_list.append(cache_dict)
 
-    if not cache_info_dict:
+    if not cache_info_list:
         raise AmdSmiLibraryException(amdsmi_wrapper.AMDSMI_STATUS_NO_DATA)
 
-    return cache_info_dict
+    return cache_info_list
 
 
 def amdsmi_get_gpu_vbios_info(

projects/amdsmi/py-interface/amdsmi_wrapper.py

@@ -746,19 +746,6 @@ amdsmi_card_form_factor_t = ctypes.c_uint32 # enum
 class struct_amdsmi_pcie_info_t(Structure):
     pass
 
-class struct_pcie_static_(Structure):
-    pass
-
-struct_pcie_static_._pack_ = 1 # source:False
-struct_pcie_static_._fields_ = [
-    ('max_pcie_lanes', ctypes.c_uint16),
-    ('PADDING_0', ctypes.c_ubyte * 2),
-    ('max_pcie_speed', ctypes.c_uint32),
-    ('pcie_interface_version', ctypes.c_uint32),
-    ('slot_type', amdsmi_card_form_factor_t),
-    ('reserved', ctypes.c_uint64 * 10),
-]
-
 class struct_pcie_metric_(Structure):
     pass
 
@@ -777,6 +764,19 @@ struct_pcie_metric_._fields_ = [
     ('reserved', ctypes.c_uint64 * 13),
 ]
 
+class struct_pcie_static_(Structure):
+    pass
+
+struct_pcie_static_._pack_ = 1 # source:False
+struct_pcie_static_._fields_ = [
+    ('max_pcie_lanes', ctypes.c_uint16),
+    ('PADDING_0', ctypes.c_ubyte * 2),
+    ('max_pcie_speed', ctypes.c_uint32),
+    ('pcie_interface_version', ctypes.c_uint32),
+    ('slot_type', amdsmi_card_form_factor_t),
+    ('reserved', ctypes.c_uint64 * 10),
+]
+
 struct_amdsmi_pcie_info_t._pack_ = 1 # source:False
 struct_amdsmi_pcie_info_t._fields_ = [
     ('pcie_static', struct_pcie_static_),
@@ -815,17 +815,17 @@ amdsmi_vbios_info_t = struct_amdsmi_vbios_info_t
 
 # values for enumeration 'amdsmi_cache_properties_type_t'
 amdsmi_cache_properties_type_t__enumvalues = {
-    1: 'CACHE_PROPERTIES_ENABLED',
-    2: 'CACHE_PROPERTIES_DATA_CACHE',
-    4: 'CACHE_PROPERTIES_INST_CACHE',
-    8: 'CACHE_PROPERTIES_CPU_CACHE',
-    16: 'CACHE_PROPERTIES_SIMD_CACHE',
+    1: 'AMDSMI_CACHE_PROPERTIES_ENABLED',
+    2: 'AMDSMI_CACHE_PROPERTIES_DATA_CACHE',
+    4: 'AMDSMI_CACHE_PROPERTIES_INST_CACHE',
+    8: 'AMDSMI_CACHE_PROPERTIES_CPU_CACHE',
+    16: 'AMDSMI_CACHE_PROPERTIES_SIMD_CACHE',
 }
-CACHE_PROPERTIES_ENABLED = 1
-CACHE_PROPERTIES_DATA_CACHE = 2
-CACHE_PROPERTIES_INST_CACHE = 4
-CACHE_PROPERTIES_CPU_CACHE = 8
-CACHE_PROPERTIES_SIMD_CACHE = 16
+AMDSMI_CACHE_PROPERTIES_ENABLED = 1
+AMDSMI_CACHE_PROPERTIES_DATA_CACHE = 2
+AMDSMI_CACHE_PROPERTIES_INST_CACHE = 4
+AMDSMI_CACHE_PROPERTIES_CPU_CACHE = 8
+AMDSMI_CACHE_PROPERTIES_SIMD_CACHE = 16
 amdsmi_cache_properties_type_t = ctypes.c_uint32 # enum
 class struct_amdsmi_gpu_cache_info_t(Structure):
     pass
@@ -835,9 +835,9 @@ class struct_cache_(Structure):
 
 struct_cache_._pack_ = 1 # source:False
 struct_cache_._fields_ = [
+    ('cache_properties', ctypes.c_uint32),
     ('cache_size', ctypes.c_uint32),
     ('cache_level', ctypes.c_uint32),
-    ('properties', ctypes.c_uint32),
     ('max_num_cu_shared', ctypes.c_uint32),
     ('num_cache_instance', ctypes.c_uint32),
     ('reserved', ctypes.c_uint32 * 3),
@@ -1780,11 +1780,11 @@ struct_amdsmi_dpm_level_t._fields_ = [
 ]
 
 amdsmi_dpm_level_t = struct_amdsmi_dpm_level_t
-class struct_amdsmi_hsmp_metric_table_t(Structure):
+class struct_amdsmi_hsmp_metrics_table_t(Structure):
     pass
 
-struct_amdsmi_hsmp_metric_table_t._pack_ = 1 # source:True
-struct_amdsmi_hsmp_metric_table_t._fields_ = [
+struct_amdsmi_hsmp_metrics_table_t._pack_ = 1 # source:True
+struct_amdsmi_hsmp_metrics_table_t._fields_ = [
     ('accumulation_counter', ctypes.c_uint32),
     ('max_socket_temperature', ctypes.c_uint32),
     ('max_vr_temperature', ctypes.c_uint32),
@@ -1845,7 +1845,7 @@ struct_amdsmi_hsmp_metric_table_t._fields_ = [
     ('gfxclk_frequency', ctypes.c_uint32 * 8),
 ]
 
-amdsmi_hsmp_metric_table_t = struct_amdsmi_hsmp_metric_table_t
+amdsmi_hsmp_metrics_table_t = struct_amdsmi_hsmp_metrics_table_t
 uint64_t = ctypes.c_uint64
 amdsmi_init = _libraries['libamd_smi.so'].amdsmi_init
 amdsmi_init.restype = amdsmi_status_t
@@ -1993,6 +1993,9 @@ amdsmi_reset_gpu.argtypes = [amdsmi_processor_handle]
 amdsmi_get_gpu_od_volt_info = _libraries['libamd_smi.so'].amdsmi_get_gpu_od_volt_info
 amdsmi_get_gpu_od_volt_info.restype = amdsmi_status_t
 amdsmi_get_gpu_od_volt_info.argtypes = [amdsmi_processor_handle, ctypes.POINTER(struct_amdsmi_od_volt_freq_data_t)]
+amdsmi_get_gpu_metrics_header_info = _libraries['libamd_smi.so'].amdsmi_get_gpu_metrics_header_info
+amdsmi_get_gpu_metrics_header_info.restype = amdsmi_status_t
+amdsmi_get_gpu_metrics_header_info.argtypes = [amdsmi_processor_handle, ctypes.POINTER(struct_amd_metrics_table_header_t)]
 amdsmi_get_gpu_metrics_info = _libraries['libamd_smi.so'].amdsmi_get_gpu_metrics_info
 amdsmi_get_gpu_metrics_info.restype = amdsmi_status_t
 amdsmi_get_gpu_metrics_info.argtypes = [amdsmi_processor_handle, ctypes.POINTER(struct_amdsmi_gpu_metrics_t)]
@@ -2285,12 +2288,12 @@ amdsmi_get_cpu_current_io_bandwidth.argtypes = [amdsmi_processor_handle, amdsmi_
 amdsmi_get_cpu_current_xgmi_bw = _libraries['libamd_smi.so'].amdsmi_get_cpu_current_xgmi_bw
 amdsmi_get_cpu_current_xgmi_bw.restype = amdsmi_status_t
 amdsmi_get_cpu_current_xgmi_bw.argtypes = [amdsmi_processor_handle, amdsmi_link_id_bw_type_t, ctypes.POINTER(ctypes.c_uint32)]
-amdsmi_get_metrics_table_version = _libraries['libamd_smi.so'].amdsmi_get_metrics_table_version
-amdsmi_get_metrics_table_version.restype = amdsmi_status_t
-amdsmi_get_metrics_table_version.argtypes = [amdsmi_processor_handle, ctypes.POINTER(ctypes.c_uint32)]
-amdsmi_get_metrics_table = _libraries['libamd_smi.so'].amdsmi_get_metrics_table
-amdsmi_get_metrics_table.restype = amdsmi_status_t
-amdsmi_get_metrics_table.argtypes = [amdsmi_processor_handle, ctypes.POINTER(struct_amdsmi_hsmp_metric_table_t)]
+amdsmi_get_hsmp_metrics_table_version = _libraries['libamd_smi.so'].amdsmi_get_hsmp_metrics_table_version
+amdsmi_get_hsmp_metrics_table_version.restype = amdsmi_status_t
+amdsmi_get_hsmp_metrics_table_version.argtypes = [amdsmi_processor_handle, ctypes.POINTER(ctypes.c_uint32)]
+amdsmi_get_hsmp_metrics_table = _libraries['libamd_smi.so'].amdsmi_get_hsmp_metrics_table
+amdsmi_get_hsmp_metrics_table.restype = amdsmi_status_t
+amdsmi_get_hsmp_metrics_table.argtypes = [amdsmi_processor_handle, ctypes.POINTER(struct_amdsmi_hsmp_metrics_table_t)]
 amdsmi_first_online_core_on_cpu_socket = _libraries['libamd_smi.so'].amdsmi_first_online_core_on_cpu_socket
 amdsmi_first_online_core_on_cpu_socket.restype = amdsmi_status_t
 amdsmi_first_online_core_on_cpu_socket.argtypes = [amdsmi_processor_handle, ctypes.POINTER(ctypes.c_uint32)]
@@ -2305,6 +2308,11 @@ amdsmi_get_esmi_err_msg.restype = amdsmi_status_t
 amdsmi_get_esmi_err_msg.argtypes = [amdsmi_status_t, ctypes.POINTER(ctypes.POINTER(ctypes.c_char))]
 __all__ = \
     ['AGG_BW0', 'AMDSMI_ARG_PTR_NULL', 'AMDSMI_AVERAGE_POWER',
+    'AMDSMI_CACHE_PROPERTIES_CPU_CACHE',
+    'AMDSMI_CACHE_PROPERTIES_DATA_CACHE',
+    'AMDSMI_CACHE_PROPERTIES_ENABLED',
+    'AMDSMI_CACHE_PROPERTIES_INST_CACHE',
+    'AMDSMI_CACHE_PROPERTIES_SIMD_CACHE',
     'AMDSMI_CARD_FORM_FACTOR_OAM', 'AMDSMI_CARD_FORM_FACTOR_PCIE',
     'AMDSMI_CARD_FORM_FACTOR_UNKNOWN', 'AMDSMI_CNTR_CMD_START',
     'AMDSMI_CNTR_CMD_STOP', 'AMDSMI_COARSE_GRAIN_GFX_ACTIVITY',
@@ -2421,29 +2429,27 @@ __all__ = \
     'AMDSMI_VRAM_VENDOR__WINBOND', 'AMDSMI_XGMI_STATUS_ERROR',
     'AMDSMI_XGMI_STATUS_MULTIPLE_ERRORS',
     'AMDSMI_XGMI_STATUS_NO_ERRORS', 'AMD_APU', 'AMD_CPU',
-    'AMD_CPU_CORE', 'AMD_GPU', 'CACHE_PROPERTIES_CPU_CACHE',
-    'CACHE_PROPERTIES_DATA_CACHE', 'CACHE_PROPERTIES_ENABLED',
-    'CACHE_PROPERTIES_INST_CACHE', 'CACHE_PROPERTIES_SIMD_CACHE',
-    'CLK_TYPE_DCEF', 'CLK_TYPE_DCLK0', 'CLK_TYPE_DCLK1',
-    'CLK_TYPE_DF', 'CLK_TYPE_FIRST', 'CLK_TYPE_GFX', 'CLK_TYPE_MEM',
-    'CLK_TYPE_PCIE', 'CLK_TYPE_SOC', 'CLK_TYPE_SYS', 'CLK_TYPE_VCLK0',
-    'CLK_TYPE_VCLK1', 'CLK_TYPE__MAX', 'COMPUTE_PARTITION_CPX',
-    'COMPUTE_PARTITION_DPX', 'COMPUTE_PARTITION_INVALID',
-    'COMPUTE_PARTITION_QPX', 'COMPUTE_PARTITION_SPX',
-    'COMPUTE_PARTITION_TPX', 'CONTAINER_DOCKER', 'CONTAINER_LXC',
-    'FW_ID_ASD', 'FW_ID_CP_CE', 'FW_ID_CP_ME', 'FW_ID_CP_MEC1',
-    'FW_ID_CP_MEC2', 'FW_ID_CP_MEC_JT1', 'FW_ID_CP_MEC_JT2',
-    'FW_ID_CP_MES', 'FW_ID_CP_PFP', 'FW_ID_CP_PM4', 'FW_ID_DFC',
-    'FW_ID_DMCU', 'FW_ID_DMCU_ERAM', 'FW_ID_DMCU_ISR',
-    'FW_ID_DRV_CAP', 'FW_ID_FIRST', 'FW_ID_IMU_DRAM',
-    'FW_ID_IMU_IRAM', 'FW_ID_ISP', 'FW_ID_MC', 'FW_ID_MES_KIQ',
-    'FW_ID_MES_STACK', 'FW_ID_MES_THREAD1', 'FW_ID_MES_THREAD1_STACK',
-    'FW_ID_MMSCH', 'FW_ID_PM', 'FW_ID_PPTABLE', 'FW_ID_PSP_BL',
-    'FW_ID_PSP_DBG', 'FW_ID_PSP_INTF', 'FW_ID_PSP_KEYDB',
-    'FW_ID_PSP_SOC', 'FW_ID_PSP_SOSDRV', 'FW_ID_PSP_SPL',
-    'FW_ID_PSP_SYSDRV', 'FW_ID_PSP_TOC', 'FW_ID_REG_ACCESS_WHITELIST',
-    'FW_ID_RLC', 'FW_ID_RLCV_LX7', 'FW_ID_RLC_P',
-    'FW_ID_RLC_RESTORE_LIST_CNTL', 'FW_ID_RLC_RESTORE_LIST_GPM_MEM',
+    'AMD_CPU_CORE', 'AMD_GPU', 'CLK_TYPE_DCEF', 'CLK_TYPE_DCLK0',
+    'CLK_TYPE_DCLK1', 'CLK_TYPE_DF', 'CLK_TYPE_FIRST', 'CLK_TYPE_GFX',
+    'CLK_TYPE_MEM', 'CLK_TYPE_PCIE', 'CLK_TYPE_SOC', 'CLK_TYPE_SYS',
+    'CLK_TYPE_VCLK0', 'CLK_TYPE_VCLK1', 'CLK_TYPE__MAX',
+    'COMPUTE_PARTITION_CPX', 'COMPUTE_PARTITION_DPX',
+    'COMPUTE_PARTITION_INVALID', 'COMPUTE_PARTITION_QPX',
+    'COMPUTE_PARTITION_SPX', 'COMPUTE_PARTITION_TPX',
+    'CONTAINER_DOCKER', 'CONTAINER_LXC', 'FW_ID_ASD', 'FW_ID_CP_CE',
+    'FW_ID_CP_ME', 'FW_ID_CP_MEC1', 'FW_ID_CP_MEC2',
+    'FW_ID_CP_MEC_JT1', 'FW_ID_CP_MEC_JT2', 'FW_ID_CP_MES',
+    'FW_ID_CP_PFP', 'FW_ID_CP_PM4', 'FW_ID_DFC', 'FW_ID_DMCU',
+    'FW_ID_DMCU_ERAM', 'FW_ID_DMCU_ISR', 'FW_ID_DRV_CAP',
+    'FW_ID_FIRST', 'FW_ID_IMU_DRAM', 'FW_ID_IMU_IRAM', 'FW_ID_ISP',
+    'FW_ID_MC', 'FW_ID_MES_KIQ', 'FW_ID_MES_STACK',
+    'FW_ID_MES_THREAD1', 'FW_ID_MES_THREAD1_STACK', 'FW_ID_MMSCH',
+    'FW_ID_PM', 'FW_ID_PPTABLE', 'FW_ID_PSP_BL', 'FW_ID_PSP_DBG',
+    'FW_ID_PSP_INTF', 'FW_ID_PSP_KEYDB', 'FW_ID_PSP_SOC',
+    'FW_ID_PSP_SOSDRV', 'FW_ID_PSP_SPL', 'FW_ID_PSP_SYSDRV',
+    'FW_ID_PSP_TOC', 'FW_ID_REG_ACCESS_WHITELIST', 'FW_ID_RLC',
+    'FW_ID_RLCV_LX7', 'FW_ID_RLC_P', 'FW_ID_RLC_RESTORE_LIST_CNTL',
+    'FW_ID_RLC_RESTORE_LIST_GPM_MEM',
     'FW_ID_RLC_RESTORE_LIST_SRM_MEM', 'FW_ID_RLC_SAVE_RESTORE_LIST',
     'FW_ID_RLC_SRLG', 'FW_ID_RLC_SRLS', 'FW_ID_RLC_V', 'FW_ID_RLX6',
     'FW_ID_RLX6_CORE1', 'FW_ID_RLX6_DRAM_BOOT',
@@ -2527,6 +2533,7 @@ __all__ = \
     'amdsmi_get_gpu_id', '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',
     'amdsmi_get_gpu_metrics_info',
     'amdsmi_get_gpu_od_volt_curve_regions',
     'amdsmi_get_gpu_od_volt_info', 'amdsmi_get_gpu_overdrive_level',
@@ -2544,9 +2551,9 @@ __all__ = \
     'amdsmi_get_gpu_total_ecc_count', 'amdsmi_get_gpu_vbios_info',
     'amdsmi_get_gpu_vendor_name', 'amdsmi_get_gpu_volt_metric',
     'amdsmi_get_gpu_vram_info', 'amdsmi_get_gpu_vram_usage',
-    'amdsmi_get_gpu_vram_vendor', 'amdsmi_get_lib_version',
-    'amdsmi_get_link_metrics', 'amdsmi_get_metrics_table',
-    'amdsmi_get_metrics_table_version',
+    'amdsmi_get_gpu_vram_vendor', 'amdsmi_get_hsmp_metrics_table',
+    'amdsmi_get_hsmp_metrics_table_version', 'amdsmi_get_lib_version',
+    'amdsmi_get_link_metrics',
     'amdsmi_get_minmax_bandwidth_between_processors',
     'amdsmi_get_pcie_info', 'amdsmi_get_power_cap_info',
     'amdsmi_get_power_info',
@@ -2562,7 +2569,7 @@ __all__ = \
     'amdsmi_gpu_counter_group_supported', 'amdsmi_gpu_create_counter',
     'amdsmi_gpu_destroy_counter', 'amdsmi_gpu_metrics_t',
     'amdsmi_gpu_read_counter', 'amdsmi_gpu_xgmi_error_status',
-    'amdsmi_hsmp_metric_table_t', 'amdsmi_init',
+    'amdsmi_hsmp_metrics_table_t', 'amdsmi_init',
     'amdsmi_init_flags_t', 'amdsmi_init_gpu_event_notification',
     'amdsmi_io_bw_encoding_t', 'amdsmi_io_link_type_t',
     'amdsmi_is_P2P_accessible',
@@ -2623,7 +2630,7 @@ __all__ = \
     'struct_amdsmi_freq_volt_region_t', 'struct_amdsmi_frequencies_t',
     'struct_amdsmi_frequency_range_t', 'struct_amdsmi_fw_info_t',
     'struct_amdsmi_gpu_cache_info_t', 'struct_amdsmi_gpu_metrics_t',
-    'struct_amdsmi_hsmp_metric_table_t',
+    'struct_amdsmi_hsmp_metrics_table_t',
     'struct_amdsmi_link_id_bw_type_t', 'struct_amdsmi_link_metrics_t',
     'struct_amdsmi_name_value_t', 'struct_amdsmi_od_vddc_point_t',
     'struct_amdsmi_od_volt_curve_t',

projects/amdsmi/rocm_smi/src/rocm_smi_gpu_metrics.cc

@@ -2685,47 +2685,42 @@ rsmi_status_t Device::dev_read_gpu_metrics_header_data()
 
   // Check if/when metrics table needs to be refreshed.
   auto now_ts = actual_timestamp_in_secs();
-  if ((!m_gpu_metrics_header.m_structure_size)    ||
-      (!m_gpu_metrics_header.m_format_revision)   ||
-      (!m_gpu_metrics_header.m_content_revision)) {
-    auto op_result = readDevInfo(DevInfoTypes::kDevGpuMetrics,
-                                 sizeof(AMDGpuMetricsHeader_v1_t),
-                                 &m_gpu_metrics_header);
-    if ((status_code = ErrnoToRsmiStatus(op_result)) !=
-        rsmi_status_t::RSMI_STATUS_SUCCESS) {
-      ostrstream << __PRETTY_FUNCTION__
-                 << " | ======= end ======= "
-                 << " | Fail "
-                 << " | Device #: " << index()
-                 << " | Metric Version: " << stringfy_metrics_header(m_gpu_metrics_header)
-                 << " | Cause: readDevInfo(kDevGpuMetrics)"
-                 << " | Returning = "
-                 << getRSMIStatusString(status_code)
-                 << " Could not read Metrics Header: "
-                 << print_unsigned_int(m_gpu_metrics_header.m_structure_size)
-                 << " |";
-      LOG_ERROR(ostrstream);
-      return status_code;
-    }
-    if ((status_code = is_gpu_metrics_version_supported(m_gpu_metrics_header)) ==
-        rsmi_status_t::RSMI_STATUS_NOT_SUPPORTED) {
-      ostrstream << __PRETTY_FUNCTION__
-                 << " | ======= end ======= "
-                 << " | Fail "
-                 << " | Device #: " << index()
-                 << " | Metric Version: " << stringfy_metrics_header(m_gpu_metrics_header)
-                 << " | Cause: gpu metric file version is not supported: "
-                 << " | Returning = "
-                 << getRSMIStatusString(status_code)
-                 << " Could not read Metrics Header: "
-                 << print_unsigned_int(m_gpu_metrics_header.m_structure_size)
-                 << " |";
-      LOG_ERROR(ostrstream);
-      return status_code;
-    }
-
-    m_gpu_metrics_updated_timestamp = actual_timestamp_in_secs();
+  auto op_result = readDevInfo(DevInfoTypes::kDevGpuMetrics,
+                                sizeof(AMDGpuMetricsHeader_v1_t),
+                                &m_gpu_metrics_header);
+  if ((status_code = ErrnoToRsmiStatus(op_result)) !=
+      rsmi_status_t::RSMI_STATUS_SUCCESS) {
+    ostrstream << __PRETTY_FUNCTION__
+                << " | ======= end ======= "
+                << " | Fail "
+                << " | Device #: " << index()
+                << " | Metric Version: " << stringfy_metrics_header(m_gpu_metrics_header)
+                << " | Cause: readDevInfo(kDevGpuMetrics)"
+                << " | Returning = "
+                << getRSMIStatusString(status_code)
+                << " Could not read Metrics Header: "
+                << print_unsigned_int(m_gpu_metrics_header.m_structure_size)
+                << " |";
+    LOG_ERROR(ostrstream);
+    return status_code;
   }
+  if ((status_code = is_gpu_metrics_version_supported(m_gpu_metrics_header)) ==
+      rsmi_status_t::RSMI_STATUS_NOT_SUPPORTED) {
+    ostrstream << __PRETTY_FUNCTION__
+                << " | ======= end ======= "
+                << " | Fail "
+                << " | Device #: " << index()
+                << " | Metric Version: " << stringfy_metrics_header(m_gpu_metrics_header)
+                << " | Cause: gpu metric file version is not supported: "
+                << " | Returning = "
+                << getRSMIStatusString(status_code)
+                << " Could not read Metrics Header: "
+                << print_unsigned_int(m_gpu_metrics_header.m_structure_size)
+                << " |";
+    LOG_ERROR(ostrstream);
+    return status_code;
+  }
+  m_gpu_metrics_updated_timestamp = actual_timestamp_in_secs();
 
   ostrstream << __PRETTY_FUNCTION__
              << " | ======= end ======= "
@@ -2847,23 +2842,21 @@ rsmi_status_t Device::setup_gpu_metrics_reading()
   }
 
   //
-  // if/in case setup_gpu_metrics_reading() was called already use the same pointer
+  m_gpu_metrics_ptr.reset();
+  m_gpu_metrics_ptr = amdgpu_metrics_factory(gpu_metrics_flag_version);
   if (!m_gpu_metrics_ptr) {
-    m_gpu_metrics_ptr = amdgpu_metrics_factory(gpu_metrics_flag_version);
-    if (!m_gpu_metrics_ptr) {
-      status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
-      ostrstream << __PRETTY_FUNCTION__
-                  << " | ======= end ======= "
-                  << " | Fail "
-                  << " | Device #: " << index()
-                  << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
-                  << " | Cause: amdgpu_metrics_factory() couldn't get a valid metric object"
-                  << " | Returning = "
-                  << getRSMIStatusString(status_code)
-                  << " |";
-      LOG_ERROR(ostrstream);
-      return status_code;
-    }
+    status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
+    ostrstream << __PRETTY_FUNCTION__
+                << " | ======= end ======= "
+                << " | Fail "
+                << " | Device #: " << index()
+                << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
+                << " | Cause: amdgpu_metrics_factory() couldn't get a valid metric object"
+                << " | Returning = "
+                << getRSMIStatusString(status_code)
+                << " |";
+    LOG_ERROR(ostrstream);
+    return status_code;
   }
 
   //
@@ -2943,23 +2936,21 @@ rsmi_status_t Device::dev_log_gpu_metrics(std::ostringstream& outstream_metrics)
   //  meaning, we didn't run any queries, and just want to
   //  print all the gpu metrics content, we need to setup
   //  the environment first.
-  if (!m_gpu_metrics_ptr) {
-    status_code = setup_gpu_metrics_reading();
-    if ((status_code != rsmi_status_t::RSMI_STATUS_SUCCESS) || (!m_gpu_metrics_ptr)) {
-      // At this point we should have a valid gpu_metrics pointer.
-      status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
-      ostrstream << __PRETTY_FUNCTION__
-                  << " | ======= end ======= "
-                  << " | Fail "
-                  << " | Device #: " << index()
-                  << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
-                  << " | Cause: Couldn't get a valid metric object"
-                  << " | Returning = "
-                  << getRSMIStatusString(status_code)
-                  << " |";
-      LOG_ERROR(ostrstream);
-      return status_code;
-    }
+  status_code = setup_gpu_metrics_reading();
+  if ((status_code != rsmi_status_t::RSMI_STATUS_SUCCESS) || (!m_gpu_metrics_ptr)) {
+    // At this point we should have a valid gpu_metrics pointer.
+    status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
+    ostrstream << __PRETTY_FUNCTION__
+                << " | ======= end ======= "
+                << " | Fail "
+                << " | Device #: " << index()
+                << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
+                << " | Cause: Couldn't get a valid metric object"
+                << " | Returning = "
+                << getRSMIStatusString(status_code)
+                << " |";
+    LOG_ERROR(ostrstream);
+    return status_code;
   }
 
   //  Header info
@@ -3105,22 +3096,20 @@ rsmi_status_t Device::run_internal_gpu_metrics_query(AMDGpuMetricsUnitType_t met
   ostrstream << __PRETTY_FUNCTION__ << " | ======= start =======";
   LOG_TRACE(ostrstream);
 
-  if (!m_gpu_metrics_ptr) {
-    status_code = setup_gpu_metrics_reading();
-    if ((status_code != rsmi_status_t::RSMI_STATUS_SUCCESS) || (!m_gpu_metrics_ptr)) {
-      status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
-      ostrstream << __PRETTY_FUNCTION__
-                  << " | ======= end ======= "
-                  << " | Fail "
-                  << " | Device #: " << index()
-                  << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
-                  << " | Cause: Couldn't get a valid metric object"
-                  << " | Returning = "
-                  << getRSMIStatusString(status_code)
-                  << " |";
-      LOG_ERROR(ostrstream);
-      return status_code;
-    }
+  status_code = setup_gpu_metrics_reading();
+  if ((status_code != rsmi_status_t::RSMI_STATUS_SUCCESS) || (!m_gpu_metrics_ptr)) {
+    status_code = rsmi_status_t::RSMI_STATUS_UNEXPECTED_DATA;
+    ostrstream << __PRETTY_FUNCTION__
+                << " | ======= end ======= "
+                << " | Fail "
+                << " | Device #: " << index()
+                << " | Metric Version: " << stringfy_metrics_header(dev_get_metrics_header())
+                << " | Cause: Couldn't get a valid metric object"
+                << " | Returning = "
+                << getRSMIStatusString(status_code)
+                << " |";
+    LOG_ERROR(ostrstream);
+    return status_code;
   }
 
   // Lookup the dynamic table

projects/amdsmi/src/amd_smi/amd_smi.cc

@@ -445,20 +445,21 @@ amdsmi_status_t amdsmi_get_gpu_cache_info(
 
     info->num_cache_types = rsmi_info.num_cache_types;
     for (unsigned int i =0; i < rsmi_info.num_cache_types; i++) {
+        // convert from sysfs type to CRAT type(HSA Cache Affinity type)
+        info->cache[i].cache_properties = 0;
+        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_DATA)
+            info->cache[i].cache_properties |= AMDSMI_CACHE_PROPERTIES_DATA_CACHE;
+        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_INSTRUCTION)
+            info->cache[i].cache_properties |= AMDSMI_CACHE_PROPERTIES_INST_CACHE;
+        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_CPU)
+            info->cache[i].cache_properties |= AMDSMI_CACHE_PROPERTIES_CPU_CACHE;
+        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_HSACU)
+            info->cache[i].cache_properties |= AMDSMI_CACHE_PROPERTIES_SIMD_CACHE;
+
         info->cache[i].cache_size = rsmi_info.cache[i].cache_size_kb;
         info->cache[i].cache_level = rsmi_info.cache[i].cache_level;
         info->cache[i].max_num_cu_shared = rsmi_info.cache[i].max_num_cu_shared;
         info->cache[i].num_cache_instance = rsmi_info.cache[i].num_cache_instance;
-        // convert from sysfs type to CRAT type(HSA Cache Affinity type)
-        info->cache[i].properties = 0;
-        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_DATA)
-            info->cache[i].properties |= CACHE_PROPERTIES_DATA_CACHE;
-        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_INSTRUCTION)
-            info->cache[i].properties |= CACHE_PROPERTIES_INST_CACHE;
-        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_CPU)
-            info->cache[i].properties |= CACHE_PROPERTIES_CPU_CACHE;
-        if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_HSACU)
-            info->cache[i].properties |= CACHE_PROPERTIES_SIMD_CACHE;
     }
 
     return AMDSMI_STATUS_SUCCESS;
@@ -1108,6 +1109,17 @@ amdsmi_status_t  amdsmi_get_gpu_ecc_status(amdsmi_processor_handle processor_han
                     reinterpret_cast<rsmi_ras_err_state_t*>(state));
 }
 
+amdsmi_status_t
+amdsmi_get_gpu_metrics_header_info(amdsmi_processor_handle processor_handle,
+                amd_metrics_table_header_t *header_value)
+{
+    AMDSMI_CHECK_INIT();
+    // nullptr api supported
+
+    return rsmi_wrapper(rsmi_dev_metrics_header_info_get, processor_handle,
+                    reinterpret_cast<metrics_table_header_t*>(header_value));
+}
+
 amdsmi_status_t  amdsmi_get_gpu_metrics_info(
         amdsmi_processor_handle processor_handle,
         amdsmi_gpu_metrics_t *pgpu_metrics) {
@@ -3004,7 +3016,7 @@ amdsmi_status_t amdsmi_get_cpu_current_xgmi_bw(amdsmi_processor_handle processor
     return AMDSMI_STATUS_SUCCESS;
 }
 
-amdsmi_status_t amdsmi_get_metrics_table_version(amdsmi_processor_handle processor_handle,
+amdsmi_status_t amdsmi_get_hsmp_metrics_table_version(amdsmi_processor_handle processor_handle,
                 uint32_t *metrics_version)
 {
     amdsmi_status_t status;
@@ -3024,8 +3036,8 @@ amdsmi_status_t amdsmi_get_metrics_table_version(amdsmi_processor_handle process
     return AMDSMI_STATUS_SUCCESS;
 }
 
-amdsmi_status_t amdsmi_get_metrics_table(amdsmi_processor_handle processor_handle,
-                amdsmi_hsmp_metric_table_t *metrics_table)
+amdsmi_status_t amdsmi_get_hsmp_metrics_table(amdsmi_processor_handle processor_handle,
+                amdsmi_hsmp_metrics_table_t *metrics_table)
 {
     amdsmi_status_t status;
     struct hsmp_metric_table metrics_tbl;
@@ -3036,7 +3048,7 @@ amdsmi_status_t amdsmi_get_metrics_table(amdsmi_processor_handle processor_handl
     if (processor_handle == nullptr)
         return AMDSMI_STATUS_INVAL;
 
-    if(sizeof(amdsmi_hsmp_metric_table_t) != sizeof(struct hsmp_metric_table))
+    if(sizeof(amdsmi_hsmp_metrics_table_t) != sizeof(struct hsmp_metric_table))
         return AMDSMI_STATUS_UNEXPECTED_SIZE;
 
     amdsmi_status_t r = amdsmi_get_processor_info(processor_handle, SIZE, proc_id);
@@ -3049,7 +3061,7 @@ amdsmi_status_t amdsmi_get_metrics_table(amdsmi_processor_handle processor_handl
     if (status != AMDSMI_STATUS_SUCCESS)
         return amdsmi_errno_to_esmi_status(status);
 
-    std::memcpy(metrics_table, &metrics_tbl, sizeof(amdsmi_hsmp_metric_table_t));
+    std::memcpy(metrics_table, &metrics_tbl, sizeof(amdsmi_hsmp_metrics_table_t));
 
     return AMDSMI_STATUS_SUCCESS;
 }

projects/amdsmi/tests/amd_smi_test/functional/gpu_metrics_read.cc

@@ -349,6 +349,23 @@ void TestGpuMetricsRead::Run(void) {
           << static_cast<uint32_t>(smu.pcie_nak_rcvd_count_acc) << "\n";
           std::cout << "pcie_replay_rover_count_acc= " << std::dec
           << static_cast<uint64_t>(smu.pcie_replay_rover_count_acc) << "\n";
+
+          // Check for constant changes/refresh metrics
+          std::cout << "\n";
+          std::cout << "\t ** -> Checking metrics with constant changes ** " << "\n";
+          constexpr uint16_t kMAX_ITER_TEST = 10;
+          amdsmi_gpu_metrics_t gpu_metrics_check;
+          for (auto idx = uint16_t(1); idx <= kMAX_ITER_TEST; ++idx) {
+            amdsmi_get_gpu_metrics_info(processor_handles_[i], &gpu_metrics_check);
+            std::cout << "\t\t -> firmware_timestamp [" << idx << "/" << kMAX_ITER_TEST << "]: " << gpu_metrics_check.firmware_timestamp << "\n";
+          }
+
+          std::cout << "\n";
+          for (auto idx = uint16_t(1); idx <= kMAX_ITER_TEST; ++idx) {
+            amdsmi_get_gpu_metrics_info(processor_handles_[i], &gpu_metrics_check);
+            std::cout << "\t\t -> system_clock_counter [" << idx << "/" << kMAX_ITER_TEST << "]: " << gpu_metrics_check.system_clock_counter << "\n";
+          }
+          std::cout << "\n";
       }
     }