AI/rocm-systems
2
0
Fork 0
Código Issues Pull Requests Ações Pacotes Projetos Versões Wiki Atividade

Merge amd-dev into amd-master 20240208

Ver Código-Fonte 
Signed-off-by: guanyu12 <guanyu12@amd.com>
Change-Id: I94f867792b9ee29054ff893e34957433c1055c25
Esse commit está contido em:
guanyu12
2024-02-08 10:19:18 +08:00
pai 505b8584c6 86d025daaa
commit 12f7dd5ea2
15 arquivos alterados com 641 adições e 82 exclusões
Baixar Arquivo de Patch Baixar Arquivo de Diff Expandir todos os arquivos Colapsar todos os arquivos
amdsmi_cli/README.md
+17 -17
Ver Arquivo
@@ -656,25 +656,25 @@ VRAM:
VRAM_VENDOR: HYNIX
VRAM_SIZE_MB: 96432 MB
CACHE:
CACHE 0:
CACHE_0:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 464
CACHE 1:
CACHE_1:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 64 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 160
CACHE 2:
CACHE_2:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32768 KB
CACHE_LEVEL: 2
MAX_NUM_CU_SHARED: 304
NUM_CACHE_INSTANCE: 1
CACHE 3:
CACHE_3:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 262144 KB
CACHE_LEVEL: 3
@@ -768,25 +768,25 @@ VRAM:
VRAM_SIZE_MB: 96432 MB
CACHE:
CACHE:
CACHE 0:
CACHE_0:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 464
CACHE 1:
CACHE_1:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 64 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 160
CACHE 2:
CACHE_2:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32768 KB
CACHE_LEVEL: 2
MAX_NUM_CU_SHARED: 304
NUM_CACHE_INSTANCE: 1
CACHE 3:
CACHE_3:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 262144 KB
CACHE_LEVEL: 3
@@ -879,25 +879,25 @@ VRAM:
VRAM_SIZE_MB: 96432 MB
CACHE:
CACHE:
CACHE 0:
CACHE_0:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 464
CACHE 1:
CACHE_1:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 64 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 160
CACHE 2:
CACHE_2:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32768 KB
CACHE_LEVEL: 2
MAX_NUM_CU_SHARED: 304
NUM_CACHE_INSTANCE: 1
CACHE 3:
CACHE_3:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 262144 KB
CACHE_LEVEL: 3
@@ -989,27 +989,27 @@ VRAM:
VRAM_VENDOR: HYNIX
VRAM_SIZE_MB: 96432 MB
CACHE:
CACHE 0:
CACHE_0:
CACHE:
CACHE 0:
CACHE_0:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 32 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 464
CACHE 1:
CACHE_1:
CACHE_PROPERTIES: INST_CACHE, SIMD_CACHE
CACHE_SIZE: 64 KB
CACHE_LEVEL: 1
MAX_NUM_CU_SHARED: 2
NUM_CACHE_INSTANCE: 160
CACHE 2:
CACHE_2:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 32768 KB
CACHE_LEVEL: 2
MAX_NUM_CU_SHARED: 304
NUM_CACHE_INSTANCE: 1
CACHE 3:
CACHE_3:
CACHE_PROPERTIES: DATA_CACHE, SIMD_CACHE
CACHE_SIZE: 262144 KB
CACHE_LEVEL: 3
amdsmi_cli/amdsmi_commands.py
+12
Ver Arquivo
@@ -1969,6 +1969,18 @@ class AMDSMICommands():
if (args.cpu_metrics_table):
static_dict["metrics_table"] = {}
try:
cpu_fam = amdsmi_interface.amdsmi_get_cpu_family()
static_dict["metrics_table"]["cpu_family"] = cpu_fam
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict["metrics_table"]["cpu_family"] = "N/A"
logging.debug("Failed to get cpu family | %s", e.get_error_info())
try:
cpu_mod = amdsmi_interface.amdsmi_get_cpu_model()
static_dict["metrics_table"]["cpu_model"] = cpu_mod
except amdsmi_exception.AmdSmiLibraryException as e:
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)
static_dict["metrics_table"]["response"] = metrics_table
docs/doxygen/Doxyfile
+1 -1
Ver Arquivo
@@ -2681,4 +2681,4 @@ GENERATE_LEGEND = YES
DOT_CLEANUP = YES
ALIASES += usecase{1}="\xrefitem usecase \"Use cases\" \"Use cases\" \1"
ALIASES += platform{1}="\xrefitem platform \"Platform\" \"Platforms\" \1"
example/amd_smi_drm_example.cc
+2 -2
Ver Arquivo
@@ -314,8 +314,8 @@ int main() {
for (unsigned int i = 0 ; i < cache_info.num_cache_types; i++) {
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_kb,
cache_info.cache[i].flags);
cache_info.cache[i].cache_size,
cache_info.cache[i].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);
include/amd_smi/amdsmi.h
+336 -9
Ver Arquivo
Diferenças do arquivo suprimidas por serem muito extensas Carregar Diff
include/amd_smi/impl/amd_smi_system.h
+3
Ver Arquivo
@@ -75,6 +75,9 @@ class AMDSmiSystem {
amdsmi_status_t gpu_index_to_handle(uint32_t gpu_index,
amdsmi_processor_handle* processor_handle);
amdsmi_status_t get_cpu_family(uint32_t *cpu_family);
amdsmi_status_t get_cpu_model(uint32_t *cpu_model);
private:
AMDSmiSystem() : init_flag_(AMDSMI_INIT_AMD_GPUS) {}
py-interface/README.md
+59 -4
Ver Arquivo
@@ -472,11 +472,26 @@ 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"}
}
}
Field | Description
---|---
`cache_index` | cache index - up to 10 caches will be available
`cache_flags` | list of up to 4 cache property type strings. Ex. data ("DATA_CACHE"), instruction ("INST_CACHE"), CPU ("CPU_CACHE"), or SIMD ("SIMD_CACHE").
`cache_index` | cache index is a string of format "cache_#" up to 10 caches will be available
`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
`max_num_cu_shared` | max number of compute units shared
@@ -2091,7 +2106,7 @@ Output: Dictionary with fields
`average_umc_activity` | Average umc (Universal Memory Controller) activity | %
`average_mm_activity` | Average mm (multimedia) engine activity | %
`average_socket_power` | Average socket power | W
`energy_accumulator` | Energy accumulated with a 15.3 uJ resolution over 1ns | uJ
`energy_accumulator` | Energy accumulated with a 15.3 uJ resolution over 1ns | uJ
`system_clock_counter` | System clock counter | ns
`average_gfxclk_frequency` | Average gfx clock frequency | MHz
`average_socclk_frequency` | Average soc clock frequency | MHz
@@ -2119,7 +2134,7 @@ Output: Dictionary with fields
`voltage_soc` | soc voltage | mV
`voltage_gfx` | gfx voltage | mV
`voltage_mem` | mem voltage | mV
`indep_throttle_status` | ASIC independent throttle status (see drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h for bit flags) |
`indep_throttle_status` | ASIC independent throttle status (see drivers/gpu/drm/amd/pm/swsmu/inc/amdgpu_smu.h for bit flags) |
`current_socket_power` | Current socket power (also known as instant socket power) | W
`vcn_activity` | List of VCN encode/decode engine utilization per AID | %
`gfxclk_lock_status` | Clock lock status. Each bit corresponds to clock instance. |
@@ -6023,3 +6038,43 @@ try:
except AmdSmiException as e:
print(e)
```
### amdsmi_get_cpu_family
Description: Get cpu family.
Output: cpu family
Exceptions that can be thrown by `amdsmi_get_cpu_family` function:
* `AmdSmiLibraryException`
Example:
```python
try:
cpu_family = amdsmi_get_cpu_family()
print(cpu_family)
except AmdSmiException as e:
print(e)
```
### amdsmi_get_cpu_model
Description: Get cpu model.
Output: cpu model
Exceptions that can be thrown by `amdsmi_get_cpu_model` function:
* `AmdSmiLibraryException`
Example:
```python
try:
cpu_model = amdsmi_get_cpu_model()
print(cpu_model)
except AmdSmiException as e:
print(e)
```
py-interface/__init__.py
+2
Ver Arquivo
@@ -72,6 +72,8 @@ try:
from .amdsmi_interface import amdsmi_get_metrics_table_version
from .amdsmi_interface import amdsmi_get_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
except AttributeError:
pass
py-interface/amdsmi_interface.py
+40 -18
Ver Arquivo
@@ -29,6 +29,7 @@ from . import amdsmi_wrapper
from .amdsmi_exception import *
import sys
import math
from time import localtime, asctime, time
MAX_NUM_PROCESSES = 1024
@@ -1413,6 +1414,10 @@ def amdsmi_get_metrics_table(
)
)
rawtime = int(mtbl.timestamp)
rawtime = time()
timeinfo = localtime(rawtime)
return {
"mtbl_accumulation_counter": mtbl.accumulation_counter,
"mtbl_max_socket_temperature": f"{round(check_msb_32(mtbl.max_socket_temperature) * fraction_q10 ,3)} °C",
@@ -1425,6 +1430,7 @@ def amdsmi_get_metrics_table(
"mtbl_max_socket_power_limit": f"{round(mtbl.max_socket_power_limit * fraction_uq10 ,3)} W",
"mtbl_socket_power": f"{round(mtbl.socket_power * fraction_uq10 ,3)} W",
"mtbl_timestamp_raw": mtbl.timestamp,
"mtbl_timestamp_readable": f"{asctime(timeinfo)}",
"mtbl_socket_energy_acc": f"{round((mtbl.socket_energy_acc * fraction_uq16)/KILO ,3)} kJ",
"mtbl_ccd_energy_acc": f"{round((mtbl.ccd_energy_acc * fraction_uq16)/KILO ,3)} kJ",
"mtbl_xcd_energy_acc": f"{round((mtbl.xcd_energy_acc * fraction_uq16)/KILO ,3)} kJ",
@@ -1459,11 +1465,11 @@ def amdsmi_get_metrics_table(
"mtbl_xgmi_write_bandwidth_acc": f"{[round(x*fraction_uq10 ,3) for x in list(mtbl.xgmi_write_bandwidth_acc)]} Gbps",
"mtbl_socket_c0_residency": f"{round(mtbl.socket_c0_residency * fraction_uq10 ,3)} %",
"mtbl_socket_gfx_busy": f"{round(mtbl.socket_gfx_busy * fraction_uq10 ,3)} %",
"mtbl_dram_bandwidth_utilization": f"{round(mtbl.dram_bandwidth_utilization * fraction_uq10 ,3)} %",
"mtbl_hbm_bandwidth_utilization": f"{round(mtbl.dram_bandwidth_utilization * fraction_uq10 ,3)} %",
"mtbl_socket_c0_residency_acc": round(mtbl.socket_c0_residency_acc * fraction_uq10 ,3),
"mtbl_socket_gfx_busy_acc": round(mtbl.socket_gfx_busy_acc * fraction_uq10 ,3),
"mtbl_dram_bandwidth_acc": f"{round(mtbl.dram_bandwidth_acc * fraction_uq10 ,3)} Gbps",
"mtbl_max_dram_bandwidth": f"{round(mtbl.max_dram_bandwidth * fraction_uq10 ,3)} Gbps",
"mtbl_hbm_bandwidth_acc": f"{round(mtbl.dram_bandwidth_acc * fraction_uq10 ,3)} Gbps",
"mtbl_max_hbm_bandwidth": f"{round(mtbl.max_dram_bandwidth * fraction_uq10 ,3)} Gbps",
"mtbl_dram_bandwidth_utilization_acc": round(mtbl.dram_bandwidth_utilization_acc * fraction_uq10 ,3),
"mtbl_pcie_bandwidth_acc": f"{[round(x*fraction_uq10 ,3) for x in list(mtbl.pcie_bandwidth_acc)]} Gbps",
"mtbl_prochot_residency_acc": mtbl.prochot_residency_acc,
@@ -1490,6 +1496,20 @@ def amdsmi_first_online_core_on_cpu_socket(
return pcore_ind.value
def amdsmi_get_cpu_family():
family = ctypes.c_uint32()
_check_res(
amdsmi_wrapper.amdsmi_get_cpu_family(ctypes.byref(family))
)
return family.value
def amdsmi_get_cpu_model():
model = ctypes.c_uint32()
_check_res(
amdsmi_wrapper.amdsmi_get_cpu_model(ctypes.byref(model))
)
return model.value
def amdsmi_init(flag=AmdSmiInitFlags.INIT_AMD_GPUS):
if not isinstance(flag, AmdSmiInitFlags):
raise AmdSmiParameterException(flag, AmdSmiInitFlags)
@@ -1619,30 +1639,32 @@ def amdsmi_get_gpu_cache_info(
cache_info_dict = {}
for cache_index in range(cache_info.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_kb,
"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_flags = cache_info.cache[cache_index].flags
data_cache = cache_flags & amdsmi_wrapper.CACHE_FLAGS_DATA_CACHE
inst_cache = cache_flags & amdsmi_wrapper.CACHE_FLAGS_INST_CACHE
cpu_cache = cache_flags & amdsmi_wrapper.CACHE_FLAGS_CPU_CACHE
simd_cache = cache_flags & amdsmi_wrapper.CACHE_FLAGS_SIMD_CACHE
# 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_flags_status = [data_cache, inst_cache, cpu_cache, simd_cache]
cache_flag_list = []
for cache_flag in cache_flags_status:
if cache_flag:
flag_name = amdsmi_wrapper.amdsmi_cache_flags_type_t__enumvalues[cache_flag]
flag_name = flag_name.replace("CACHE_FLAGS_", "")
cache_flag_list.append(flag_name)
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_", "")
cache_property_list.append(property_name)
cache_dict["cache_properties"] = cache_flag_list
cache_info_dict[f"cache {cache_index}"] = cache_dict
cache_dict["cache_properties"] = cache_property_list
cache_info_dict[f"cache_{cache_index}"] = cache_dict
if not cache_info_dict:
raise AmdSmiLibraryException(amdsmi_wrapper.AMDSMI_STATUS_NO_DATA)
py-interface/amdsmi_wrapper.py
+28 -21
Ver Arquivo
@@ -813,20 +813,20 @@ struct_amdsmi_vbios_info_t._fields_ = [
amdsmi_vbios_info_t = struct_amdsmi_vbios_info_t
# values for enumeration 'amdsmi_cache_flags_type_t'
amdsmi_cache_flags_type_t__enumvalues = {
1: 'CACHE_FLAGS_ENABLED',
2: 'CACHE_FLAGS_DATA_CACHE',
4: 'CACHE_FLAGS_INST_CACHE',
8: 'CACHE_FLAGS_CPU_CACHE',
16: 'CACHE_FLAGS_SIMD_CACHE',
# 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',
}
CACHE_FLAGS_ENABLED = 1
CACHE_FLAGS_DATA_CACHE = 2
CACHE_FLAGS_INST_CACHE = 4
CACHE_FLAGS_CPU_CACHE = 8
CACHE_FLAGS_SIMD_CACHE = 16
amdsmi_cache_flags_type_t = ctypes.c_uint32 # enum
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_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_size_kb', ctypes.c_uint32),
('cache_size', ctypes.c_uint32),
('cache_level', ctypes.c_uint32),
('flags', 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),
@@ -2294,6 +2294,12 @@ amdsmi_get_metrics_table.argtypes = [amdsmi_processor_handle, ctypes.POINTER(str
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)]
amdsmi_get_cpu_family = _libraries['libamd_smi.so'].amdsmi_get_cpu_family
amdsmi_get_cpu_family.restype = amdsmi_status_t
amdsmi_get_cpu_family.argtypes = [ctypes.POINTER(ctypes.c_uint32)]
amdsmi_get_cpu_model = _libraries['libamd_smi.so'].amdsmi_get_cpu_model
amdsmi_get_cpu_model.restype = amdsmi_status_t
amdsmi_get_cpu_model.argtypes = [ctypes.POINTER(ctypes.c_uint32)]
amdsmi_get_esmi_err_msg = _libraries['libamd_smi.so'].amdsmi_get_esmi_err_msg
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))]
@@ -2415,9 +2421,9 @@ __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_FLAGS_CPU_CACHE',
'CACHE_FLAGS_DATA_CACHE', 'CACHE_FLAGS_ENABLED',
'CACHE_FLAGS_INST_CACHE', 'CACHE_FLAGS_SIMD_CACHE',
'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',
@@ -2465,7 +2471,7 @@ __all__ = \
'VRAM_TYPE_GDDR6', 'VRAM_TYPE_HBM', 'VRAM_TYPE_UNKNOWN',
'VRAM_TYPE__MAX', 'WR_BW0', 'amd_metrics_table_header_t',
'amdsmi_asic_info_t', 'amdsmi_bdf_t', 'amdsmi_bit_field_t',
'amdsmi_board_info_t', 'amdsmi_cache_flags_type_t',
'amdsmi_board_info_t', 'amdsmi_cache_properties_type_t',
'amdsmi_card_form_factor_t', 'amdsmi_clk_info_t',
'amdsmi_clk_type_t', 'amdsmi_compute_partition_type_t',
'amdsmi_container_types_t', 'amdsmi_counter_command_t',
@@ -2490,8 +2496,9 @@ __all__ = \
'amdsmi_get_cpu_current_xgmi_bw', 'amdsmi_get_cpu_ddr_bw',
'amdsmi_get_cpu_dimm_power_consumption',
'amdsmi_get_cpu_dimm_temp_range_and_refresh_rate',
'amdsmi_get_cpu_dimm_thermal_sensor', 'amdsmi_get_cpu_fclk_mclk',
'amdsmi_get_cpu_hsmp_proto_ver', 'amdsmi_get_cpu_prochot_status',
'amdsmi_get_cpu_dimm_thermal_sensor', 'amdsmi_get_cpu_family',
'amdsmi_get_cpu_fclk_mclk', 'amdsmi_get_cpu_hsmp_proto_ver',
'amdsmi_get_cpu_model', 'amdsmi_get_cpu_prochot_status',
'amdsmi_get_cpu_pwr_svi_telemetry_all_rails',
'amdsmi_get_cpu_smu_fw_version',
'amdsmi_get_cpu_socket_c0_residency',
rocm_smi/src/rocm_smi.cc
+1 -1
Ver Arquivo
@@ -738,7 +738,7 @@ rsmi_dev_ecc_count_get(uint32_t dv_ind, rsmi_gpu_block_t block,
ret = GetDevValueVec(type, dv_ind, &val_vec);
if (val_vec.size() < 2 ) ret = RSMI_STATUS_FILE_ERROR;
if (ret == RSMI_STATUS_FILE_ERROR || val_vec.size() != 2) {
if (ret == RSMI_STATUS_FILE_ERROR) {
ss << __PRETTY_FUNCTION__ << " | ======= end ======="
<< ", GetDevValueVec() ret was RSMI_STATUS_FILE_ERROR "
<< "-> reporting RSMI_STATUS_NOT_SUPPORTED";
src/amd_smi/amd_smi.cc
+38 -7
Ver Arquivo
@@ -445,23 +445,22 @@ 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++) {
info->cache[i].cache_size_kb = rsmi_info.cache[i].cache_size_kb;
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].flags = 0;
info->cache[i].properties = 0;
if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_DATA)
info->cache[i].flags |= CACHE_FLAGS_DATA_CACHE;
info->cache[i].properties |= CACHE_PROPERTIES_DATA_CACHE;
if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_INSTRUCTION)
info->cache[i].flags |= CACHE_FLAGS_INST_CACHE;
info->cache[i].properties |= CACHE_PROPERTIES_INST_CACHE;
if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_CPU)
info->cache[i].flags |= CACHE_FLAGS_CPU_CACHE;
info->cache[i].properties |= CACHE_PROPERTIES_CPU_CACHE;
if (rsmi_info.cache[i].flags & HSA_CACHE_TYPE_HSACU)
info->cache[i].flags |= CACHE_FLAGS_SIMD_CACHE;
info->cache[i].properties |= CACHE_PROPERTIES_SIMD_CACHE;
}
return AMDSMI_STATUS_SUCCESS;
}
@@ -3082,6 +3081,38 @@ amdsmi_status_t amdsmi_first_online_core_on_cpu_socket(amdsmi_processor_handle p
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_family(uint32_t *cpu_family)
{
amdsmi_status_t status;
uint32_t family;
AMDSMI_CHECK_INIT();
status = amd::smi::AMDSmiSystem::getInstance().get_cpu_family(&family);
if (status != AMDSMI_STATUS_SUCCESS)
return amdsmi_errno_to_esmi_status(status);
*cpu_family = family;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_model(uint32_t *cpu_model)
{
amdsmi_status_t status;
uint32_t model;
AMDSMI_CHECK_INIT();
status = amd::smi::AMDSmiSystem::getInstance().get_cpu_model(&model);
if (status != AMDSMI_STATUS_SUCCESS)
return amdsmi_errno_to_esmi_status(status);
*cpu_model = model;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_esmi_err_msg(amdsmi_status_t status, const char **status_string)
{
for (auto& iter : amd::smi::esmi_status_map) {
src/amd_smi/amd_smi_system.cc
+2 -2
Ver Arquivo
@@ -56,7 +56,7 @@ namespace smi {
#define AMD_SMI_INIT_FLAG_RESRV_TEST1 0x800000000000000 //!< Reserved for test
static amdsmi_status_t get_cpu_family(uint32_t *cpu_family) {
amdsmi_status_t AMDSmiSystem::get_cpu_family(uint32_t *cpu_family) {
amdsmi_status_t ret;
ret = static_cast<amdsmi_status_t>(esmi_cpu_family_get(cpu_family));
@@ -68,7 +68,7 @@ static amdsmi_status_t get_cpu_family(uint32_t *cpu_family) {
}
static amdsmi_status_t get_cpu_model(uint32_t *cpu_model) {
amdsmi_status_t AMDSmiSystem::get_cpu_model(uint32_t *cpu_model) {
amdsmi_status_t ret;
ret = static_cast<amdsmi_status_t>(esmi_cpu_model_get(cpu_model));
src/amd_smi/amd_smi_utils.cc
+1
Ver Arquivo
@@ -522,6 +522,7 @@ amdsmi_status_t smi_amdgpu_get_market_name_from_dev_id(uint32_t device_id, char
break;
case 0x73a1:
case 0x73ae:
case 0x73bf:
strcpy(market_name, "NAVI21");
break;
case 0x74b4:
tools/get_platform.py
+99
Ver Arquivo
@@ -0,0 +1,99 @@
import re
import argparse
def get_platforms(file_name):
# removing the new line characters
with open(file_name) as f:
lines = [line.rstrip() for line in f]
platform_map = {}
platform_lines = ""
function_line = ""
for line in lines:
if "@platform" in line:
platform_lines += line.replace("*","").lstrip()
if platform_lines =="":
continue
l = line.lstrip()
if len(l) == 0:
continue
if l[0] == '*' or l[0]== '#' or (len(l)>1 and l[0] == '/' and l[1] == '*'):
if function_line != "":
func = ' '.join(function_line.split())
platforms = re.findall(r"\{(.*?)\}", platform_lines)
for p in platforms:
if p not in platform_map:
platform_map[p]=[func]
else:
platform_map[p].append(func)
platform_lines = ""
function_line = ""
continue
function_line += line + " "
return platform_map
# string in list1 but not in list2
def diff(list1, list2):
result = []
for l1 in list1:
if l1 not in list2:
result.append(l1)
return result
def common(list1, list2):
result = []
for l1 in list1:
if l1 in list2:
result.append(l1)
return result
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='amdsmi platform classifier')
parser.add_argument('--platforms', default=False, action='store_true', help='list supported platforms')
parser.add_argument('--list', default="", metavar="platform" , help='List function in a platform')
parser.add_argument('--diff' , default=None, metavar="platform", nargs=2, help='Find the APIs in platform1 but not in platform2')
parser.add_argument('--common' , default=None, metavar="platform", nargs=2, help='Find the common APIs of two platforms')
file_name="../include/amd_smi/amdsmi.h"
args = parser.parse_args()
platform_map=get_platforms(file_name)
if args.platforms:
for p in platform_map:
print (p)
exit(0)
if args.list:
if args.list not in platform_map:
print("Unknown platform ", args.list)
exit(1)
for f in platform_map[args.list]:
print (f)
exit(0)
if args.diff != None:
if args.diff[0] not in platform_map or args.diff[1] not in platform_map:
print("Unknown platforms ", args.diff)
exit(1)
platforms0 = platform_map[args.diff[0]]
platforms1 = platform_map[args.diff[1]]
result = diff(platforms0, platforms1)
print("APIs in", args.diff[0], "but not in", args.diff[1])
for f in result:
print(f)
exit(0)
if args.common != None:
if args.common[0] not in platform_map or args.common[1] not in platform_map:
print("Unknown platforms ", args.common)
exit(1)
platforms0 = platform_map[args.common[0]]
platforms1 = platform_map[args.common[1]]
result = common(platforms0, platforms1)
print("APIs in both ", args.common[0], "and", args.common[1])
for f in result:
print(f)
exit(0)
parser.print_help()