AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge amd-staging into amd-master 20230907

Browse Source 
Signed-off-by: Hao Zhou <Hao.Zhou@amd.com>
Change-Id: I834fb22c825182a745e1ea83071e5b481825270d
This commit is contained in:
Hao Zhou
2023-09-07 10:18:14 +08:00
parent c9413dc931 489991a322
commit e98d5273df
19 changed files with 1228 additions and 591 deletions
Download Patch File Download Diff File Expand all files Collapse all files
example/amd_smi_drm_example.cc
+20 -13
View File
@@ -40,18 +40,20 @@
* DEALINGS WITH THE SOFTWARE.
*
*/
#include <assert.h>
#include <stdint.h>
#include <unistd.h>
#include "amd_smi/amdsmi.h"
#include <bitset>
#include <iostream>
#include <pwd.h>
#include <sys/stat.h>
#include <string.h>
#include <unistd.h>
#include <bitset>
#include <cassert>
#include <cstdint>
#include <cstring>
#include <iostream>
#include <vector>
#include "amd_smi/amdsmi.h"
#define CHK_AMDSMI_RET(RET) \
{ \
@@ -272,8 +274,10 @@ int main() {
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_gpu_device_bdf:\n");
printf("\tDevice[%d] BDF %04lx:%02x:%02x.%d\n\n", i,
bdf.domain_number, bdf.bus_number, bdf.device_number,
bdf.function_number);
bdf.fields.domain_number,
bdf.fields.bus_number,
bdf.fields.device_number,
bdf.fields.function_number);
// Get handle from BDF
amdsmi_processor_handle dev_handle;
@@ -407,7 +411,7 @@ int main() {
// Get temperature measurements
// amdsmi_temperature_t edge_temp, junction_temp, vram_temp,
// plx_temp;
int64_t temp_measurements[4];
int64_t temp_measurements[TEMPERATURE_TYPE__MAX + 1];
amdsmi_temperature_type_t temp_types[4] = {
TEMPERATURE_TYPE_EDGE, TEMPERATURE_TYPE_JUNCTION,
TEMPERATURE_TYPE_VRAM, TEMPERATURE_TYPE_PLX};
@@ -507,8 +511,11 @@ int main() {
uint64_t mem = 0, gtt_mem = 0, cpu_mem = 0, vram_mem = 0;
uint64_t gfx = 0, enc = 0;
char bdf_str[20];
sprintf(bdf_str, "%04lx:%02x:%02x.%d", bdf.domain_number,
bdf.bus_number, bdf.device_number, bdf.function_number);
sprintf(bdf_str, "%04lx:%02x:%02x.%d",
bdf.fields.domain_number,
bdf.fields.bus_number,
bdf.fields.device_number,
bdf.fields.function_number);
int num = 0;
ret = amdsmi_get_gpu_process_list(processor_handles[j], &num_process,
process_list);
@@ -543,7 +550,7 @@ int main() {
"==+=========================================+\n");
for (int it = 0; it < num; it++) {
char command[30];
struct passwd *pwd = NULL;
struct passwd *pwd = nullptr;
struct stat st;
sprintf(command, "/proc/%d", info_list[it].pid);
example/amd_smi_nodrm_example.cc
+15 -16
View File
@@ -40,17 +40,21 @@
* DEALINGS WITH THE SOFTWARE.
*
*/
#include <assert.h>
#include <stdint.h>
#include <unistd.h>
#include "amd_smi/amdsmi.h"
#include <bitset>
#include <iostream>
#include <pwd.h>
#include <sys/stat.h>
#include <unistd.h>
#include <bitset>
#include <cassert>
#include <cstdint>
#include <cstring>
#include <iostream>
#include <vector>
#include "amd_smi/amdsmi.h"
#define CHK_AMDSMI_RET(RET) \
{ \
if (RET != AMDSMI_STATUS_SUCCESS) { \
@@ -123,8 +127,10 @@ int main() {
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_gpu_device_bdf:\n");
printf("\tDevice[%d] BDF %04lx:%02x:%02x.%d\n\n", i,
bdf.domain_number, bdf.bus_number, bdf.device_number,
bdf.function_number);
bdf.fields.domain_number,
bdf.fields.bus_number,
bdf.fields.device_number,
bdf.fields.function_number);
amdsmi_asic_info_t asic_info = {};
ret = amdsmi_get_gpu_asic_info(processor_handles[j], &asic_info);
@@ -219,7 +225,7 @@ int main() {
.fw_version);
// Get temperature measurements
int64_t temp_measurements[4];
int64_t temp_measurements[TEMPERATURE_TYPE__MAX + 1];
amdsmi_temperature_type_t temp_types[4] = {
TEMPERATURE_TYPE_EDGE, TEMPERATURE_TYPE_JUNCTION,
TEMPERATURE_TYPE_VRAM, TEMPERATURE_TYPE_PLX};
@@ -275,13 +281,6 @@ int main() {
printf("\tCorrectable errors: %lu\n", err_cnt_info.correctable_count);
printf("\tUncorrectable errors: %lu\n\n",
err_cnt_info.uncorrectable_count);
// Get process list
auto compare = [](const void *a, const void *b) -> int {
return (*(amdsmi_proc_info_t *)a).pid >
(*(amdsmi_proc_info_t *)b).pid
? 1
: -1;
};
// Get device name
amdsmi_board_info_t board_info = {};
example/amdsmi_esmi_intg_example.cc
+44 -4
View File
@@ -59,10 +59,11 @@ using namespace std;
{ \
if (RET != AMDSMI_STATUS_SUCCESS) { \
const char *err_str; \
const char **status_str; \
std::cout << "AMDSMI call returned " << RET << " at line " \
<< __LINE__ << std::endl; \
amdsmi_status_code_to_string(RET, &err_str); \
std::cout << err_str << std::endl; \
*status_str = amdsmi_get_esmi_err_msg(RET, &err_str); \
std::cout << *status_str << std::endl; \
return RET; \
} \
}
@@ -113,14 +114,14 @@ int main(int argc, char **argv) {
CHK_AMDSMI_RET(ret)
std::cout << "core_count=" << core_count << std::endl;
ret = amdsmi_get_hsmp_proto_ver(&proto_ver);
ret = amdsmi_get_cpu_hsmp_proto_ver(sockets[i], &proto_ver);
CHK_AMDSMI_RET(ret)
cout<<"\n------------------------------------------";
cout<<"\n| HSMP Proto Version | "<< proto_ver <<"\t\t |"<< endl;
cout<<"------------------------------------------\n";
ret = amdsmi_get_smu_fw_version(&smu_fw);
ret = amdsmi_get_cpu_smu_fw_version(sockets[i], &smu_fw);
CHK_AMDSMI_RET(ret)
cout<<"\n------------------------------------------";
@@ -289,7 +290,46 @@ int main(int argc, char **argv) {
cout<<" NA (Err:" <<ret<<" |";
}
cout<<"\n-------------------------------------------------\n";
uint32_t input_power;
powermax = 0;
cout<<"\nEnter the max power to be set:\n";
cin>>input_power;
ret = amdsmi_get_cpu_socket_power_cap_max(sockets[i], i, &powermax);
CHK_AMDSMI_RET(ret)
if ((ret == AMDSMI_STATUS_SUCCESS) && (input_power > powermax)) {
cout<<"Input power is more than max power limit,"
" limiting to "<<static_cast<double>(powermax)/1000<<"Watts\n";
input_power = powermax;
}
ret = amdsmi_set_cpu_socket_power_cap(sockets[i], i, input_power);
CHK_AMDSMI_RET(ret)
if (!ret) {
cout<<"Socket["<<i<<"] power_limit set to "
<<fixed<<setprecision(3)<<static_cast<double>(input_power)/1000<<" Watts successfully\n";
}
cout<<"\n-------------------------------------------------\n";
#if 0
uint8_t mode;
const char *err_str;
cout <<"Enter the power efficiency mode to be set[0, 1 or 2]:\n";
cin>>mode;
ret = amdsmi_set_cpu_pwr_efficiency_mode(sockets[i], i, mode);
CHK_AMDSMI_RET(ret)
if (ret != AMDSMI_STATUS_SUCCESS) {
cout<<"Failed to set power efficiency mode for socket["<<i<<"], Err["
<<ret<<"]: "<<amdsmi_get_esmi_err_msg(ret, &err_str)<<"\n";
return ret;
}
if (!ret)
cout<<"Power efficiency profile policy is set to "<<mode<<"successfully\n";
cout<<"\n-------------------------------------------------\n";
uint32_t svi_power;
cout<<"\n| SVI Power Telemetry (mWatts) \t |";
include/amd_smi/amdsmi.h
+42 -17
View File
@@ -352,12 +352,12 @@ typedef struct {
} amdsmi_frequency_range_t;
typedef union {
struct {
struct fields_ {
uint64_t function_number : 3;
uint64_t device_number : 5;
uint64_t bus_number : 8;
uint64_t domain_number : 48;
};
} fields;
uint64_t as_uint;
} amdsmi_bdf_t;
@@ -380,7 +380,7 @@ typedef struct {
typedef struct {
uint8_t num_fw_info;
struct {
struct fw_info_list_ {
amdsmi_fw_block_t fw_id;
uint64_t fw_version;
uint64_t reserved[2];
@@ -391,11 +391,12 @@ typedef struct {
typedef struct {
char market_name[AMDSMI_MAX_STRING_LENGTH];
uint32_t vendor_id; //< Use 32 bit to be compatible with other platform.
char vendor_name[AMDSMI_MAX_STRING_LENGTH];
uint32_t subvendor_id; //< The subsystem vendor id
uint64_t device_id; //< The device id of a GPU
uint32_t rev_id;
char asic_serial[AMDSMI_NORMAL_STRING_LENGTH];
uint32_t reserved[3];
uint32_t reserved[19];
} amdsmi_asic_info_t;
typedef struct {
@@ -439,15 +440,15 @@ typedef struct {
typedef uint32_t amdsmi_process_handle_t;
typedef struct {
char name[AMDSMI_NORMAL_STRING_LENGTH];
amdsmi_process_handle_t pid;
uint64_t mem; /** in bytes */
struct {
uint64_t gfx;
uint64_t enc;
char name[AMDSMI_NORMAL_STRING_LENGTH];
amdsmi_process_handle_t pid;
uint64_t mem; /** in bytes */
struct engine_usage_ {
uint64_t gfx;
uint64_t enc;
uint32_t reserved[12];
} engine_usage; /** How much time the process spend using these engines in ns */
struct {
} engine_usage; /** How much time the process spend using these engines in ns */
struct memory_usage_ {
uint64_t gtt_mem;
uint64_t cpu_mem;
uint64_t vram_mem;
@@ -3685,20 +3686,24 @@ amdsmi_status_t amdsmi_get_cpu_socket_energy(amdsmi_cpusocket_handle socket_hand
/**
* @brief Get SMU Firmware Version.
*
* @param[in] socket_handle Cpu socket which to query
* @param[in,out] amdsmi_smu_fw - Input buffer to return the firmware version
*
* @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
*/
amdsmi_status_t amdsmi_get_smu_fw_version(amdsmi_smu_fw_version_t *amdsmi_smu_fw);
amdsmi_status_t amdsmi_get_cpu_smu_fw_version(amdsmi_cpusocket_handle socket_handle,
amdsmi_smu_fw_version_t *amdsmi_smu_fw);
/**
* @brief Get HSMP protocol Version.
*
* @param[in] socket_handle Cpu socket which to query
* @param[in,out] proto_ver - Input buffer to return the protocol version
*
* @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
*/
amdsmi_status_t amdsmi_get_hsmp_proto_ver(uint32_t *proto_ver);
amdsmi_status_t amdsmi_get_cpu_hsmp_proto_ver(amdsmi_cpusocket_handle socket_handle,
uint32_t *proto_ver);
/**
* @brief Get normalized status of the processor's PROCHOT status.
@@ -3938,11 +3943,13 @@ amdsmi_status_t amdsmi_set_cpu_socket_boostlimit(amdsmi_cpusocket_handle socket_
/**
* @brief Get the DDR bandwidth data.
*
* @param[in] socket_handle Cpu socket which to query
* @param[in,out] ddr_bw - Input buffer to fill ddr bandwidth data
*
* @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
*/
amdsmi_status_t amdsmi_get_cpu_ddr_bw(amdsmi_ddr_bw_metrics_t *ddr_bw);
amdsmi_status_t amdsmi_get_cpu_ddr_bw(amdsmi_cpusocket_handle socket_handle,
amdsmi_ddr_bw_metrics_t *ddr_bw);
/** @} */
@@ -4154,12 +4161,14 @@ amdsmi_status_t amdsmi_get_cpu_current_io_bandwidth(amdsmi_cpusocket_handle sock
/**
* @brief Get current input output bandwidth.
*
* @param[in] socket_handle Cpu socket which to query
* @param[in] link - link id and bw type to which xgmi bandwidth to be obtained
* @param[in,out] xgmi_bw - Input buffer to fill bandwidth data
*
* @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
*/
amdsmi_status_t amdsmi_get_cpu_current_xgmi_bw(amdsmi_link_id_bw_type_t link, uint32_t *xgmi_bw);
amdsmi_status_t amdsmi_get_cpu_current_xgmi_bw(amdsmi_cpusocket_handle socket_handle,
amdsmi_link_id_bw_type_t link, uint32_t *xgmi_bw);
/** @} */
@@ -4216,14 +4225,30 @@ amdsmi_status_t amdsmi_get_number_of_cpu_sockets(uint32_t sockets);
/**
* @brief Get first online core on socket.
*
* @param[in] socket_handle Cpu socket which to query
* @param[in] sock_ind - socket index.
*
* @param[in,out] sockets - Input buffer to fill first online core on socket data
*
* @return ::amdsmi_status_t | ::AMDSMI_STATUS_SUCCESS on success, non-zero on fail
*/
amdsmi_status_t amdsmi_first_online_core_on_cpu_socket(uint32_t sock_ind, uint32_t *pcore_ind);
amdsmi_status_t amdsmi_first_online_core_on_cpu_socket(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t *pcore_ind);
/**
* @brief Get a description of provided AMDSMI error status for esmi errors.
*
* @details Set the provided pointer to a const char *, @p status_string, to
* a string containing a description of the provided error code @p status.
*
* @param[in] status - The error status for which a description is desired.
*
* @param[in,out] status_string - A pointer to a const char * which will be made
* to point to a description of the provided error code
*
* @return const char* returned on success
*/
const char* amdsmi_get_esmi_err_msg(amdsmi_status_t status, const char **status_string);
#endif
/** @} */
#ifdef __cplusplus
py-interface/CMakeLists.txt
+5 -5
View File
@@ -1,7 +1,6 @@
# Generate py-interface and package targets
# match this version to your clang
# too new won't work, too old won't work either
# CLANG installed must be 14.0 or above
set(clang_ver 14.0)
set(ctypeslib_ver 2.3.2)
@@ -12,7 +11,6 @@ set(PY_PACKAGE_DIR "${PY_BUILD_DIR}/amdsmi")
set(PY_WRAPPER_INSTALL_DIR "${SHARE_INSTALL_PREFIX}" CACHE STRING "Wrapper installation directory")
# TODO: Figure out how python-clang and clang are related
# Currently only a very specific combination works
# try to find clang of the right version
set(GOOD_CLANG_FOUND FALSE)
@@ -22,7 +20,7 @@ if(BUILD_WRAPPER)
# extract clang version manually because find_package(clang) doesn't work
execute_process(COMMAND ${clang} --version OUTPUT_VARIABLE clang_full_version_string)
string (REGEX REPLACE ".*clang version ([0-9]+\\.[0-9]+).*" "\\1" CLANG_VERSION_STRING ${clang_full_version_string})
if(CLANG_VERSION_STRING VERSION_EQUAL clang_ver)
if((CLANG_VERSION_STRING VERSION_GREATER clang_ver) OR (CLANG_VERSION_STRING VERSION_EQUAL clang_ver))
message("GOOD CLANG VERSION: ${CLANG_VERSION_STRING}")
set(GOOD_CLANG_FOUND TRUE)
else()
@@ -47,9 +45,11 @@ if(NOT GOOD_CLANG_FOUND)
COMMAND ln -Pf ${CMAKE_CURRENT_BINARY_DIR}/amdsmi_wrapper.py ${PY_PACKAGE_DIR}/)
else()
find_package(Python3 3.7 COMPONENTS Interpreter Development REQUIRED)
# --break-system-packages is needed for python 3.11
# see: https://peps.python.org/pep-0668/
add_custom_target(
python_pre_reqs
COMMAND ${Python3_EXECUTABLE} -m pip install clang==${clang_ver} ctypeslib2==${ctypeslib_ver})
COMMAND ${Python3_EXECUTABLE} -m pip install --break-system-packages clang==${clang_ver} ctypeslib2==${ctypeslib_ver})
# generate new wrapper
configure_file(${PROJECT_SOURCE_DIR}/tools/generator.py generator.py @ONLY COPYONLY)
add_custom_command(
py-interface/amdsmi_interface.py
+10 -53
View File
@@ -445,10 +445,10 @@ def _format_bdf(amdsmi_bdf: amdsmi_wrapper.amdsmi_bdf_t) -> str:
Returns:
`str`: String containing BDF data in a readable format.
"""
domain = hex(amdsmi_bdf.c__UA_amdsmi_bdf_t_0.domain_number)[2:].zfill(4)
bus = hex(amdsmi_bdf.c__UA_amdsmi_bdf_t_0.bus_number)[2:].zfill(2)
device = hex(amdsmi_bdf.c__UA_amdsmi_bdf_t_0.device_number)[2:].zfill(2)
function = hex(amdsmi_bdf.c__UA_amdsmi_bdf_t_0.function_number)[2:]
domain = hex(amdsmi_bdf.fields.domain_number)[2:].zfill(4)
bus = hex(amdsmi_bdf.fields.bus_number)[2:].zfill(2)
device = hex(amdsmi_bdf.fields.device_number)[2:].zfill(2)
function = hex(amdsmi_bdf.fields.function_number)[2:]
return domain + ":" + bus + ":" + device + "." + function
@@ -495,10 +495,10 @@ def _make_amdsmi_bdf_from_list(bdf):
if len(bdf) != 4:
return None
amdsmi_bdf = amdsmi_wrapper.amdsmi_bdf_t()
amdsmi_bdf.c__UA_amdsmi_bdf_t_0.function_number = bdf[3]
amdsmi_bdf.c__UA_amdsmi_bdf_t_0.device_number = bdf[2]
amdsmi_bdf.c__UA_amdsmi_bdf_t_0.bus_number = bdf[1]
amdsmi_bdf.c__UA_amdsmi_bdf_t_0.domain_number = bdf[0]
amdsmi_bdf.fields.function_number = bdf[3]
amdsmi_bdf.fields.device_number = bdf[2]
amdsmi_bdf.fields.bus_number = bdf[1]
amdsmi_bdf.fields.domain_number = bdf[0]
return amdsmi_bdf
@@ -1422,32 +1422,6 @@ def amdsmi_set_gpu_perf_level(
processor_handle, perf_level))
def amdsmi_get_gpu_power_profile_presets(
processor_handle: amdsmi_wrapper.amdsmi_processor_handle, sensor_idx: int
) -> Dict[str, Any]:
if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
raise AmdSmiParameterException(
processor_handle, amdsmi_wrapper.amdsmi_processor_handle
)
if not isinstance(sensor_idx, int):
raise AmdSmiParameterException(sensor_idx, int)
sensor_idx = ctypes.c_uint32(sensor_idx)
status = amdsmi_wrapper.amdsmi_power_profile_status_t()
_check_res(
amdsmi_wrapper.amdsmi_get_gpu_power_profile_presets(
processor_handle, sensor_idx, ctypes.byref(status)
)
)
return {
"available_profiles": status.available_profiles,
"current": status.current,
"num_profiles": status.num_profiles,
}
def amdsmi_reset_gpu(processor_handle: amdsmi_wrapper.amdsmi_processor_handle):
if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
raise AmdSmiParameterException(
@@ -1457,23 +1431,6 @@ def amdsmi_reset_gpu(processor_handle: amdsmi_wrapper.amdsmi_processor_handle):
_check_res(amdsmi_wrapper.amdsmi_reset_gpu(processor_handle))
def amdsmi_set_gpu_perf_determinism_mode(
processor_handle: amdsmi_wrapper.amdsmi_processor_handle, clock_value: int
):
if not isinstance(processor_handle, amdsmi_wrapper.amdsmi_processor_handle):
raise AmdSmiParameterException(
processor_handle, amdsmi_wrapper.amdsmi_processor_handle
)
if not isinstance(clock_value, int):
raise AmdSmiParameterException(clock_value, int)
clock_value = ctypes.c_uint64(clock_value)
_check_res(
amdsmi_wrapper.amdsmi_set_gpu_perf_determinism_mode(
processor_handle, clock_value)
)
def amdsmi_set_gpu_fan_speed(
processor_handle: amdsmi_wrapper.amdsmi_processor_handle, sensor_idx: int, fan_speed: int
):
@@ -1998,7 +1955,7 @@ def amdsmi_get_utilization_count(
result = [{"timestamp": timestamp.value}]
for idx in range(count.value):
counter_type = amdsmi_wrapper.c__EA_AMDSMI_UTILIZATION_COUNTER_TYPE__enumvalues[
counter_type = amdsmi_wrapper.AMDSMI_UTILIZATION_COUNTER_TYPE__enumvalues[
util_counter_list[idx].type
]
if counter_type == "AMDSMI_UTILIZATION_COUNTER_LAST":
@@ -2024,7 +1981,7 @@ def amdsmi_get_gpu_perf_level(
processor_handle, ctypes.byref(perf))
)
result = amdsmi_wrapper.c__EA_amdsmi_dev_perf_level_t__enumvalues[perf.value]
result = amdsmi_wrapper.amdsmi_dev_perf_level_t__enumvalues[perf.value]
if result == "AMDSMI_DEV_PERF_LEVEL_FIRST":
result = "AMDSMI_DEV_PERF_LEVEL_AUTO"
if result == "AMDSMI_DEV_PERF_LEVEL_LAST":
py-interface/amdsmi_wrapper.py
+394 -453
View File
File diff suppressed because it is too large Load Diff
rocm_smi/example/rocm_smi_example.cc
+5
View File
@@ -722,6 +722,11 @@ int main() {
CHK_RSMI_RET_I(ret)
std::cout << "\t**Dev.Rev.ID: 0x" << std::hex << val_ui16 << std::endl;
char pcie_vendor_name[256];
ret = rsmi_dev_pcie_vendor_name_get(i, pcie_vendor_name, 256);
CHK_RSMI_RET_I(ret)
std::cout << "\t**PCIe vendor name: " << pcie_vendor_name << std::endl;
char current_compute_partition[256];
current_compute_partition[0] = '\0';
ret = rsmi_dev_compute_partition_get(i, current_compute_partition, 256);
rocm_smi/include/rocm_smi/rocm_smi.h
+37
View File
@@ -1155,6 +1155,43 @@ rsmi_status_t rsmi_dev_sku_get(uint32_t dv_ind, uint16_t *sku);
*/
rsmi_status_t rsmi_dev_vendor_id_get(uint32_t dv_ind, uint16_t *id);
/**
* @brief Get the name string for a give PCIe vendor ID
*
* @details Given a device index @p dv_ind, a pointer to a caller provided
* char buffer @p name, and a length of this buffer @p len, this function will
* write the name of the PCIe vendor (up to @p len characters) buffer @p name.
*
* If the integer ID associated with the PCIe vendor is not found in one of the
* system files containing device name information (e.g.
* /usr/share/misc/pci.ids), then this function will return RSMI_STATUS_NOT_FOUND.
* Updating the system name files can be accompplished with
* "sudo update-pciids".
*
* @param[in] dv_ind a device index
*
* @param[inout] name a pointer to a caller provided char buffer to which the
* name will be written
* If this parameter is nullptr, this function will return
* ::RSMI_STATUS_INVALID_ARGS if the function is supported with the provided,
* arguments and ::RSMI_STATUS_NOT_SUPPORTED if it is not supported with the
* provided arguments.
*
* @param[in] len the length of the caller provided buffer @p name.
*
* @retval ::RSMI_STATUS_SUCCESS call was successful
* @retval ::RSMI_STATUS_NOT_SUPPORTED installed software or hardware does not
* support this function with the given arguments
* @retval ::RSMI_STATUS_INVALID_ARGS the provided arguments are not valid
* @retval ::RSMI_STATUS_NOT_FOUND the vnedor name are not found
* @retval ::RSMI_STATUS_INSUFFICIENT_SIZE is returned if @p len bytes is not
* large enough to hold the entire name. In this case, only @p len bytes will
* be written.
*
*/
rsmi_status_t rsmi_dev_pcie_vendor_name_get(uint32_t dv_ind, char *name,
size_t len);
/**
* @brief Get the name string of a gpu device.
*
rocm_smi/include/rocm_smi/rocm_smi_device.h
+6 -1
View File
@@ -167,7 +167,12 @@ enum DevInfoTypes {
kDevGpuReset,
kDevAvailableComputePartition,
kDevComputePartition,
kDevMemoryPartition
kDevMemoryPartition,
// The information read from pci core sysfs
kDevPCieTypeStart = 1000,
kDevPCieVendorID = kDevPCieTypeStart,
kDevPCieTypeEND = 2000,
};
typedef struct {
rocm_smi/include/rocm_smi/rocm_smi_utils.h
+1
View File
@@ -103,6 +103,7 @@ std::tuple<bool, std::string, std::string, std::string, std::string,
getSystemDetails(void);
void logSystemDetails(void);
rsmi_status_t getBDFString(uint64_t bdf_id, std::string& bfd_str);
rsmi_status_t getBDFWithDomain(uint64_t bdf_id, std::string& bfd_str);
void logHexDump(const char *desc, const void *addr, const size_t len,
size_t perLine);
bool isSystemBigEndian();
rocm_smi/src/rocm_smi.cc
+51
View File
@@ -1979,6 +1979,28 @@ static rsmi_status_t get_dev_name_from_file(uint32_t dv_ind, char *name,
return RSMI_STATUS_SUCCESS;
}
// return empty if cannot find it in pci.ids
static std::string get_vendor_name_from_id(uint16_t vendor_id) {
for (auto fl : pci_name_files) {
std::ifstream id_file_strm(fl);
std::string ln;
while (std::getline(id_file_strm, ln)) {
// parse line
if (ln == "" || ln[0] == '#' || ln[0] == '\t') {
continue;
}
std::istringstream ln_str(ln);
std::string val_str = get_id_name_str_from_line(vendor_id, ln, &ln_str);
if (val_str != "") {
return val_str;
}
}
}
return "";
}
// Parse pci.ids files. Comment lines have # in first column. Otherwise,
// Syntax:
// vendor vendor_name
@@ -2157,6 +2179,35 @@ rsmi_dev_name_get(uint32_t dv_ind, char *name, size_t len) {
CATCH
}
rsmi_status_t
rsmi_dev_pcie_vendor_name_get(uint32_t dv_ind, char *name, size_t len) {
rsmi_status_t ret;
TRY
std::ostringstream ss;
ss << __PRETTY_FUNCTION__ << "| ======= start =======";
LOG_TRACE(ss);
CHK_SUPPORT_NAME_ONLY(name)
if (len == 0) {
return RSMI_STATUS_INVALID_ARGS;
}
DEVICE_MUTEX
uint16_t id = 0;
ret = get_id(dv_ind, amd::smi::kDevPCieVendorID, &id);
if (ret != RSMI_STATUS_SUCCESS) return ret;
std::string vendor_name = get_vendor_name_from_id(id);
if (vendor_name == "") {
return RSMI_STATUS_NOT_FOUND;
}
memset(name, 0, len);
strncpy(name, vendor_name.c_str(), len-1);
return ret;
CATCH
}
rsmi_status_t
rsmi_dev_brand_get(uint32_t dv_ind, char *brand, uint32_t len) {
TRY
rocm_smi/src/rocm_smi_device.cc
+21 -1
View File
@@ -81,6 +81,9 @@ static const char *kPathDebugRootFName = "/sys/kernel/debug/dri/";
// Device debugfs file names
static const char *kDevGpuResetFName = "amdgpu_gpu_recover";
// PCI sysfs file names
static const char *kDevPCieVendorIDFName = "vendor";
// Device sysfs file names
static const char *kDevPerfLevelFName = "power_dpm_force_performance_level";
static const char *kDevDevProdNameFName = "product_name";
@@ -242,6 +245,7 @@ static const std::map<DevInfoTypes, const char *> kDevAttribNameMap = {
{kDevDevID, kDevDevIDFName},
{kDevDevRevID, kDevDevRevIDFName},
{kDevVendorID, kDevVendorIDFName},
{kDevPCieVendorID, kDevPCieVendorIDFName},
{kDevSubSysDevID, kDevSubSysDevIDFName},
{kDevSubSysVendorID, kDevSubSysVendorIDFName},
{kDevGPUMClk, kDevGPUMClkFName},
@@ -589,6 +593,20 @@ int Device::openSysfsFileStream(DevInfoTypes type, T *fs, const char *str) {
sysfs_path += "/device/";
sysfs_path += kDevAttribNameMap.at(type);
// For the file under PCI sysfs
if (type >= kDevPCieTypeStart && type <= kDevPCieTypeEND) {
sysfs_path = "/sys/bus/pci/devices/";
std::string bdf_str;
if (getBDFWithDomain(bdfid_, bdf_str) != RSMI_STATUS_SUCCESS) {
ss << "Fail to craft the bdf string";
LOG_ERROR(ss);
return 1;
}
sysfs_path += bdf_str;
sysfs_path += "/";
sysfs_path += kDevAttribNameMap.at(type);
}
DBG_FILE_ERROR(sysfs_path, str);
bool reg_file;
@@ -611,7 +629,7 @@ int Device::openSysfsFileStream(DevInfoTypes type, T *fs, const char *str) {
fs->open(sysfs_path);
if (!fs->is_open()) {
if (!fs) {
ss << "Could not open - SYSFS file (" << sysfs_path << ") for "
<< "DevInfoInfoType (" << RocmSMI::devInfoTypesStrings.at(type) << "), "
<< ", returning " << std::to_string(errno) << " ("
@@ -901,6 +919,7 @@ int Device::readDevInfo(DevInfoTypes type, uint64_t *val) {
case kDevSubSysDevID:
case kDevSubSysVendorID:
case kDevVendorID:
case kDevPCieVendorID:
case kDevErrCntFeatures:
ret = readDevInfoStr(type, &tempStr);
RET_IF_NONZERO(ret);
@@ -1038,6 +1057,7 @@ int Device::readDevInfo(DevInfoTypes type, std::string *val) {
case kDevSubSysDevID:
case kDevSubSysVendorID:
case kDevVendorID:
case kDevPCieVendorID:
case kDevVramVendor:
case kDevVBiosVer:
case kDevPCIEThruPut:
rocm_smi/src/rocm_smi_main.cc
+2 -1
View File
@@ -150,7 +150,8 @@ amd::smi::RocmSMI::devInfoTypesStrings = {
{amd::smi::kDevAvailableComputePartition, amdSMI +
"kDevAvailableComputePartition"},
{amd::smi::kDevComputePartition, amdSMI + "kDevComputePartition"},
{amd::smi::kDevMemoryPartition, amdSMI + "kDevMemoryPartition"}
{amd::smi::kDevMemoryPartition, amdSMI + "kDevMemoryPartition"},
{amd::smi::kDevPCieVendorID, amdSMI + "kDevPCieVendorID"},
};
namespace amd {
rocm_smi/src/rocm_smi_utils.cc
+24
View File
@@ -54,6 +54,7 @@
#include <cstdint>
#include <iomanip>
#include <iostream>
#include <iomanip>
#include <sstream>
#include <algorithm>
#include <vector>
@@ -792,6 +793,29 @@ bool isSystemBigEndian() {
return isBigEndian;
}
rsmi_status_t getBDFWithDomain(uint64_t bdf_id, std::string& bfd_str)
{
auto result = rsmi_status_t::RSMI_STATUS_SUCCESS;
auto domain_id = static_cast<uint32_t>(bdf_id >> 32);
auto bus_id = static_cast<uint8_t>((bdf_id & 0x0000FF00) >> 8);
auto dev_id = static_cast<uint8_t>((bdf_id & 0x000000F8) >> 3);
auto func_id = static_cast<uint8_t>(bdf_id & 0x00000003);
bfd_str = std::string();
if (!(bus_id > 0)) {
result = rsmi_status_t::RSMI_STATUS_NO_DATA;
return result;
}
std::stringstream bdf_sstream;
bdf_sstream << std::hex << std::setfill('0') << std::setw(4) << +domain_id << ":";
bdf_sstream << std::hex << std::setfill('0') << std::setw(2) << +bus_id << ":";
bdf_sstream << std::hex << std::setfill('0') << std::setw(2) << +dev_id << ".";
bdf_sstream << std::hex << std::setfill('0') << +func_id;
bfd_str = bdf_sstream.str();
return result;
}
rsmi_status_t getBDFString(uint64_t bdf_id, std::string& bfd_str)
{
auto result = rsmi_status_t::RSMI_STATUS_SUCCESS;
src/amd_smi/amd_smi.cc
+535 -10
View File
@@ -671,11 +671,10 @@ amdsmi_get_gpu_asic_info(amdsmi_processor_handle processor_handle, amdsmi_asic_i
info->rev_id = dev_info.pci_rev;
info->vendor_id = gpu_device->get_vendor_id();
}
// For other sysfs related information, get from rocm-smi
else {
uint64_t dv_uid = 0;
status = rsmi_wrapper(rsmi_dev_unique_id_get, processor_handle, &dv_uid);
if (status == AMDSMI_STATUS_SUCCESS) snprintf(info->asic_serial, sizeof(info->asic_serial), "%d", dv_uid);
if (status == AMDSMI_STATUS_SUCCESS) snprintf(info->asic_serial, sizeof(info->asic_serial), "%lu", dv_uid);
status = rsmi_wrapper(rsmi_dev_brand_get, processor_handle,
info->market_name, AMDSMI_NORMAL_STRING_LENGTH);
@@ -688,6 +687,9 @@ amdsmi_get_gpu_asic_info(amdsmi_processor_handle processor_handle, amdsmi_asic_i
&subvendor_id);
if (status == AMDSMI_STATUS_SUCCESS) info->subvendor_id = subvendor_id;
}
// For other sysfs related information, get from rocm-smi
status = rsmi_wrapper(rsmi_dev_pcie_vendor_name_get, processor_handle,
info->vendor_name, AMDSMI_MAX_STRING_LENGTH);
return AMDSMI_STATUS_SUCCESS;
}
@@ -1847,10 +1849,10 @@ amdsmi_status_t amdsmi_get_processor_handle_from_bdf(amdsmi_bdf_t bdf,
return status;
}
amdsmi_bdf_t found_bdf = gpu_device->get_bdf();
if (bdf.bus_number == found_bdf.bus_number &&
bdf.device_number == found_bdf.device_number &&
bdf.domain_number == found_bdf.domain_number &&
bdf.function_number == found_bdf.function_number) {
if ((bdf.fields.bus_number == found_bdf.fields.bus_number) &&
(bdf.fields.device_number == found_bdf.fields.device_number) &&
(bdf.fields.domain_number == found_bdf.fields.domain_number) &&
(bdf.fields.function_number == found_bdf.fields.function_number)) {
*processor_handle = devs[idx];
return AMDSMI_STATUS_SUCCESS;
}
@@ -1861,14 +1863,20 @@ amdsmi_status_t amdsmi_get_processor_handle_from_bdf(amdsmi_bdf_t bdf,
}
#ifdef ENABLE_ESMI_LIB
amdsmi_status_t amdsmi_get_hsmp_proto_ver(uint32_t *proto_ver)
amdsmi_status_t amdsmi_get_cpu_hsmp_proto_ver(amdsmi_cpusocket_handle socket_handle,
uint32_t *proto_ver)
{
amdsmi_status_t status;
uint32_t hsmp_proto_ver;
if (proto_ver == nullptr)
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_hsmp_proto_ver_get(&hsmp_proto_ver));
*proto_ver = hsmp_proto_ver;
@@ -1878,14 +1886,20 @@ amdsmi_status_t amdsmi_get_hsmp_proto_ver(uint32_t *proto_ver)
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_smu_fw_version(amdsmi_smu_fw_version_t *amdsmi_smu_fw)
amdsmi_status_t amdsmi_get_cpu_smu_fw_version(amdsmi_cpusocket_handle socket_handle,
amdsmi_smu_fw_version_t *amdsmi_smu_fw)
{
amdsmi_status_t status;
struct smu_fw_version smu_fw;
if (amdsmi_smu_fw == nullptr)
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_smu_fw_version_get(&smu_fw));
amdsmi_smu_fw->major = smu_fw.major;
@@ -2179,6 +2193,483 @@ amdsmi_status_t amdsmi_get_cpu_pwr_svi_telemetry_all_rails(amdsmi_cpusocket_hand
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_socket_power_cap(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t pcap)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_power_cap_set(sock_ind, pcap));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_pwr_efficiency_mode(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t mode)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_pwr_efficiency_mode_set(sock_ind, mode));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_core_boostlimit(amdsmi_processor_handle processor_handle,
uint32_t core_ind, uint32_t *pboostlimit)
{
amdsmi_status_t status;
uint32_t boostlimit;
if (processor_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuCore* core = nullptr;
amdsmi_status_t r = get_cpu_core_from_handle(processor_handle, &core);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_core_boostlimit_get(core_ind, &boostlimit));
*pboostlimit = boostlimit;
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_socket_c0_residency(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t *pc0_residency)
{
amdsmi_status_t status;
uint32_t res;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_c0_residency_get(sock_ind, &res));
*pc0_residency = res;
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_core_boostlimit(amdsmi_processor_handle processor_handle,
uint32_t core_ind, uint32_t boostlimit)
{
amdsmi_status_t status;
if (processor_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuCore* core = nullptr;
amdsmi_status_t r = get_cpu_core_from_handle(processor_handle, &core);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_core_boostlimit_set(core_ind, boostlimit));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_socket_boostlimit(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t boostlimit)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_boostlimit_set(sock_ind, boostlimit));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_ddr_bw(amdsmi_cpusocket_handle socket_handle,
amdsmi_ddr_bw_metrics_t *ddr_bw)
{
amdsmi_status_t status;
struct ddr_bw_metrics ddr;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_ddr_bw_get(&ddr));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
ddr_bw->max_bw = ddr.max_bw;
ddr_bw->utilized_bw = ddr.utilized_bw;
ddr_bw->utilized_pct = ddr.utilized_pct;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_socket_temperature(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t *ptmon)
{
amdsmi_status_t status;
uint32_t tmon;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_temperature_get(sock_ind, &tmon));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
*ptmon = tmon;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_dimm_temp_range_and_refresh_rate(
amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t dimm_addr, amdsmi_temp_range_refresh_rate_t *rate)
{
amdsmi_status_t status;
struct temp_range_refresh_rate dimm_rate;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_dimm_temp_range_and_refresh_rate_get(
sock_ind, dimm_addr, &dimm_rate));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
rate->range = dimm_rate.range;
rate->ref_rate = dimm_rate.ref_rate;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_dimm_power_consumption(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t dimm_addr, amdsmi_dimm_power_t *dimm_pow)
{
amdsmi_status_t status;
struct dimm_power d_power;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_dimm_power_consumption_get(sock_ind,
dimm_addr, &d_power));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
dimm_pow->power = d_power.power;
dimm_pow->update_rate = d_power.update_rate;
dimm_pow->dimm_addr = d_power.dimm_addr;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_dimm_thermal_sensor(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t dimm_addr, amdsmi_dimm_thermal_t *dimm_temp)
{
amdsmi_status_t status;
struct dimm_thermal d_sensor;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_dimm_thermal_sensor_get(sock_ind,
dimm_addr, &d_sensor));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
dimm_temp->temp = d_sensor.temp;
dimm_temp->update_rate = d_sensor.update_rate;
dimm_temp->dimm_addr = d_sensor.dimm_addr;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_xgmi_width(amdsmi_cpusocket_handle socket_handle,
uint8_t min, uint8_t max)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_xgmi_width_set(min, max));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_gmi3_link_width_range(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t min_link_width, uint8_t max_link_width)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_gmi3_link_width_range_set(sock_ind,
min_link_width, max_link_width));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_cpu_apb_enable(amdsmi_cpusocket_handle socket_handle, uint32_t sock_ind)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_apb_enable(sock_ind));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_cpu_apb_disable(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint8_t pstate)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_apb_disable(sock_ind, pstate));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_socket_lclk_dpm_level(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint8_t nbio_id, uint8_t min, uint8_t max)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_lclk_dpm_level_set(sock_ind,
nbio_id, min, max));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_socket_lclk_dpm_level(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t nbio_id, amdsmi_dpm_level_t *nbio)
{
amdsmi_status_t status;
struct dpm_level nb;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_socket_lclk_dpm_level_get(sock_ind,
nbio_id, &nb));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
nbio->min_dpm_level = nb.min_dpm_level;
nbio->max_dpm_level = nb.max_dpm_level;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_pcie_link_rate(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t rate_ctrl, uint8_t *prev_mode)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_pcie_link_rate_set(sock_ind,
rate_ctrl, prev_mode));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_set_cpu_df_pstate_range(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, uint8_t max_pstate, uint8_t min_pstate)
{
amdsmi_status_t status;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_df_pstate_range_set(sock_ind,
max_pstate, min_pstate));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_current_io_bandwidth(amdsmi_cpusocket_handle socket_handle,
uint8_t sock_ind, amdsmi_link_id_bw_type_t link, uint32_t *io_bw)
{
amdsmi_status_t status;
uint32_t bw;
struct link_id_bw_type io_link;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_current_io_bandwidth_get(sock_ind,
io_link, &bw));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
link.link_name = io_link.link_name;
link.bw_type= io_link.bw_type;
*io_bw = bw;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_cpu_current_xgmi_bw(amdsmi_cpusocket_handle socket_handle,
amdsmi_link_id_bw_type_t link, uint32_t *xgmi_bw)
{
amdsmi_status_t status;
uint32_t bw;
struct link_id_bw_type io_link;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_current_xgmi_bw_get(io_link, &bw));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
link.link_name = io_link.link_name;
link.bw_type= io_link.bw_type;
*xgmi_bw = bw;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_number_of_cpu_sockets(uint32_t sockets)
{
amdsmi_status_t status = amd::smi::AMDSmiSystem::getInstance().get_cpu_sockets(sockets);
@@ -2223,4 +2714,38 @@ amdsmi_status_t amdsmi_get_cpu_model(uint32_t model)
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_first_online_core_on_cpu_socket(amdsmi_cpusocket_handle socket_handle,
uint32_t sock_ind, uint32_t *pcore_ind)
{
amdsmi_status_t status;
uint32_t online_core;
if (socket_handle == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiCpuSocket* socket = nullptr;
amdsmi_status_t r = get_cpu_socket_from_handle(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
status = static_cast<amdsmi_status_t>(esmi_first_online_core_on_socket(sock_ind, &online_core));
if (status != AMDSMI_STATUS_SUCCESS)
return status;
*pcore_ind = online_core;
return AMDSMI_STATUS_SUCCESS;
}
const char* amdsmi_get_esmi_err_msg(amdsmi_status_t status, const char **status_string)
{
for (auto& iter : amd::smi::esmi_status_map) {
if (iter.first == status) {
*status_string = esmi_get_err_msg(static_cast<esmi_status_t>(iter.first));
break;
}
}
return *status_string;
}
#endif
src/amd_smi/amd_smi_drm.cc
+4 -4
View File
@@ -171,10 +171,10 @@ amdsmi_status_t AMDSmiDrm::init() {
continue;
}
bdf.function_number = device->businfo.pci->func;
bdf.device_number = device->businfo.pci->dev;
bdf.bus_number = device->businfo.pci->bus;
bdf.domain_number = device->businfo.pci->domain;
bdf.fields.function_number = device->businfo.pci->func;
bdf.fields.device_number = device->businfo.pci->dev;
bdf.fields.bus_number = device->businfo.pci->bus;
bdf.fields.domain_number = device->businfo.pci->domain;
vendor_id = device->deviceinfo.pci->vendor_id;
src/amd_smi/fdinfo.cc
+10 -8
View File
@@ -73,10 +73,11 @@ amdsmi_status_t gpuvsmi_get_pids(const amdsmi_bdf_t &bdf, std::vector<long int>
struct dirent *dir;
/* 0000:00:00.0 */
snprintf(bdf_str, 13, "%04x:%02x:%02x.%d", bdf.domain_number & 0xffff,
bdf.bus_number & 0xff,
bdf.device_number & 0x1f,
bdf.function_number & 0x7);
snprintf(bdf_str, 13, "%04x:%02x:%02x.%d",
bdf.fields.domain_number & 0xffff,
bdf.fields.bus_number & 0xff,
bdf.fields.device_number & 0x1f,
bdf.fields.function_number & 0x7);
d = opendir("/proc");
if (!d)
@@ -121,10 +122,11 @@ amdsmi_status_t gpuvsmi_get_pid_info(const amdsmi_bdf_t &bdf, long int pid,
struct dirent *dir;
/* 0000:00:00.0 */
snprintf(bdf_str, 13, "%04x:%02x:%02x.%d", bdf.domain_number & 0xffff,
bdf.bus_number & 0xff,
bdf.device_number & 0x1f,
bdf.function_number & 0x7);
snprintf(bdf_str, 13, "%04x:%02x:%02x.%d",
bdf.fields.domain_number & 0xffff,
bdf.fields.bus_number & 0xff,
bdf.fields.device_number & 0x1f,
bdf.fields.function_number & 0x7);
std::string path = "/proc/" + std::to_string(pid) + "/fdinfo/";
tests/amd_smi_test/functional/sys_info_read.cc
+2 -5
View File
@@ -163,11 +163,8 @@ void TestSysInfoRead::Run(void) {
} else {
if (err == AMDSMI_STATUS_SUCCESS) {
IF_VERB(STANDARD) {
// TODO(bliu): read unique_id
/*
std::cout << "\t**GPU Unique ID : " << std::hex << asci_info.unique_id <<
std::endl;
*/
std:: cout << "\t**GPU PCIe Vendor : "
<< asci_info.vendor_name << std::endl;
}
// Verify api support checking functionality is working
err = amdsmi_get_gpu_asic_info(processor_handles_[i], nullptr);