SWDEV-353742 - Port smilib function to amdsmi

Change-Id: I99df249755a5c665a8dd1777fa82d046e139bd77
Signed-off-by: Dalibor Stanisavljevic <Dalibor.Stanisavljevic@amd.com>
This commit is contained in:
Dalibor Stanisavljevic
2022-09-05 15:26:03 +02:00
committed by Bill(Shuzhou) Liu
parent 7fa3b36d7f
commit 3daf9c1063
18 changed files with 2577 additions and 449 deletions
+9 -2
View File
@@ -141,6 +141,8 @@ set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/amd_smi_socket.cc")
set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/amd_smi_system.cc")
set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/amd_smi_drm.cc")
set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/amd_smi_lib_loader.cc")
set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/amd_smi_utils.cc")
set(CMN_SRC_LIST ${CMN_SRC_LIST} "${AMDSMI_SRC_DIR}/fdinfo.cc")
set(CMN_INC_LIST "${ROCM_INC_DIR}/rocm_smi_device.h")
set(CMN_INC_LIST ${CMN_INC_LIST} "${ROCM_INC_DIR}/rocm_smi_main.h")
@@ -166,11 +168,16 @@ set(CMN_SRC_LIST ${CMN_SRC_LIST} "${COMMON_INC_DIR}/impl/amd_smi_lib_loader.h")
add_subdirectory("rocm_smi")
# Examples and docs
set(SMI_EXAMPLE_EXE "amd_smi_ex")
add_executable(${SMI_EXAMPLE_EXE} "example/amd_smi_example.cc")
set(SMI_EXAMPLE_EXE "amd_smi_drm_ex")
add_executable(${SMI_EXAMPLE_EXE} "example/amd_smi_drm_example.cc")
target_link_libraries(${SMI_EXAMPLE_EXE} ${ROCM_SMI_TARGET})
add_dependencies(${SMI_EXAMPLE_EXE} ${ROCM_SMI_TARGET})
set(SMI_NODRM_EXAMPLE_EXE "amd_smi_nodrm_ex")
add_executable(${SMI_NODRM_EXAMPLE_EXE} "example/amd_smi_nodrm_example.cc")
target_link_libraries(${SMI_NODRM_EXAMPLE_EXE} ${ROCM_SMI_TARGET})
add_dependencies(${SMI_NODRM_EXAMPLE_EXE} ${ROCM_SMI_TARGET})
# Generate Doxygen documentation
find_package(Doxygen)
find_package(LATEX COMPONENTS PDFLATEX)
+535
View File
@@ -0,0 +1,535 @@
/*
* =============================================================================
* The University of Illinois/NCSA
* Open Source License (NCSA)
*
* Copyright (c) 2022, Advanced Micro Devices, Inc.
* All rights reserved.
*
* Developed by:
*
* AMD Research and AMD ROC Software Development
*
* Advanced Micro Devices, Inc.
*
* www.amd.com
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal with the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimers.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimers in
* the documentation and/or other materials provided with the distribution.
* - Neither the names of <Name of Development Group, Name of Institution>,
* nor the names of its contributors may be used to endorse or promote
* products derived from this Software without specific prior written
* permission.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS WITH THE SOFTWARE.
*
*/
#include <assert.h>
#include <stdint.h>
#include <unistd.h>
#include "amd_smi/amd_smi.h"
#include <bitset>
#include <iostream>
#include <pwd.h>
#include <sys/stat.h>
#include <vector>
#define CHK_AMDSMI_RET(RET) \
{ \
if (RET != AMDSMI_STATUS_SUCCESS) { \
const char *err_str; \
std::cout << "AMDSMI call returned " << RET << " at line " \
<< __LINE__ << std::endl; \
amdsmi_status_string(RET, &err_str); \
std::cout << err_str << std::endl; \
return RET; \
} \
}
int main() {
amdsmi_status_t ret;
// Init amdsmi for sockets and devices.
// Here we are only interested in AMD_GPUS.
ret = amdsmi_init(AMDSMI_INIT_AMD_GPUS);
CHK_AMDSMI_RET(ret)
// Get all sockets
uint32_t socket_count = 0;
amdsmi_socket_handle *sockets = nullptr;
ret = amdsmi_get_socket_handles(&socket_count, &sockets);
CHK_AMDSMI_RET(ret)
std::cout << "Total Socket: " << socket_count << std::endl;
// For each socket, get identifier and devices
for (uint32_t i = 0; i < socket_count; i++) {
// Get Socket info
char socket_name[128];
ret = amdsmi_get_socket_info(sockets[i], socket_name, 128);
CHK_AMDSMI_RET(ret)
std::cout << "Socket " << socket_name << std::endl;
// Get all devices of the socket
uint32_t device_count = 0;
amdsmi_device_handle *device_handles = nullptr;
ret = amdsmi_get_device_handles(sockets[i], &device_count,
&device_handles);
CHK_AMDSMI_RET(ret)
// For each device of the socket, get name and temperature.
for (uint32_t j = 0; j < device_count; j++) {
// Get device type. Since the amdsmi is initialized with
// AMD_SMI_INIT_AMD_GPUS, the device_type must be AMD_GPU.
device_type_t device_type = {};
ret = amdsmi_get_device_type(device_handles[j], &device_type);
CHK_AMDSMI_RET(ret)
if (device_type != AMD_GPU) {
std::cout << "Expect AMD_GPU device type!\n";
return AMDSMI_STATUS_NOT_SUPPORTED;
}
amdsmi_bdf_t bdf = {};
ret = amdsmi_get_device_bdf(device_handles[j], &bdf);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_device_bdf:\n");
printf("\tDevice[%d] BDF %04x:%02x:%02x.%d\n\n", i,
bdf.domain_number, bdf.bus_number, bdf.device_number,
bdf.function_number);
amdsmi_asic_info_t asic_info = {};
ret = amdsmi_get_asic_info(device_handles[j], &asic_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_asic_info:\n");
printf("\tMarket Name: %s\n", asic_info.market_name);
printf("\tFamilyID: 0x%x\n", asic_info.family);
printf("\tDeviceID: 0x%lx\n", asic_info.device_id);
printf("\tVendorID: 0x%x\n", asic_info.vendor_id);
printf("\tRevisionID: 0x%x\n", asic_info.rev_id);
printf("\tAsic serial: 0x%s\n\n", asic_info.asic_serial);
// Get VBIOS info
amdsmi_vbios_info_t vbios_info = {};
ret = amdsmi_get_vbios_info(device_handles[j], &vbios_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_vbios_info:\n");
printf("\tVBios Name: %s\n", vbios_info.name);
printf("\tBuild Date: %s\n", vbios_info.build_date);
printf("\tPart Number: %s\n", vbios_info.part_number);
printf("\tVBios Version: %d\n", vbios_info.vbios_version);
printf("\tVBios Version String: %s\n\n",
vbios_info.vbios_version_string);
// Get engine usage info
amdsmi_engine_usage_t engine_usage = {};
ret = amdsmi_get_gpu_activity(device_handles[j], &engine_usage);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_gpu_activity:\n");
printf("\tAverage GFX Activity: %d\n",
engine_usage.average_gfx_activity);
printf("\tAverage MM Activity: %d\n",
engine_usage.average_mm_activity[0]);
printf("\tAverage UMC Activity: %d\n\n",
engine_usage.average_umc_activity);
// Get firmware info
amdsmi_fw_info_t fw_information = {};
ret = amdsmi_get_fw_info(device_handles[j], &fw_information);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_fw_info:\n");
printf("\tFirmware version: %d\n", fw_information.num_fw_info);
printf("\tSMU: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SMU]
.fw_version);
printf("\tSMC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SMC]
.fw_version);
printf("\tVCN: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_VCN]
.fw_version);
printf("\tCP_ME: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_ME]
.fw_version);
printf("\tCP_PFP: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_PFP]
.fw_version);
printf("\tCP_CE: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_CE]
.fw_version);
printf("\tRLC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_RLC]
.fw_version);
printf("\tCP_MEC1: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_MEC1]
.fw_version);
printf("\tCP_MEC2: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_MEC2]
.fw_version);
printf("\tSDMA0: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SDMA0]
.fw_version);
printf("\tMC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_MC]
.fw_version);
printf("\tRLC RESTORE LIST CNTL: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_CNTL]
.fw_version);
printf("\tRLC RESTORE LIST GPM MEM: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_GPM_MEM]
.fw_version);
printf("\tRLC RESTORE LIST SRM MEM: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_SRM_MEM]
.fw_version);
printf(
"\tPSP SOSDRV: %ld\n\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_PSP_SOSDRV]
.fw_version);
// Get GPU power limit info
amdsmi_power_limit_t power_limit = {};
ret = amdsmi_get_power_limit(device_handles[j], &power_limit);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_power_limit:\n");
printf("\tGPU Power limit: %d\n\n", power_limit.limit);
// Get GFX clock measurements
amdsmi_clock_measure_t gfx_clk_values = {};
ret = amdsmi_get_clock_measure(device_handles[j], CLOCK_TYPE_GFX,
&gfx_clk_values);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_clock_measure:\n");
printf("\tGPU GFX Max Clock: %d\n", gfx_clk_values.max_clk);
printf("\tGPU GFX Average Clock: %d\n", gfx_clk_values.avg_clk);
printf("\tGPU GFX Current Clock: %d\n", gfx_clk_values.cur_clk);
// Get MEM clock measurements
amdsmi_clock_measure_t mem_clk_values = {};
ret = amdsmi_get_clock_measure(device_handles[j], CLOCK_TYPE_MEM,
&mem_clk_values);
CHK_AMDSMI_RET(ret)
printf("\tGPU MEM Max Clock: %d\n", mem_clk_values.max_clk);
printf("\tGPU MEM Average Clock: %d\n", mem_clk_values.avg_clk);
printf("\tGPU MEM Current Clock: %d\n\n", mem_clk_values.cur_clk);
// Get VRAM temperature limit
amdsmi_temperature_limit_t mem_temp_limit = {};
ret = amdsmi_get_temperature_limit(
device_handles[j], TEMPERATURE_TYPE_VRAM, &mem_temp_limit);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_temperature_limit:\n");
printf("\tGPU VRAM temp limit: %d\n", mem_temp_limit.limit);
// Get GFX temperature limit
amdsmi_temperature_limit_t gfx_temp_limit = {};
ret = amdsmi_get_temperature_limit(
device_handles[j], TEMPERATURE_TYPE_EDGE, &gfx_temp_limit);
CHK_AMDSMI_RET(ret)
printf("\tGPU GFX temp limit: %d\n\n", gfx_temp_limit.limit);
// Get temperature measurements
// amdsmi_temperature_t edge_temp, junction_temp, vram_temp,
// plx_temp;
amdsmi_temperature_t temp_measurements[4];
amdsmi_temperature_type_t temp_types[4] = {
TEMPERATURE_TYPE_EDGE, TEMPERATURE_TYPE_JUNCTION,
TEMPERATURE_TYPE_VRAM, TEMPERATURE_TYPE_PLX};
for (const auto &temp_type : temp_types) {
ret = amdsmi_get_temperature_measure(
device_handles[j], temp_type,
&temp_measurements[(int)(temp_type)]);
CHK_AMDSMI_RET(ret)
}
printf(" Output of amdsmi_get_temperature_measure:\n");
printf("\tGPU Edge temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_EDGE].cur_temp);
printf("\tGPU Junction temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_JUNCTION].cur_temp);
printf("\tGPU VRAM temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_VRAM].cur_temp);
printf("\tGPU PLX temp measurement: %d\n\n",
temp_measurements[TEMPERATURE_TYPE_PLX].cur_temp);
// Get RAS features enabled
char block_names[14][10] = {"UMC", "SDMA", "GFX", "MMHUB",
"ATHUB", "PCIE_BIF", "HDP", "XGMI_WAFL",
"DF", "SMN", "SEM", "MP0",
"MP1", "FUSE"};
char status_names[7][10] = {"NONE", "DISABLED", "PARITY",
"SING_C", "MULT_UC", "POISON",
"ENABLED"};
amdsmi_ras_err_state_t state = {};
int index = 0;
printf(" Output of amdsmi_get_ras_features_enabled:\n");
for (auto block = AMDSMI_GPU_BLOCK_FIRST;
block <= AMDSMI_GPU_BLOCK_LAST;
block = (amdsmi_gpu_block_t)(block * 2)) {
ret = amdsmi_get_ras_features_enabled(device_handles[j], block,
&state);
CHK_AMDSMI_RET(ret)
printf("\tBlock: %s\n", block_names[index]);
printf("\tStatus: %s\n", status_names[state]);
index++;
}
printf("\n");
// Get bad pages
char bad_page_status_names[3][15] = {"RESERVED", "PENDING",
"UNRESERVABLE"};
uint32_t num_pages = 0;
ret = amdsmi_get_bad_page_info(device_handles[j], &num_pages,
nullptr);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_bad_page_info:\n");
if (!num_pages) {
printf("\tNo bad pages found.\n");
} else {
amdsmi_retired_page_record_t bad_page_info[num_pages] = {};
ret = amdsmi_get_bad_page_info(device_handles[j], &num_pages,
bad_page_info);
CHK_AMDSMI_RET(ret)
for (uint32_t page_it = 0; page_it < num_pages; page_it += 1) {
printf(" Page[%d]\n", page_it);
printf("\tAddress: %lu\n",
bad_page_info[page_it].page_address);
printf("\tSize: %lu\n", bad_page_info[page_it].page_size);
printf(
"\tStatus: %s\n",
bad_page_status_names[bad_page_info[page_it].status]);
}
}
printf("\n");
// Get ECC error counts
amdsmi_error_count_t err_cnt_info = {};
ret = amdsmi_get_ecc_error_count(device_handles[j], &err_cnt_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_ecc_error_count:\n");
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;
};
auto sum_item = [](uint16_t *a) -> float {
float b = 0;
for (int iterator = 0; iterator < AMDSMI_MAX_MM_IP_COUNT;
iterator += 1) {
b += (float)a[iterator] / 100.0;
}
return b;
};
// Get frequency ranges
amdsmi_frequency_range_t freq_ranges = {};
ret = amdsmi_get_target_frequency_range(
device_handles[j], CLOCK_TYPE_GFX, &freq_ranges);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_target_frequency_range:\n");
printf("\tSupported min freq: %lu\n",
freq_ranges.supported_freq_range.lower_bound);
printf("\tSupported max freq: %lu\n",
freq_ranges.supported_freq_range.upper_bound);
printf("\tCurrent min freq: %lu\n",
freq_ranges.current_freq_range.lower_bound);
printf("\tCurrent max freq: %lu\n\n",
freq_ranges.current_freq_range.upper_bound);
uint32_t num_process = 0;
ret = amdsmi_get_process_list(device_handles[j], nullptr,
&num_process);
CHK_AMDSMI_RET(ret)
if (!num_process) {
printf("No processes found.\n");
} else {
amdsmi_process_handle process_list[num_process];
amdsmi_proc_info_t info_list[num_process];
amdsmi_proc_info_t process = {};
uint64_t mem = 0, gtt_mem = 0, cpu_mem = 0, vram_mem = 0;
float gfx = 0, comp = 0, dma = 0, enc = 0, dec = 0;
char bdf_str[20];
sprintf(bdf_str, "%04x:%02x:%02x.%d", bdf.domain_number,
bdf.bus_number, bdf.device_number, bdf.function_number);
int num = 0;
ret = amdsmi_get_process_list(device_handles[j], process_list,
&num_process);
CHK_AMDSMI_RET(ret)
for (uint32_t it = 0; it < num_process; it += 1) {
if (getpid() == process_list[it]) {
continue;
}
ret = amdsmi_get_process_info(device_handles[j],
process_list[it], &process);
if (ret != AMDSMI_STATUS_SUCCESS) {
printf("amdsmi_get_process_info() failed for "
"process_list[%d], returned %d\n",
it, ret);
continue;
}
info_list[num++] = process;
}
qsort(info_list, num, sizeof(info_list[0]), compare);
printf("+=======+==================+============+=============="
"+=============+=============+=============+============"
"==+=========================================+\n");
printf(
"| pid | name | user | gpu bdf | "
"fb usage | gtt memory | cpu memory | vram memory | "
"ring usage (%%) |\n");
printf("| | | | "
"| | | | "
" | gfx comp dma enc dec |\n");
printf("+=======+==================+============+=============="
"+=============+=============+=============+============"
"==+=========================================+\n");
for (int it = 0; it < num; it++) {
char command[30];
struct passwd *pwd = NULL;
struct stat st;
sprintf(command, "/proc/%d", info_list[it].pid);
if (stat(command, &st))
continue;
pwd = getpwuid(st.st_uid);
if (!pwd)
printf("| %5d | %16s | %10d | %s | %7ld KiB | %7ld KiB "
"| %7ld KiB | %7ld KiB | %6.2f %6.2f %6.2f "
"%6.2f %6.2f |\n",
info_list[it].pid, info_list[it].name, st.st_uid,
bdf_str, info_list[it].mem / 1024,
info_list[it].memory_usage.gtt_mem / 1024,
info_list[it].memory_usage.cpu_mem / 1024,
info_list[it].memory_usage.vram_mem / 1024,
sum_item(info_list[it].engine_usage.gfx),
sum_item(info_list[it].engine_usage.compute),
sum_item(info_list[it].engine_usage.sdma),
sum_item(info_list[it].engine_usage.enc),
sum_item(info_list[it].engine_usage.dec));
else
printf("| %5d | %16s | %10s | %s | %7ld KiB | %7ld KiB "
"| %7ld KiB | %7ld KiB | %6.2f %6.2f %6.2f "
"%6.2f %6.2f |\n",
info_list[it].pid, info_list[it].name,
pwd->pw_name, bdf_str, info_list[it].mem / 1024,
info_list[it].memory_usage.gtt_mem / 1024,
info_list[it].memory_usage.cpu_mem / 1024,
info_list[it].memory_usage.vram_mem / 1024,
sum_item(info_list[it].engine_usage.gfx),
sum_item(info_list[it].engine_usage.compute),
sum_item(info_list[it].engine_usage.sdma),
sum_item(info_list[it].engine_usage.enc),
sum_item(info_list[it].engine_usage.dec));
mem += info_list[it].mem / 1024;
gtt_mem += info_list[it].memory_usage.gtt_mem / 1024;
cpu_mem += info_list[it].memory_usage.cpu_mem / 1024;
vram_mem += info_list[it].memory_usage.vram_mem / 1024;
gfx += sum_item(info_list[it].engine_usage.gfx);
comp += sum_item(info_list[it].engine_usage.compute);
dma += sum_item(info_list[it].engine_usage.sdma);
enc += sum_item(info_list[it].engine_usage.enc);
dec += sum_item(info_list[it].engine_usage.dec);
printf(
"+-------+------------------+------------+-------------"
"-+-------------+-------------+-------------+----------"
"----+-----------------------------------------+\n");
}
printf("| TOTAL:| %s | %7ld "
"KiB | %7ld KiB | %7ld KiB | %7ld KiB | %6.2f %6.2f "
"%6.2f %6.2f %6.2f |\n",
bdf_str, mem, gtt_mem, cpu_mem, vram_mem, gfx, comp, dma,
enc, dec);
printf("+=======+==================+============+=============="
"+=============+=============+=============+============"
"=+==========================================+\n");
}
// Get device name
amdsmi_board_info board_info = {};
ret = amdsmi_get_board_info(device_handles[j], &board_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_board_info:\n");
std::cout << "\tdevice [" << j
<< "]\n\t\tProduct name: " << board_info.product_name
<< "\n"
<< "\t\tProduct number: " << board_info.product_serial
<< "\n"
<< "\t\tSerial number: " << board_info.serial_number
<< "\n\n";
// Get temperature
int64_t val_i64 = 0;
ret = amdsmi_dev_temp_metric_get(device_handles[j], 0,
AMDSMI_TEMP_CURRENT, &val_i64);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_dev_temp_metric_get:\n");
std::cout << "\t\tTemperature: " << val_i64 / 1000 << "C"
<< "\n\n";
// Get frame buffer
amdsmi_vram_info_t vram_usage = {};
ret = amdsmi_get_vram_usage(device_handles[j], &vram_usage);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_vram_usage:\n");
std::cout << "\t\tFrame buffer usage (MB): " << vram_usage.vram_used
<< "/" << vram_usage.vram_total << "\n\n";
// Get Cap info
amdsmi_gpu_caps_t caps_info = {};
ret = amdsmi_get_caps_info(device_handles[j], &caps_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_caps_info:\n");
std::cout << "\t\tGFX IP Major: " << caps_info.gfx.gfxip_major
<< "\n"
<< "\t\tGFX IP Minor: " << caps_info.gfx.gfxip_minor
<< "\n"
<< "\t\tCU IP Count: " << caps_info.gfx.gfxip_cu_count
<< "\n"
<< "\t\tDMA IP Count: " << caps_info.dma_ip_count << "\n"
<< "\t\tGFX IP Count: " << caps_info.gfx_ip_count << "\n"
<< "\t\tMM IP Count: " << int(caps_info.mm.mm_ip_count)
<< "\n\n";
amdsmi_power_cap_info cap_info = {};
ret = amdsmi_get_power_cap_info(device_handles[j], 0, &cap_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_power_cap_info:\n");
std::cout << "\t\t Power Cap: " << cap_info.power_cap
<< "W\n";
std::cout << "\t\t Dpm Cap: " << cap_info.dpm_cap
<< "\n\n";
}
}
// Clean up resources allocated at amdsmi_init. It will invalidate sockets
// and devices pointers
ret = amdsmi_shut_down();
CHK_AMDSMI_RET(ret)
return 0;
}
-146
View File
@@ -1,146 +0,0 @@
/*
* =============================================================================
* The University of Illinois/NCSA
* Open Source License (NCSA)
*
* Copyright (c) 2022, Advanced Micro Devices, Inc.
* All rights reserved.
*
* Developed by:
*
* AMD Research and AMD ROC Software Development
*
* Advanced Micro Devices, Inc.
*
* www.amd.com
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal with the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimers.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimers in
* the documentation and/or other materials provided with the distribution.
* - Neither the names of <Name of Development Group, Name of Institution>,
* nor the names of its contributors may be used to endorse or promote
* products derived from this Software without specific prior written
* permission.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS WITH THE SOFTWARE.
*
*/
#include <assert.h>
#include <stdint.h>
#include <unistd.h>
#include <vector>
#include <iostream>
#include <bitset>
#include "amd_smi/amd_smi.h"
#define CHK_AMDSMI_RET(RET) { \
if (RET != AMDSMI_STATUS_SUCCESS) { \
const char *err_str; \
std::cout << "AMDSMI call returned " << RET \
<< " at line " << __LINE__ << std::endl; \
amdsmi_status_string(RET, &err_str); \
std::cout << err_str << std::endl; \
return RET; \
} \
}
int main() {
amdsmi_status_t ret;
// Init amdsmi for sockets and devices.
// Here we are only interested in AMD_GPUS.
ret = amdsmi_init(AMDSMI_INIT_AMD_GPUS);
CHK_AMDSMI_RET(ret)
// Get all sockets
uint32_t socket_count = 0;
amdsmi_socket_handle* sockets = nullptr;
ret = amdsmi_get_socket_handles(&socket_count, &sockets);
CHK_AMDSMI_RET(ret)
std::cout << "Total Socket: " << socket_count << std::endl;
// For each socket, get identifier and devices
for (uint32_t i = 0; i < socket_count; i++) {
// Get Socket info
char socket_name[128];
ret = amdsmi_get_socket_info(sockets[i], socket_name, 128);
CHK_AMDSMI_RET(ret)
std::cout << "Socket " << socket_name << std::endl;
// Get all devices of the socket
uint32_t device_count = 0;
amdsmi_device_handle* device_handles = nullptr;
ret = amdsmi_get_device_handles(sockets[i],
&device_count, &device_handles);
CHK_AMDSMI_RET(ret)
// For each device of the socket, get name and temperature.
for (uint32_t j = 0; j < device_count; j++) {
// Get device type. Since the amdsmi is initialized with
// AMDSMI_INIT_AMD_GPUS, the device_type must be AMD_GPU.
device_type_t device_type;
ret = amdsmi_get_device_type(device_handles[j], &device_type);
CHK_AMDSMI_RET(ret)
if (device_type != AMD_GPU) {
std::cout << "Expect AMD_GPU device type!\n";
return 1;
}
// Get device name
amdsmi_board_info board_info;
ret = amdsmi_get_board_info(device_handles[j], &board_info);
CHK_AMDSMI_RET(ret)
std::cout << "\tdevice "
<< j << "\n\t\tName:" << board_info.product_name << std::endl;
// Get temperature
int64_t val_i64 = 0;
ret = amdsmi_dev_temp_metric_get(device_handles[j], 0,
AMDSMI_TEMP_CURRENT, &val_i64);
CHK_AMDSMI_RET(ret)
std::cout << "\t\tTemperature: " << val_i64/1000 << "C" << std::endl;
// Get frame buffer
amdsmi_vram_info_t vram_usage;
ret = amdsmi_get_vram_usage(device_handles[j], &vram_usage);
CHK_AMDSMI_RET(ret)
std::cout << "\t\tFrame buffer usage (MB): " << vram_usage.vram_used << "/"
<< vram_usage.vram_total << std::endl;
// Get Cap info
amdsmi_gpu_caps_t caps_info = {};
ret = amdsmi_get_caps_info(device_handles[j], &caps_info);
CHK_AMDSMI_RET(ret)
std::cout << "\t\tGFX IP Major: " << caps_info.gfx.gfxip_major << "\n";
std::cout << "\t\tGFX IP Minor: " << caps_info.gfx.gfxip_minor << "\n";
std::cout << "\t\tCU IP Count: " << caps_info.gfx.gfxip_cu_count << "\n";
std::cout << "\t\tDMA IP Count: " << caps_info.dma_ip_count << "\n";
std::cout << "\t\tGFX IP Count: " << caps_info.gfx_ip_count << "\n";
std::cout << "\t\tMM IP Count: " << int(caps_info.mm.mm_ip_count) << "\n";
}
}
// Clean up resources allocated at amdsmi_init. It will invalidate sockets
// and devices pointers
ret = amdsmi_shut_down();
CHK_AMDSMI_RET(ret)
return 0;
}
+335
View File
@@ -0,0 +1,335 @@
/*
* =============================================================================
* The University of Illinois/NCSA
* Open Source License (NCSA)
*
* Copyright (c) 2022, Advanced Micro Devices, Inc.
* All rights reserved.
*
* Developed by:
*
* AMD Research and AMD ROC Software Development
*
* Advanced Micro Devices, Inc.
*
* www.amd.com
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to
* deal with the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimers.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimers in
* the documentation and/or other materials provided with the distribution.
* - Neither the names of <Name of Development Group, Name of Institution>,
* nor the names of its contributors may be used to endorse or promote
* products derived from this Software without specific prior written
* permission.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE CONTRIBUTORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
* DEALINGS WITH THE SOFTWARE.
*
*/
#include <assert.h>
#include <stdint.h>
#include <unistd.h>
#include "amd_smi/amd_smi.h"
#include <bitset>
#include <iostream>
#include <pwd.h>
#include <sys/stat.h>
#include <vector>
#define CHK_AMDSMI_RET(RET) \
{ \
if (RET != AMDSMI_STATUS_SUCCESS) { \
const char *err_str; \
std::cout << "AMDSMI call returned " << RET << " at line " \
<< __LINE__ << std::endl; \
amdsmi_status_string(RET, &err_str); \
std::cout << err_str << std::endl; \
return RET; \
} \
}
int main() {
amdsmi_status_t ret;
// Init amdsmi for sockets and devices.
// Here we are only interested in AMD_GPUS.
ret = amdsmi_init(AMDSMI_INIT_AMD_GPUS);
CHK_AMDSMI_RET(ret)
// Get all sockets
uint32_t socket_count = 0;
amdsmi_socket_handle *sockets = nullptr;
ret = amdsmi_get_socket_handles(&socket_count, &sockets);
CHK_AMDSMI_RET(ret)
std::cout << "Total Socket: " << socket_count << std::endl;
// For each socket, get identifier and devices
for (uint32_t i = 0; i < socket_count; i++) {
// Get Socket info
char socket_name[128];
ret = amdsmi_get_socket_info(sockets[i], socket_name, 128);
CHK_AMDSMI_RET(ret)
std::cout << "Socket " << socket_name << std::endl;
// Get all devices of the socket
uint32_t device_count = 0;
amdsmi_device_handle *device_handles = nullptr;
ret = amdsmi_get_device_handles(sockets[i], &device_count,
&device_handles);
CHK_AMDSMI_RET(ret)
// For each device of the socket, get name and temperature.
for (uint32_t j = 0; j < device_count; j++) {
// Get device type. Since the amdsmi is initialized with
// AMD_SMI_INIT_AMD_GPUS, the device_type must be AMD_GPU.
device_type_t device_type = {};
ret = amdsmi_get_device_type(device_handles[j], &device_type);
CHK_AMDSMI_RET(ret)
if (device_type != AMD_GPU) {
std::cout << "Expect AMD_GPU device type!\n";
return AMDSMI_STATUS_NOT_SUPPORTED;
}
amdsmi_bdf_t bdf = {};
ret = amdsmi_get_device_bdf(device_handles[j], &bdf);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_device_bdf:\n");
printf("\tDevice[%d] BDF %04x:%02x:%02x.%d\n\n", i,
bdf.domain_number, bdf.bus_number, bdf.device_number,
bdf.function_number);
amdsmi_asic_info_t asic_info = {};
ret = amdsmi_get_asic_info(device_handles[j], &asic_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_asic_info:\n");
printf("\tMarket Name: %s\n", asic_info.market_name);
printf("\tFamilyID: 0x%x\n", asic_info.family);
printf("\tDeviceID: 0x%x\n", asic_info.device_id);
printf("\tVendorID: 0x%x\n", asic_info.vendor_id);
printf("\tRevisionID: 0x%x\n", asic_info.rev_id);
printf("\tAsic serial: 0x%s\n\n", asic_info.asic_serial);
// Get VBIOS info
amdsmi_vbios_info_t vbios_info = {};
ret = amdsmi_get_vbios_info(device_handles[j], &vbios_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_vbios_info:\n");
printf("\tVBios Name: %s\n", vbios_info.name);
printf("\tBuild Date: %s\n", vbios_info.build_date);
printf("\tPart Number: %s\n", vbios_info.part_number);
printf("\tVBios Version: %d\n", vbios_info.vbios_version);
printf("\tVBios Version String: %s\n\n",
vbios_info.vbios_version_string);
// Get engine usage info
amdsmi_engine_usage_t engine_usage = {};
ret = amdsmi_get_gpu_activity(device_handles[j], &engine_usage);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_gpu_activity:\n");
printf("\tAverage GFX Activity: %d\n",
engine_usage.average_gfx_activity);
printf("\tAverage MM Activity: %d\n",
engine_usage.average_mm_activity[0]);
printf("\tAverage UMC Activity: %d\n\n",
engine_usage.average_umc_activity);
// Get firmware info
amdsmi_fw_info_t fw_information = {};
ret = amdsmi_get_fw_info(device_handles[j], &fw_information);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_fw_info:\n");
printf("\tFirmware version: %d\n", fw_information.num_fw_info);
printf("\tSMU: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SMU]
.fw_version);
printf("\tSMC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SMC]
.fw_version);
printf("\tVCN: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_VCN]
.fw_version);
printf("\tCP_ME: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_ME]
.fw_version);
printf("\tCP_PFP: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_PFP]
.fw_version);
printf("\tCP_CE: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_CE]
.fw_version);
printf("\tRLC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_RLC]
.fw_version);
printf("\tCP_MEC1: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_MEC1]
.fw_version);
printf("\tCP_MEC2: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_CP_MEC2]
.fw_version);
printf("\tSDMA0: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_SDMA0]
.fw_version);
printf("\tMC: %ld\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_MC]
.fw_version);
printf("\tRLC RESTORE LIST CNTL: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_CNTL]
.fw_version);
printf("\tRLC RESTORE LIST GPM MEM: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_GPM_MEM]
.fw_version);
printf("\tRLC RESTORE LIST SRM MEM: %ld\n",
fw_information
.fw_info_list
[amdsmi_fw_block_t::FW_ID_RLC_RESTORE_LIST_SRM_MEM]
.fw_version);
printf(
"\tPSP SOSDRV: %ld\n\n",
fw_information.fw_info_list[amdsmi_fw_block_t::FW_ID_PSP_SOSDRV]
.fw_version);
// Get temperature measurements
amdsmi_temperature_t temp_measurements[4];
amdsmi_temperature_type_t temp_types[4] = {
TEMPERATURE_TYPE_EDGE, TEMPERATURE_TYPE_JUNCTION,
TEMPERATURE_TYPE_VRAM, TEMPERATURE_TYPE_PLX};
for (const auto &temp_type : temp_types) {
ret = amdsmi_get_temperature_measure(
device_handles[j], temp_type,
&temp_measurements[(int)(temp_type)]);
CHK_AMDSMI_RET(ret)
}
printf(" Output of amdsmi_get_temperature_measure:\n");
printf("\tGPU Edge temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_EDGE].cur_temp);
printf("\tGPU Junction temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_JUNCTION].cur_temp);
printf("\tGPU VRAM temp measurement: %d\n",
temp_measurements[TEMPERATURE_TYPE_VRAM].cur_temp);
printf("\tGPU PLX temp measurement: %d\n\n",
temp_measurements[TEMPERATURE_TYPE_PLX].cur_temp);
// Get bad pages
char bad_page_status_names[3][15] = {"RESERVED", "PENDING",
"UNRESERVABLE"};
uint32_t num_pages = 0;
ret = amdsmi_get_bad_page_info(device_handles[j], &num_pages,
nullptr);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_bad_page_info:\n");
if (!num_pages) {
printf("\tNo bad pages found.\n");
} else {
amdsmi_retired_page_record_t bad_page_info[num_pages] = {};
ret = amdsmi_get_bad_page_info(device_handles[j], &num_pages,
bad_page_info);
CHK_AMDSMI_RET(ret)
for (uint32_t page_it = 0; page_it < num_pages; page_it += 1) {
printf(" Page[%d]\n", page_it);
printf("\tAddress: %lu\n",
bad_page_info[page_it].page_address);
printf("\tSize: %lu\n", bad_page_info[page_it].page_size);
printf(
"\tStatus: %s\n",
bad_page_status_names[bad_page_info[page_it].status]);
}
}
printf("\n");
// Get ECC error counts
amdsmi_error_count_t err_cnt_info = {};
ret = amdsmi_get_ecc_error_count(device_handles[j], &err_cnt_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_ecc_error_count:\n");
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 board_info = {};
ret = amdsmi_get_board_info(device_handles[j], &board_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_board_info:\n");
std::cout << "\tdevice [" << j
<< "]\n\t\tProduct name: " << board_info.product_name
<< "\n"
<< "\t\tProduct number: " << board_info.product_serial
<< "\n"
<< "\t\tSerial number: " << board_info.serial_number
<< "\n\n";
// Get temperature
int64_t val_i64 = 0;
ret = amdsmi_dev_temp_metric_get(device_handles[j], 0,
AMDSMI_TEMP_CURRENT, &val_i64);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_dev_temp_metric_get:\n");
std::cout << "\t\tTemperature: " << val_i64 / 1000 << "C"
<< "\n\n";
// Get frame buffer
amdsmi_vram_info_t vram_usage = {};
ret = amdsmi_get_vram_usage(device_handles[j], &vram_usage);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_vram_usage:\n");
std::cout << "\t\tFrame buffer usage (MB): " << vram_usage.vram_used
<< "/" << vram_usage.vram_total << "\n\n";
// Get Cap info
amdsmi_gpu_caps_t caps_info = {};
ret = amdsmi_get_caps_info(device_handles[j], &caps_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_caps_info:\n");
std::cout << "\t\tGFX IP Major: " << caps_info.gfx.gfxip_major
<< "\n"
<< "\t\tGFX IP Minor: " << caps_info.gfx.gfxip_minor
<< "\n"
<< "\t\tCU IP Count: " << caps_info.gfx.gfxip_cu_count
<< "\n"
<< "\t\tDMA IP Count: " << caps_info.dma_ip_count << "\n"
<< "\t\tGFX IP Count: " << caps_info.gfx_ip_count << "\n"
<< "\t\tMM IP Count: " << int(caps_info.mm.mm_ip_count)
<< "\n\n";
amdsmi_power_cap_info cap_info = {};
ret = amdsmi_get_power_cap_info(device_handles[j], 0, &cap_info);
CHK_AMDSMI_RET(ret)
printf(" Output of amdsmi_get_power_cap_info:\n");
std::cout << "\t\t Power Cap: " << cap_info.power_cap / 1000000
<< "W\n\n";
}
}
// Clean up resources allocated at amdsmi_init. It will invalidate sockets
// and devices pointers
ret = amdsmi_shut_down();
CHK_AMDSMI_RET(ret)
return 0;
}
+222 -174
View File
@@ -104,7 +104,7 @@ typedef enum device_type {
/**
* @brief Error codes retured by amd_smi_lib functions
*/
typedef enum amdsmi_status {
typedef enum amdsmi_status_t {
AMDSMI_STATUS_SUCCESS = 0, /**< Call succeeded */
AMDSMI_STATUS_INVAL, /**< Invalid parameters */
AMDSMI_STATUS_NOT_SUPPORTED, /**< Command not supported */
@@ -334,15 +334,15 @@ typedef struct amdsmi_board_info {
} amdsmi_board_info_t;
typedef struct amdsmi_temperature {
uint16_t cur_temp;
uint32_t cur_temp;
} amdsmi_temperature_t;
typedef struct amdsmi_temperature_limit {
uint16_t limit;
uint32_t limit;
} amdsmi_temperature_limit_t;
typedef struct amdsmi_power_limit {
uint16_t limit;
uint32_t limit;
} amdsmi_power_limit_t;
typedef struct amdsmi_power_measure {
@@ -369,17 +369,22 @@ typedef struct amdsmi_engine_usage {
typedef uint32_t amdsmi_process_handle;
typedef struct amdsmi_process_info {
char name[AMDSMI_NORMAL_STRING_LENGTH];
amdsmi_process_handle pid;
uint64_t mem; /** in bytes */
char name[AMDSMI_NORMAL_STRING_LENGTH];
amdsmi_process_handle pid;
uint64_t mem; /** in bytes */
struct {
uint16_t gfx[AMDSMI_MAX_MM_IP_COUNT];
uint16_t compute[AMDSMI_MAX_MM_IP_COUNT];
uint16_t sdma[AMDSMI_MAX_MM_IP_COUNT];
uint16_t enc[AMDSMI_MAX_MM_IP_COUNT];
uint16_t dec[AMDSMI_MAX_MM_IP_COUNT];
} engine_usage; /** percentage 0-100% times 100 */
struct {
uint16_t gfx[AMDSMI_MAX_MM_IP_COUNT];
uint16_t compute[AMDSMI_MAX_MM_IP_COUNT];
uint16_t sdma[AMDSMI_MAX_MM_IP_COUNT];
uint16_t enc[AMDSMI_MAX_MM_IP_COUNT];
uint16_t dec[AMDSMI_MAX_MM_IP_COUNT];
} usage; /** percentage 0-100% times 100 */
char container_name[AMDSMI_NORMAL_STRING_LENGTH];
uint64_t gtt_mem;
uint64_t cpu_mem;
uint64_t vram_mem;
} memory_usage; /** in bytes */
char container_name[AMDSMI_NORMAL_STRING_LENGTH];
} amdsmi_proc_info_t;
//! Guaranteed maximum possible number of supported frequencies
@@ -1071,8 +1076,8 @@ typedef struct {
* @brief This structure holds error counts.
*/
typedef struct {
uint64_t correctable_err; //!< Accumulated correctable errors
uint64_t uncorrectable_err; //!< Accumulated uncorrectable errors
uint64_t correctable_count; //!< Accumulated correctable errors
uint64_t uncorrectable_count; //!< Accumulated uncorrectable errors
} amdsmi_error_count_t;
/**
@@ -1795,7 +1800,50 @@ amdsmi_dev_memory_total_get(amdsmi_device_handle device_handle, amdsmi_memory_ty
amdsmi_status_t
amdsmi_dev_memory_usage_get(amdsmi_device_handle device_handle, amdsmi_memory_type_t mem_type,
uint64_t *used);
/**
* @brief The first call to this API returns the number of bad pages which
* should be used to allocate the buffer that should contain the bad page
* records.
* @details This call will query the device @p device_handle for the
* number of bad pages (written to @p num_pages address). The results are
* written to address held by the @p info pointer.
* @param[in] device_handle a device handle
* @param[out] num_pages Number of bad page records.
* @param[out] info Pointer to amdsmi_retired_page_record_t to which the
* results will be written to.
* @retval ::AMDSMI_STATUS_SUCCESS call was successful
* @retval ::AMDSMI_STATUS_INVAL the parameters are not valid or nullptr
* @retval ::AMDSMI_STATUS_NOT_SUPPORTED API not supported
*/
amdsmi_status_t
amdsmi_get_bad_page_info(amdsmi_device_handle device_handle, uint32_t *num_pages, amdsmi_retired_page_record_t *info);
/**
* @brief Returns if RAS features are enabled or disabled for given block
*
* @details Given a device handle @p device_handle, this function queries the
* state of RAS features for a specific block @p block. Result will be written
* to address held by pointer @p state.
*
* @param[in] device_handle Device handle which to query
*
* @param[in] block Block which to query
*
* @param[inout] state A pointer to amdsmi_ras_err_state_t to which the state
* of block will be written.
* If this parameter is nullptr, this function will return
* ::AMDSMI_STATUS_INVALID_ARGS if the function is supported with the provided
* arguments and ::AMDSMI_STATUS_NOT_SUPPORTED if it is not supported with the
* provided arguments.
*
* @retval ::AMDSMI_STATUS_SUCCESS call was successful
* @retval ::AMDSMI_STATUS_NOT_SUPPORTED installed software or hardware does not
* support this function with the given arguments.
* @retval ::AMDSMI_STATUS_INVALID_ARGS the provided arguments are not valid
*
*/
amdsmi_status_t
amdsmi_get_ras_features_enabled(amdsmi_device_handle device_handle, amdsmi_gpu_block block,
amdsmi_ras_err_state_t *state);
/**
* @brief Get percentage of time any device memory is being used
*
@@ -3676,11 +3724,11 @@ amdsmi_status_t amdsmi_event_notification_stop(amdsmi_device_handle device_handl
* \param [out] bdf - Reference to BDF. Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are invalid
* * -::SMI_ERR_NOT_FOUND - Device cannot be found
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are invalid
* * -::AMDSMI_STATUS_NOT_FOUND - Device cannot be found
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status_t
amdsmi_get_device_bdf(amdsmi_device_handle dev, amdsmi_bdf_t *bdf);
@@ -3698,13 +3746,13 @@ amdsmi_get_device_bdf(amdsmi_device_handle dev, amdsmi_bdf_t *bdf);
* allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are invalid
* * -::SMI_ERR_NOT_FOUND - Device cannot be found
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are invalid
* * -::AMDSMI_STATUS_NOT_FOUND - Device cannot be found
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_device_uuid(amdsmi_device_handle dev, unsigned int *uuid_length, char *uuid);
/** @} */
@@ -3726,13 +3774,13 @@ amdsmi_get_device_uuid(amdsmi_device_handle dev, unsigned int *uuid_length, char
* allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NOT_FOUND - Device cannot be found
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NOT_FOUND - Device cannot be found
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_driver_version(amdsmi_device_handle dev, int *length, char *version);
/** @} */
@@ -3750,15 +3798,15 @@ amdsmi_get_driver_version(amdsmi_device_handle dev, int *length, char *version);
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_asic_info(amdsmi_device_handle dev, amdsmi_asic_info_t *info);
/**
@@ -3770,15 +3818,15 @@ amdsmi_get_asic_info(amdsmi_device_handle dev, amdsmi_asic_info_t *info);
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialize
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialize
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_board_info(amdsmi_device_handle dev, amdsmi_board_info_t *info);
/**
@@ -3792,15 +3840,15 @@ amdsmi_get_board_info(amdsmi_device_handle dev, amdsmi_board_info_t *info);
* allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_power_cap_info(amdsmi_device_handle dev, uint32_t sensor_ind,
amdsmi_power_cap_info_t *info);
@@ -3814,15 +3862,15 @@ amdsmi_get_power_cap_info(amdsmi_device_handle dev, uint32_t sensor_ind,
*
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_xgmi_info(amdsmi_device_handle dev, amdsmi_xgmi_info_t *info);
/**
@@ -3835,15 +3883,15 @@ amdsmi_get_xgmi_info(amdsmi_device_handle dev, amdsmi_xgmi_info_t *info);
* allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_caps_info(amdsmi_device_handle dev, amdsmi_gpu_caps_t *info);
/** @} */
@@ -3860,15 +3908,15 @@ amdsmi_get_caps_info(amdsmi_device_handle dev, amdsmi_gpu_caps_t *info);
* \param [out] info - Reference to the fw info. Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_fw_info(amdsmi_device_handle dev, amdsmi_fw_info_t *info);
/**
@@ -3880,15 +3928,15 @@ amdsmi_get_fw_info(amdsmi_device_handle dev, amdsmi_fw_info_t *info);
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_vbios_info(amdsmi_device_handle dev, amdsmi_vbios_info_t *info);
/** @} */
@@ -3906,15 +3954,15 @@ amdsmi_get_vbios_info(amdsmi_device_handle dev, amdsmi_vbios_info_t *info);
* \param [out] info - Reference to the gpu engine usage structure. Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_gpu_activity(amdsmi_device_handle dev, amdsmi_engine_usage_t *info);
/**
@@ -3926,15 +3974,15 @@ amdsmi_get_gpu_activity(amdsmi_device_handle dev, amdsmi_engine_usage_t *info);
* \param [out] info - Reference to the gpu power structure. Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_power_measure(amdsmi_device_handle dev, amdsmi_power_measure_t *info);
/**
@@ -3950,15 +3998,15 @@ amdsmi_get_power_measure(amdsmi_device_handle dev, amdsmi_power_measure_t *info)
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_clock_measure(amdsmi_device_handle dev, amdsmi_clk_type_t clk_type, amdsmi_clock_measure_t *info);
/**
@@ -3973,15 +4021,15 @@ amdsmi_get_clock_measure(amdsmi_device_handle dev, amdsmi_clk_type_t clk_type, a
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_temperature_measure(amdsmi_device_handle dev, amdsmi_temperature_type_t temp_type, amdsmi_temperature_t *info);
/**
@@ -3996,16 +4044,16 @@ amdsmi_get_temperature_measure(amdsmi_device_handle dev, amdsmi_temperature_type
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_get_temperature_limit(amdsmi_device_handle dev, amdsmi_temperature_t temp_type, amdsmi_temperature_limit_t *limit);
amdsmi_status_t
amdsmi_get_temperature_limit(amdsmi_device_handle dev, amdsmi_temperature_type_t temp_type, amdsmi_temperature_limit_t *limit);
/**
* \brief Returns power limit of the GPU.
@@ -4017,15 +4065,15 @@ amdsmi_get_temperature_limit(amdsmi_device_handle dev, amdsmi_temperature_t temp
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_power_limit(amdsmi_device_handle dev, amdsmi_power_limit_t *limit);
/**
@@ -4039,15 +4087,15 @@ amdsmi_get_power_limit(amdsmi_device_handle dev, amdsmi_power_limit_t *limit);
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_vram_usage(amdsmi_device_handle dev, amdsmi_vram_info_t *info);
/** @} */
@@ -4072,15 +4120,15 @@ amdsmi_get_vram_usage(amdsmi_device_handle dev, amdsmi_vram_info_t *info);
*
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialize
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_INVAL - Parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialize
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_INVAL - Parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_target_frequency_range(amdsmi_device_handle dev, amdsmi_clk_type_t clk_type, amdsmi_frequency_range_t *range);
/** @} */
@@ -4109,13 +4157,13 @@ amdsmi_get_target_frequency_range(amdsmi_device_handle dev, amdsmi_clk_type_t cl
* in list or the number of running processes if equal to 0.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - The parameters are not valid or NULL
* * -::SMI_ERR_NOMEM - Provided buffer is not large enough
* * -::SMI_ERR_NOT_SUPPORTED - API not supported
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - The parameters are not valid or NULL
* * -::AMDSMI_STATUS_NOMEM - Provided buffer is not large enough
* * -::AMDSMI_STATUS_NOT_SUPPORTED - API not supported
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_process_list(amdsmi_device_handle dev, amdsmi_process_handle *list, uint32_t *max_processes);
/**
@@ -4129,12 +4177,12 @@ amdsmi_get_process_list(amdsmi_device_handle dev, amdsmi_process_handle *list, u
* information. Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - The parameters are not valid or NULL
* * -::SMI_ERR_NOT_SUPPORTED - API not supported
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - The parameters are not valid or NULL
* * -::AMDSMI_STATUS_NOT_SUPPORTED - API not supported
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_process_info(amdsmi_device_handle dev, amdsmi_process_handle process, amdsmi_proc_info_t *info);
/** @} */
@@ -4154,15 +4202,15 @@ amdsmi_get_process_info(amdsmi_device_handle dev, amdsmi_process_handle process,
* Must be allocated by user.
*
* \return
* * ::SMI_SUCCESS - Successful
* * -::SMI_ERR_RETRY - Device is busy. Please retry
* * -::SMI_ERR_NO_PERM - Library was not initialized
* * -::SMI_ERR_INVAL - The parameters are not valid or NULL
* * -::SMI_ERR_IO - Device is in an unrecoverable state
* * -::SMI_ERR_NOT_INIT - Device is uninitialized
* * -::SMI_ERR_API_FAILED - Other errors
* * ::AMDSMI_STATUS_SUCCESS - Successful
* * -::AMDSMI_STATUS_RETRY - Device is busy. Please retry
* * -::AMDSMI_STATUS_NO_PERM - Library was not initialized
* * -::AMDSMI_STATUS_INVAL - The parameters are not valid or NULL
* * -::AMDSMI_STATUS_IO - Device is in an unrecoverable state
* * -::AMDSMI_STATUS_NOT_INIT - Device is uninitialized
* * -::AMDSMI_STATUS_API_FAILED - Other errors
*/
amdsmi_status
amdsmi_status_t
amdsmi_get_ecc_error_count(amdsmi_device_handle dev, amdsmi_error_count_t *ec);
#ifdef __cplusplus
+10 -1
View File
@@ -58,7 +58,13 @@ class AMDSmiDrm {
public:
amdsmi_status_t init();
amdsmi_status_t cleanup();
int get_drm_fd_by_index(uint32_t gpu_index) const;
amdsmi_status_t get_drm_fd_by_index(uint32_t gpu_index, uint32_t *fd_info) const;
amdsmi_status_t get_bdf_by_index(uint32_t gpu_index, amdsmi_bdf_t *bdf_info) const;
amdsmi_status_t get_drm_path_by_index(uint32_t gpu_index, std::string *drm_path) const;
std::vector<amdsmi_bdf_t> get_bdfs();
std::vector<std::string>& get_drm_paths();
bool check_if_drm_is_supported();
amdsmi_status_t amdgpu_query_info(int fd, unsigned info_id,
unsigned size, void *value);
amdsmi_status_t amdgpu_query_fw(int fd, unsigned info_id, unsigned fw_type,
@@ -70,6 +76,9 @@ class AMDSmiDrm {
private:
using DrmCmdWriteFunc = int (*)(int, unsigned long, void *, unsigned long);
std::vector<int> drm_fds_; // drm file descriptor by gpu_index
std::vector<std::string> drm_paths_; // drm path (renderD128 for example)
std::vector<amdsmi_bdf_t> drm_bdfs_; // bdf
AMDSmiLibraryLoader lib_loader_; // lazy load libdrm
DrmCmdWriteFunc drm_cmd_write_; // drmCommandWrite
std::mutex drm_mutex_;
+22 -2
View File
@@ -47,16 +47,32 @@
#include "amd_smi/amd_smi.h"
#include "amd_smi/impl/amd_smi_device.h"
#include "amd_smi/impl/amd_smi_drm.h"
#include "shared_mutex.h" // NOLINT
namespace amd {
namespace smi {
class AMDSmiGPUDevice: public AMDSmiDevice {
public:
explicit AMDSmiGPUDevice(uint32_t gpu_id, AMDSmiDrm& drm):
AMDSmiDevice(AMD_GPU), gpu_id_(gpu_id), drm_(drm) {}
AMDSmiGPUDevice(uint32_t gpu_id, uint32_t fd, std::string path, amdsmi_bdf_t bdf, AMDSmiDrm& drm):
AMDSmiDevice(AMD_GPU), gpu_id_(gpu_id), fd_(fd), path_(path), bdf_(bdf), drm_(drm) {}
AMDSmiGPUDevice(uint32_t gpu_id, AMDSmiDrm& drm):
AMDSmiDevice(AMD_GPU), gpu_id_(gpu_id), drm_(drm) {
if (check_if_drm_is_supported()) this->get_drm_data();
}
~AMDSmiGPUDevice() {
if (check_if_drm_is_supported()) shared_mutex_close(mutex_);
}
amdsmi_status_t get_drm_data();
pthread_mutex_t* get_mutex();
uint32_t get_gpu_id() const;
uint32_t get_gpu_fd() const;
std::string& get_gpu_path();
amdsmi_bdf_t get_bdf();
bool check_if_drm_is_supported() { return drm_.check_if_drm_is_supported(); }
amdsmi_status_t amdgpu_query_info(unsigned info_id,
unsigned size, void *value) const;
amdsmi_status_t amdgpu_query_hw_ip(unsigned info_id, unsigned hw_ip_type,
@@ -66,7 +82,11 @@ class AMDSmiGPUDevice: public AMDSmiDevice {
amdsmi_status_t amdgpu_query_vbios(void *info) const;
private:
uint32_t gpu_id_;
uint32_t fd_;
std::string path_;
amdsmi_bdf_t bdf_;
AMDSmiDrm& drm_;
shared_mutex_t mutex_;
};
+1
View File
@@ -60,6 +60,7 @@ class AMDSmiSocket {
const std::string& get_socket_id() const { return socket_identifier_;}
void add_device(AMDSmiDevice* device) { devices_.push_back(device); }
std::vector<AMDSmiDevice*>& get_devices() { return devices_;}
amdsmi_status_t get_device_count(uint32_t* device_count) const;
private:
std::string socket_identifier_;
std::vector<AMDSmiDevice*> devices_;
+44
View File
@@ -0,0 +1,44 @@
/* * Copyright (C) 2022 Advanced Micro Devices. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy of
* this software and associated documentation files (the "Software"), to deal in
* the Software without restriction, including without limitation the rights to
* use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
* the Software, and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
* FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
* COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
* IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#ifndef AMD_SMI_INCLUDE_AMD_SMI_UTILS_H_
#define AMD_SMI_INCLUDE_AMD_SMI_UTILS_H_
#include "amd_smi/amd_smi.h"
#include "amd_smi/impl/amd_smi_gpu_device.h"
#include "rocm_smi/rocm_smi_utils.h"
#define SMIGPUDEVICE_MUTEX(MUTEX) \
amd::smi::pthread_wrap _pw(*(MUTEX)); \
amd::smi::ScopedPthread _lock(_pw, true); \
if (_lock.mutex_not_acquired()) { \
return AMDSMI_STATUS_BUSY; \
}
amdsmi_status_t smi_amdgpu_find_hwmon_dir(amd::smi::AMDSmiGPUDevice* device, std::string* full_path);
amdsmi_status_t smi_amdgpu_get_board_info(amd::smi::AMDSmiGPUDevice* device, amdsmi_board_info_t *info);
amdsmi_status_t smi_amdgpu_get_power_cap(amd::smi::AMDSmiGPUDevice* device, int *cap);
amdsmi_status_t smi_amdgpu_get_ranges(amd::smi::AMDSmiGPUDevice* device, amdsmi_clk_type_t domain, int *max_freq, int *min_freq, int *num_dpm);
amdsmi_status_t smi_amdgpu_get_enabled_blocks(amd::smi::AMDSmiGPUDevice* device, uint64_t *enabled_blocks);
amdsmi_status_t smi_amdgpu_get_bad_page_info(amd::smi::AMDSmiGPUDevice* device, uint32_t *num_pages, amdsmi_retired_page_record_t *info);
amdsmi_status_t smi_amdgpu_get_ecc_error_count(amd::smi::AMDSmiGPUDevice* device, amdsmi_error_count_t *err_cnt);
#endif //
+38
View File
@@ -0,0 +1,38 @@
/*
* Copyright 2022 Advanced Micro Devices, Inc.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
*/
#ifndef __FDINFO__
#define __FDINFO__
#ifdef __cplusplus
extern "C" {
#endif
amdsmi_status_t gpuvsmi_get_pids(const amdsmi_bdf_t &bdf, std::vector<long int> &pids, uint64_t *size);
amdsmi_status_t gpuvsmi_get_pid_info(const amdsmi_bdf_t &bdf, long int pid, amdsmi_proc_info_t &info);
#ifdef __cplusplus
} // extern "C"
#endif
#endif
+514 -76
View File
@@ -50,12 +50,14 @@
#include <sstream>
#include <iomanip>
#include <iostream>
#include <fstream>
#include <vector>
#include <set>
#include <map>
#include <memory>
#include <xf86drm.h>
#include "amd_smi/amd_smi.h"
#include "amd_smi/impl/fdinfo.h"
#include "amd_smi/impl/amd_smi_common.h"
#include "amd_smi/impl/amd_smi_system.h"
#include "amd_smi/impl/amd_smi_socket.h"
@@ -63,6 +65,7 @@
#include "rocm_smi/rocm_smi.h"
#include "rocm_smi/rocm_smi_common.h"
#include "amd_smi/impl/amdgpu_drm.h"
#include "amd_smi/impl/amd_smi_utils.h"
// TODO(bliu): One to one map to all status code
static amdsmi_status_t rsmi_to_amdsmi_status(rsmi_status_t status) {
@@ -178,7 +181,9 @@ amdsmi_status_t amdsmi_get_device_handles(amdsmi_socket_handle socket_handle,
.handle_to_socket(socket_handle, &socket);
if (r != AMDSMI_STATUS_SUCCESS) return r;
*device_count = static_cast<uint32_t>(socket->get_devices().size());
r = socket->get_device_count(device_count);
if (r != AMDSMI_STATUS_SUCCESS) return r;
*device_handles = reinterpret_cast<amdsmi_device_handle*>(
socket->get_devices().data());
return AMDSMI_STATUS_SUCCESS;
@@ -198,54 +203,44 @@ amdsmi_status_t amdsmi_get_device_type(amdsmi_device_handle device_handle ,
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t amdsmi_get_board_info(amdsmi_device_handle device_handle,
amdsmi_board_info_t *info) {
if (info == NULL) {
amdsmi_status_t
amdsmi_get_device_bdf(amdsmi_device_handle device_handle, amdsmi_bdf_t *bdf) {
if (bdf == NULL) {
return AMDSMI_STATUS_INVAL;
}
memset(info, 0, sizeof(amdsmi_board_info_t));
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
// ignore errors so that if the function is not supported,
// it will continue to add other info.
auto r = rsmi_wrapper(rsmi_dev_name_get, device_handle,
info->product_name, AMDSMI_PRODUCT_NAME_LENGTH);
r = rsmi_wrapper(rsmi_dev_serial_number_get, device_handle,
info->product_serial, AMDSMI_NORMAL_STRING_LENGTH);
if (gpu_device->check_if_drm_is_supported()) {
*bdf = gpu_device->get_bdf();
}
else {
//TODO
}
return AMDSMI_STATUS_SUCCESS;
}
// TODO(bliu) : add other asic info
amdsmi_status amdsmi_get_asic_info(amdsmi_device_handle device_handle,
amdsmi_asic_info_t *info) {
if (info == nullptr)
amdsmi_status_t amdsmi_get_board_info(amdsmi_device_handle device_handle, amdsmi_board_info_t *board_info) {
if (board_info == NULL) {
return AMDSMI_STATUS_INVAL;
memset(info, 0, sizeof(amdsmi_asic_info_t));
}
auto r = rsmi_wrapper(rsmi_dev_serial_number_get, device_handle,
info->asic_serial, AMDSMI_NORMAL_STRING_LENGTH);
amdsmi_status_t status;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
r = rsmi_wrapper(rsmi_dev_brand_get, device_handle,
info->market_name, AMDSMI_NORMAL_STRING_LENGTH);
if (gpu_device->check_if_drm_is_supported()) {
status = smi_amdgpu_get_board_info(gpu_device, board_info);
}
else {
status = rsmi_wrapper(rsmi_dev_name_get, device_handle, board_info->product_name, AMDSMI_PRODUCT_NAME_LENGTH);
status = rsmi_wrapper(rsmi_dev_serial_number_get, device_handle, board_info->product_serial, AMDSMI_NORMAL_STRING_LENGTH);
}
uint16_t vendor_id = 0;
r = rsmi_wrapper(rsmi_dev_vendor_id_get, device_handle,
&vendor_id);
if ( r == AMDSMI_STATUS_SUCCESS)
info->vendor_id = static_cast<uint32_t>(vendor_id);
r = rsmi_wrapper(rsmi_dev_unique_id_get, device_handle,
&(info->device_id));
vendor_id = 0;
r = rsmi_wrapper(rsmi_dev_subsystem_vendor_id_get, device_handle,
&vendor_id);
if ( r == AMDSMI_STATUS_SUCCESS)
info->subvendor_id = static_cast<uint32_t>(vendor_id);
return AMDSMI_STATUS_SUCCESS;
return status;
}
amdsmi_status_t amdsmi_dev_temp_metric_get(amdsmi_device_handle device_handle,
@@ -333,7 +328,7 @@ amdsmi_status_t amdsmi_get_caps_info(amdsmi_device_handle device_handle,
sizeof(struct drm_amdgpu_info_device), &device);
if (r != AMDSMI_STATUS_SUCCESS) return r;
info->gfx.gfxip_cu_count = device.cu_active_number;
info->gfx.gfxip_cu_count = (uint16_t)device.cu_active_number;
r = gpu_device->amdgpu_query_hw_ip(AMDGPU_INFO_HW_IP_INFO,
AMDGPU_HW_IP_GFX, sizeof(ip), &ip);
@@ -392,16 +387,6 @@ amdsmi_status_t amdsmi_get_caps_info(amdsmi_device_handle device_handle,
return AMDSMI_STATUS_SUCCESS;
}
// TODO(bliu): add more vbios info
amdsmi_status amdsmi_get_vbios_info(amdsmi_device_handle device_handle,
amdsmi_vbios_info_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
return rsmi_wrapper(rsmi_dev_vbios_version_get, device_handle,
info->vbios_version_string, AMDSMI_NORMAL_STRING_LENGTH);
}
amdsmi_status_t amdsmi_dev_fan_rpms_get(amdsmi_device_handle device_handle,
uint32_t sensor_ind, int64_t *speed) {
return rsmi_wrapper(rsmi_dev_fan_rpms_get, device_handle, sensor_ind,
@@ -437,7 +422,7 @@ amdsmi_status_t amdsmi_dev_id_get(amdsmi_device_handle device_handle,
}
// TODO(bliu) : add fw info from libdrm
amdsmi_status amdsmi_get_fw_info(amdsmi_device_handle dev,
amdsmi_status_t amdsmi_get_fw_info(amdsmi_device_handle dev,
amdsmi_fw_info_t *info) {
const std::map<amdsmi_fw_block, rsmi_fw_block_t> fw_in_rsmi = {
{ FW_ID_ASD, RSMI_FW_BLOCK_ASD},
@@ -477,6 +462,66 @@ amdsmi_status amdsmi_get_fw_info(amdsmi_device_handle dev,
info->num_fw_info++;
}
}
return AMDSMI_STATUS_SUCCESS;
}
// TODO(bliu) : add other asic info
amdsmi_status_t
amdsmi_get_asic_info(amdsmi_device_handle device_handle, amdsmi_asic_info_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
struct drm_amdgpu_info_device dev_info = {};
struct drm_amdgpu_info_vbios vbios = {};
char* name;
char *tmp;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
amdsmi_status_t status;
if (gpu_device->check_if_drm_is_supported()){
status = gpu_device->amdgpu_query_info(AMDGPU_INFO_DEV_INFO, sizeof(struct drm_amdgpu_info_device), &dev_info);
if (status != AMDSMI_STATUS_SUCCESS) return status;
status = gpu_device->amdgpu_query_vbios(&vbios);
if (status != AMDSMI_STATUS_SUCCESS) return status;
SMIGPUDEVICE_MUTEX(gpu_device->get_mutex())
std::string path = "/sys/class/drm/" + gpu_device->get_gpu_path() + "/device/unique_id";
FILE *fp = fopen(path.c_str(), "r");
if (fp) {
fscanf(fp, "%s", &info->asic_serial);
fclose(fp);
}
name = strtok_r((char *) vbios.name, " ", &tmp);
if (name)
strncpy(info->market_name, name, AMDSMI_MAX_STRING_LENGTH);
info->device_id = dev_info.device_id;
info->family = dev_info.family;
info->rev_id = dev_info.pci_rev;
}
else {
uint16_t vendor_id = 0;
amdsmi_status_t status = rsmi_wrapper(rsmi_dev_serial_number_get, device_handle,
info->asic_serial, AMDSMI_NORMAL_STRING_LENGTH);
status = rsmi_wrapper(rsmi_dev_brand_get, device_handle,
info->market_name, AMDSMI_NORMAL_STRING_LENGTH);
status = rsmi_wrapper(rsmi_dev_vendor_id_get, device_handle,
&vendor_id);
if (status == AMDSMI_STATUS_SUCCESS) info->vendor_id = vendor_id;
vendor_id = 0;
status = rsmi_wrapper(rsmi_dev_subsystem_vendor_id_get, device_handle,
&vendor_id);
if (status == AMDSMI_STATUS_SUCCESS) info->subvendor_id = vendor_id;
return status;
}
return AMDSMI_STATUS_SUCCESS;
}
@@ -664,7 +709,7 @@ amdsmi_is_P2P_accessible(amdsmi_device_handle device_handle_src,
}
// TODO(bliu) : other xgmi related information
amdsmi_status
amdsmi_status_t
amdsmi_get_xgmi_info(amdsmi_device_handle device_handle, amdsmi_xgmi_info_t *info) {
if (info == nullptr)
return AMDSMI_STATUS_INVAL;
@@ -882,29 +927,48 @@ amdsmi_status_t amdsmi_dev_gpu_metrics_info_get(
reinterpret_cast<rsmi_gpu_metrics_t*>(pgpu_metrics));
}
// TODO(bliu): read from libdrm
amdsmi_status
amdsmi_status_t
amdsmi_get_power_cap_info(amdsmi_device_handle device_handle,
uint32_t sensor_ind,
amdsmi_power_cap_info_t *info) {
if (info == nullptr)
return AMDSMI_STATUS_INVAL;
amd::smi::AMDSmiGPUDevice* gpudevice = nullptr;
amdsmi_status_t r = get_gpu_device_from_handle(device_handle, &gpudevice);
if (r != AMDSMI_STATUS_SUCCESS)
return r;
amd::smi::AMDSmiGPUDevice* gpudevice =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
amdsmi_status_t status;
if (gpudevice->check_if_drm_is_supported()){
int power_cap = 0;
int dpm = 0;
// Ignore errors to get as much as possible info.
memset(info, 0, sizeof(amdsmi_power_cap_info_t));
auto rsmi_status = rsmi_dev_power_cap_default_get(gpudevice->get_gpu_id(),
&(info->default_power_cap));
rsmi_status = rsmi_dev_power_cap_range_get(gpudevice->get_gpu_id(),
sensor_ind, &(info->max_power_cap), &(info->min_power_cap));
rsmi_status = rsmi_dev_power_cap_get(gpudevice->get_gpu_id(),
sensor_ind, &(info->power_cap));
status = smi_amdgpu_get_power_cap(gpudevice, &power_cap);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
info->power_cap = power_cap;
status = smi_amdgpu_get_ranges(gpudevice, CLOCK_TYPE_GFX,
NULL, NULL, &dpm);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
info->dpm_cap = dpm;
return AMDSMI_STATUS_SUCCESS;
}
else {
// Ignore errors to get as much as possible info.
memset(info, 0, sizeof(amdsmi_power_cap_info_t));
auto rsmi_status = rsmi_dev_power_cap_default_get(gpudevice->get_gpu_id(),
&(info->default_power_cap));
rsmi_status = rsmi_dev_power_cap_range_get(gpudevice->get_gpu_id(),
sensor_ind, &(info->max_power_cap), &(info->min_power_cap));
rsmi_status = rsmi_dev_power_cap_get(gpudevice->get_gpu_id(),
sensor_ind, &(info->power_cap));
// TODO(bliu) : dpm_cap
}
// TODO(bliu) : dpm_cap
return AMDSMI_STATUS_SUCCESS;
}
@@ -1192,19 +1256,393 @@ amdsmi_status_t amdsmi_version_str_get(amdsmi_sw_component_t component,
return rsmi_to_amdsmi_status(status);
}
amdsmi_status amdsmi_get_gpu_activity(amdsmi_device_handle dev,
amdsmi_engine_usage_t *info) {
if (info == nullptr)
amdsmi_status_t
amdsmi_get_vbios_info(amdsmi_device_handle dev, amdsmi_vbios_info_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
struct drm_amdgpu_info_vbios vbios = {};
// Get gpu activity from the gpu_metrics table
amdsmi_gpu_metrics_t gpu_metrics_info;
auto r = amdsmi_dev_gpu_metrics_info_get(dev, &gpu_metrics_info);
if ( r == AMDSMI_STATUS_SUCCESS ) {
info->average_gfx_activity = gpu_metrics_info.average_gfx_activity;
info->average_umc_activity = gpu_metrics_info.average_umc_activity;
info->average_mm_activity[0] = gpu_metrics_info.average_mm_activity;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
if (gpu_device->check_if_drm_is_supported()){
status = gpu_device->amdgpu_query_vbios(&vbios);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
strncpy(info->name, (char *) vbios.name, AMDSMI_MAX_STRING_LENGTH);
strncpy(info->build_date, (char *) vbios.date, AMDSMI_MAX_DATE_LENGTH);
strncpy(info->part_number, (char *) vbios.vbios_pn, AMDSMI_MAX_STRING_LENGTH);
strncpy(info->vbios_version_string, (char *) vbios.vbios_ver_str, AMDSMI_NORMAL_STRING_LENGTH);
info->vbios_version = vbios.version;
}
else {
// rocm
}
return r;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_gpu_activity(amdsmi_device_handle dev, amdsmi_engine_usage_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
amdsmi_gpu_metrics_t metrics = {};
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
status = amdsmi_dev_gpu_metrics_info_get(dev, &metrics);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
info->average_gfx_activity = metrics.average_gfx_activity;
info->average_mm_activity[0] = metrics.average_mm_activity;
info->average_umc_activity = metrics.average_umc_activity;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_power_limit(amdsmi_device_handle dev, amdsmi_power_limit_t *limit) {
if (limit == nullptr) {
return AMDSMI_STATUS_INVAL;
}
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
int power_limit;
status = smi_amdgpu_get_power_cap(gpu_device, &power_limit);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
limit->limit = (uint16_t)(power_limit);
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_clock_measure(amdsmi_device_handle dev, amdsmi_clk_type_t clk_type, amdsmi_clock_measure_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
if (clk_type >= CLOCK_TYPE__MAX) {
printf("Domain value greater or equals CLOCK_TYPE__MAX value. Return code: %d", AMDSMI_STATUS_INVAL);
return AMDSMI_STATUS_INVAL;
}
amdsmi_gpu_metrics_t metrics = {};
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
status = amdsmi_dev_gpu_metrics_info_get(dev, &metrics);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
int max_freq;
status = smi_amdgpu_get_ranges(gpu_device, clk_type,
&max_freq, NULL, NULL);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
info->max_clk = max_freq;
switch (clk_type) {
case CLOCK_TYPE_GFX:
info->avg_clk = metrics.average_gfxclk_frequency;
info->cur_clk = metrics.current_gfxclk;
break;
case CLOCK_TYPE_MEM:
info->avg_clk = metrics.average_uclk_frequency;
info->cur_clk = metrics.current_uclk;
break;
case CLOCK_TYPE_VCLK0:
info->avg_clk = metrics.average_vclk0_frequency;
info->cur_clk = metrics.current_vclk0;
break;
case CLOCK_TYPE_VCLK1:
info->avg_clk = metrics.average_vclk1_frequency;
info->cur_clk = metrics.current_vclk1;
break;
default:
return AMDSMI_STATUS_INVAL;
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_temperature_limit(amdsmi_device_handle dev, amdsmi_temperature_type_t temp_type, amdsmi_temperature_limit_t *limit) {
if (limit == nullptr || temp_type >= TEMPERATURE_TYPE__MAX) {
return AMDSMI_STATUS_INVAL;
}
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
std::string name;
std::string path;
switch (temp_type) {
case TEMPERATURE_TYPE_EDGE:
name = "edge";
break;
case TEMPERATURE_TYPE_JUNCTION:
name = "junction";
break;
case TEMPERATURE_TYPE_VRAM:
name = "mem";
break;
default:
return AMDSMI_STATUS_INVAL;
}
status = smi_amdgpu_find_hwmon_dir(gpu_device, &path);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
SMIGPUDEVICE_MUTEX(gpu_device->get_mutex())
for (int count = 1; ; count++) {
std::string local_path = path + "/temp" +
std::to_string(count);
std::string temp = local_path + "_label";
char f_name[10];
std::ifstream file(temp.c_str(), std::ifstream::in);
if (!file.is_open()) {
printf("Failed to open file: %s \n", temp.c_str());
return AMDSMI_STATUS_API_FAILED;
}
file.getline(f_name, 10);
if (!strstr(name.c_str(), f_name)) {
int readTemp = 0;
temp = local_path + "_crit";
std::ifstream file2(temp.c_str(), std::ifstream::in);
if (!file2.is_open()) {
printf("Failed to open file: %s \n", temp.c_str());
return AMDSMI_STATUS_API_FAILED;
}
file2.getline(f_name, 10);
if (!sscanf(f_name, "%d", &readTemp)) {
return AMDSMI_STATUS_API_FAILED;
}
limit->limit = (uint16_t)(readTemp / 1000);
break;
}
file.close();
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_temperature_measure(amdsmi_device_handle dev, amdsmi_temperature_type_t temp_type, amdsmi_temperature_t *info) {
if (info == nullptr || temp_type > TEMPERATURE_TYPE__MAX) {
return AMDSMI_STATUS_INVAL;
}
amdsmi_gpu_metrics_t metrics;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
status = amdsmi_dev_gpu_metrics_info_get(dev, &metrics);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
switch (temp_type) {
case TEMPERATURE_TYPE_EDGE:
info->cur_temp = metrics.temperature_edge;
break;
case TEMPERATURE_TYPE_JUNCTION:
info->cur_temp = metrics.temperature_hotspot;
break;
case TEMPERATURE_TYPE_VRAM:
info->cur_temp = metrics.temperature_mem;
break;
case TEMPERATURE_TYPE_PLX:
info->cur_temp = metrics.temperature_vrsoc;
break;
default:
return AMDSMI_STATUS_INVAL;
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_ras_features_enabled(amdsmi_device_handle device_handle, amdsmi_gpu_block block, amdsmi_ras_err_state_t *state) {
if (state == nullptr || block > AMDSMI_GPU_BLOCK_LAST) {
return AMDSMI_STATUS_INVAL;
}
uint64_t features_mask = 0;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
amdsmi_status_t status;
status = smi_amdgpu_get_enabled_blocks(gpu_device, &features_mask);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
*state = (features_mask & block) ? AMDSMI_RAS_ERR_STATE_ENABLED : AMDSMI_RAS_ERR_STATE_DISABLED;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_bad_page_info(amdsmi_device_handle device_handle, uint32_t *num_pages, amdsmi_retired_page_record_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(device_handle);
amdsmi_status_t status;
if (gpu_device->check_if_drm_is_supported()){
status = smi_amdgpu_get_bad_page_info(gpu_device, num_pages, info);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
}
else {
// rocm
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_ecc_error_count(amdsmi_device_handle dev, amdsmi_error_count_t *ec) {
if (ec == nullptr) {
return AMDSMI_STATUS_INVAL;
}
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
if (gpu_device->check_if_drm_is_supported()){
status = smi_amdgpu_get_ecc_error_count(gpu_device, ec);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
}
else {
// rocm
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_process_list(amdsmi_device_handle dev, amdsmi_process_handle *list, uint32_t *max_processes) {
if (max_processes == nullptr) {
return AMDSMI_STATUS_INVAL;
}
std::vector<long int> pids;
uint32_t i = 0;
uint64_t size = 0;
amdsmi_status_t status;
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
if (gpu_device->check_if_drm_is_supported()){
amdsmi_bdf_t bdf = gpu_device->get_bdf();
status = gpuvsmi_get_pids(bdf, pids, &size);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
if (*max_processes == 0 || (pids.size() == 0)) {
*max_processes = (uint32_t)pids.size();
return AMDSMI_STATUS_SUCCESS;
}
if (!list) {
return AMDSMI_STATUS_INVAL;
}
if (*max_processes < pids.size()) {
return AMDSMI_STATUS_OUT_OF_RESOURCES;
}
for (auto &pid : pids) {
if (i >= *max_processes) {
break;
}
list[i++] = (uint32_t)pid;
}
*max_processes = (uint32_t)pids.size();
}
else {
// rocm
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_process_info(amdsmi_device_handle dev, amdsmi_process_handle process, amdsmi_proc_info_t *info) {
if (info == nullptr) {
return AMDSMI_STATUS_INVAL;
}
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
if (gpu_device->check_if_drm_is_supported()) {
status = gpuvsmi_get_pid_info(gpu_device->get_bdf(), process, *info);
if (status != AMDSMI_STATUS_SUCCESS) return status;
}
else {
// rocm
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t
amdsmi_get_target_frequency_range(amdsmi_device_handle dev, amdsmi_clk_type_t clk_type, amdsmi_frequency_range_t *range) {
if (range == nullptr || clk_type > CLOCK_TYPE__MAX) {
return AMDSMI_STATUS_INVAL;
}
amdsmi_gpu_metrics_t metrics = {};
amd::smi::AMDSmiGPUDevice* gpu_device =
static_cast<amd::smi::AMDSmiGPUDevice*>(dev);
amdsmi_status_t status;
int min = 0, max = 0;
status = amdsmi_dev_gpu_metrics_info_get(dev, &metrics);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
status = smi_amdgpu_get_ranges(gpu_device, clk_type, &max, &min, nullptr);
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
range->supported_freq_range.lower_bound = (long)min;
range->current_freq_range.lower_bound = (long)min;
range->supported_freq_range.upper_bound = (long)max;
max = 0;
switch (clk_type) {
case CLOCK_TYPE_GFX:
max = metrics.current_gfxclk;
break;
case CLOCK_TYPE_MEM:
max = metrics.current_uclk;
break;
case CLOCK_TYPE_VCLK0:
max = metrics.current_vclk0;
break;
case CLOCK_TYPE_VCLK1:
max = metrics.current_vclk1;
break;
default:
return AMDSMI_STATUS_INVAL;
}
range->current_freq_range.upper_bound = (long)max;
return AMDSMI_STATUS_SUCCESS;
}
+71 -5
View File
@@ -55,27 +55,40 @@ namespace smi {
amdsmi_status_t AMDSmiDrm::init() {
// A few RAII handler
using dir_ptr = std::unique_ptr<DIR, decltype(&closedir)>;
using drm_version_ptr = std::unique_ptr<drmVersion,
decltype(&drmFreeVersion)>;
// using drm_device_ptr = std::unique_ptr(drmDevicePtr,
// decltype(&drmFreeDevice));
struct dirent *dir = nullptr;
int fd = -1;
amdsmi_status_t status = lib_loader_.load("libdrm.so");
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
// load symbol from libdrm
drm_cmd_write_ = nullptr;
status = lib_loader_.load_symbol(&drm_cmd_write_, "drmCommandWrite");
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
using drmGetVersionType = drmVersionPtr (*)(int); // drmGetVersion
using drmFreeVersionType = void (*)(drmVersionPtr); // drmFreeVersion
using drmGetVersionType = drmVersionPtr (*)(int); // drmGetVersion
using drmFreeVersionType = void (*)(drmVersionPtr); // drmFreeVersion
using drmGetDeviceType = int(*)(int, drmDevicePtr*); // drmGetDevice
using drmFreeDeviceType = void(*)(drmDevicePtr*); // drmFreeDevice
drmGetVersionType drm_get_version = nullptr;
drmFreeVersionType drm_free_version = nullptr;
drmGetDeviceType drm_get_device = nullptr;
drmFreeDeviceType drm_free_device = nullptr;
status = lib_loader_.load_symbol(&drm_get_version, "drmGetVersion");
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
@@ -85,10 +98,20 @@ amdsmi_status_t AMDSmiDrm::init() {
return status;
}
status = lib_loader_.load_symbol(&drm_get_device, "drmGetDevice");
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
status = lib_loader_.load_symbol(&drm_free_device, "drmFreeDevice");
if (status != AMDSMI_STATUS_SUCCESS) {
return status;
}
auto d = dir_ptr(opendir("/dev/dri/"), &closedir);
if (d == nullptr) return AMDSMI_STATUS_NOT_INIT;
drmDevicePtr device;
while ((dir = readdir(d.get())) != NULL) {
char* name_cstr = new char[sizeof(dir->d_name) + 10];
auto name = std::unique_ptr<char[]>(name_cstr);
@@ -105,7 +128,22 @@ amdsmi_status_t AMDSmiDrm::init() {
continue;
}
if (drm_get_device(fd, &device) != 0) {
drm_free_device(&device);
return AMDSMI_STATUS_DRM_ERROR;
}
drm_fds_.push_back(fd);
drm_paths_.push_back(dir->d_name);
amdsmi_bdf_t bdf;
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;
drm_bdfs_.push_back(bdf);
drm_free_device(&device);
}
return AMDSMI_STATUS_SUCCESS;
@@ -115,7 +153,10 @@ amdsmi_status_t AMDSmiDrm::cleanup() {
for (unsigned int i=0; i < drm_fds_.size(); i++) {
close(drm_fds_[i]);
}
drm_fds_.clear();
drm_paths_.clear();
drm_bdfs_.clear();
lib_loader_.unload();
return AMDSMI_STATUS_SUCCESS;
}
@@ -190,9 +231,34 @@ amdsmi_status_t AMDSmiDrm::amdgpu_query_vbios(int fd, void *info) {
}
int AMDSmiDrm::get_drm_fd_by_index(uint32_t gpu_index) const {
if (gpu_index + 1 > drm_fds_.size()) return -1;
return drm_fds_[gpu_index];
amdsmi_status_t AMDSmiDrm::get_drm_fd_by_index(uint32_t gpu_index, uint32_t *fd_info) const {
if (gpu_index + 1 > drm_fds_.size()) return AMDSMI_STATUS_NOT_SUPPORTED;
*fd_info = drm_fds_[gpu_index];
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t AMDSmiDrm::get_bdf_by_index(uint32_t gpu_index, amdsmi_bdf_t *bdf_info) const {
if (gpu_index + 1 > drm_bdfs_.size()) return AMDSMI_STATUS_NOT_SUPPORTED;
*bdf_info = drm_bdfs_[gpu_index];
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t AMDSmiDrm::get_drm_path_by_index(uint32_t gpu_index, std::string *drm_path) const {
if (gpu_index + 1 > drm_paths_.size()) return AMDSMI_STATUS_NOT_SUPPORTED;
*drm_path = drm_paths_[gpu_index];
return AMDSMI_STATUS_SUCCESS;
}
std::vector<std::string>& AMDSmiDrm::get_drm_paths() {
return drm_paths_;
}
bool AMDSmiDrm::check_if_drm_is_supported() {
return drm_cmd_write_ != NULL ? true : false;
}
std::vector<amdsmi_bdf_t> AMDSmiDrm::get_bdfs() {
return drm_bdfs_;
}
} // namespace smi
+53 -8
View File
@@ -52,33 +52,78 @@ uint32_t AMDSmiGPUDevice::get_gpu_id() const {
return gpu_id_;
}
uint32_t AMDSmiGPUDevice::get_gpu_fd() const {
return fd_;
}
std::string& AMDSmiGPUDevice::get_gpu_path() {
return path_;
}
amdsmi_bdf_t AMDSmiGPUDevice::get_bdf() {
return bdf_;
}
amdsmi_status_t AMDSmiGPUDevice::get_drm_data() {
amdsmi_status_t ret;
uint32_t fd = 0;
std::string path;
amdsmi_bdf_t bdf;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
ret = drm_.get_drm_path_by_index(gpu_id_, &path);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
ret = drm_.get_bdf_by_index(gpu_id_, &bdf);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
mutex_ = shared_mutex_init(path.c_str(), 0777);
if (mutex_.ptr == nullptr) {
printf("Failed to create shared mem. mutex.");
return AMDSMI_STATUS_INIT_ERROR;
}
bdf_ = bdf, path_ = path, fd_ = fd;
return AMDSMI_STATUS_SUCCESS;
}
pthread_mutex_t* AMDSmiGPUDevice::get_mutex() {
return mutex_.ptr;
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_info(unsigned info_id,
unsigned size, void *value) const {
int fd = drm_.get_drm_fd_by_index(gpu_id_);
if (fd == -1) return AMDSMI_STATUS_NOT_SUPPORTED;
amdsmi_status_t ret;
uint32_t fd = 0;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
return drm_.amdgpu_query_info(fd, info_id, size, value);
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_hw_ip(unsigned info_id,
unsigned hw_ip_type, unsigned size, void *value) const {
int fd = drm_.get_drm_fd_by_index(gpu_id_);
if (fd == -1) return AMDSMI_STATUS_NOT_SUPPORTED;
amdsmi_status_t ret;
uint32_t fd = 0;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
return drm_.amdgpu_query_hw_ip(fd, info_id, hw_ip_type, size, value);
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_fw(unsigned info_id,
unsigned fw_type, unsigned size, void *value) const {
int fd = drm_.get_drm_fd_by_index(gpu_id_);
if (fd == -1) return AMDSMI_STATUS_NOT_SUPPORTED;
amdsmi_status_t ret;
uint32_t fd = 0;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;
return drm_.amdgpu_query_fw(fd, info_id, fw_type, size, value);
}
amdsmi_status_t AMDSmiGPUDevice::amdgpu_query_vbios(void *info) const {
int fd = drm_.get_drm_fd_by_index(gpu_id_);
if (fd == -1) return AMDSMI_STATUS_NOT_SUPPORTED;
amdsmi_status_t ret;
uint32_t fd = 0;
ret = drm_.get_drm_fd_by_index(gpu_id_, &fd);
if (ret != AMDSMI_STATUS_SUCCESS) return AMDSMI_STATUS_NOT_SUPPORTED;;
return drm_.amdgpu_query_vbios(fd, info);
}
+5
View File
@@ -55,6 +55,11 @@ AMDSmiSocket::~AMDSmiSocket() {
devices_.clear();
}
amdsmi_status_t AMDSmiSocket::get_device_count(uint32_t* device_count) const {
*device_count = static_cast<uint32_t>(devices_.size());
return AMDSMI_STATUS_SUCCESS;
}
} // namespace smi
} // namespace amd
+71 -34
View File
@@ -45,6 +45,7 @@
#include "amd_smi/impl/amd_smi_system.h"
#include "amd_smi/impl/amd_smi_gpu_device.h"
#include "rocm_smi/rocm_smi.h"
#include "rocm_smi/rocm_smi_main.h"
namespace amd {
@@ -54,55 +55,91 @@ namespace smi {
amdsmi_status_t AMDSmiSystem::init(uint64_t flags) {
init_flag_ = flags;
amdsmi_status_t amd_smi_status;
// populate sockets and devices
if (flags & AMDSMI_INIT_AMD_GPUS) {
drm_.init();
amd_smi_status = drm_.init();
// init rsmi
rsmi_status_t ret = rsmi_init(flags);
if (ret != RSMI_STATUS_SUCCESS) {
return static_cast<amdsmi_status_t>(ret);
}
uint32_t device_count = 0;
ret = rsmi_num_monitor_devices(&device_count);
if (ret != RSMI_STATUS_SUCCESS) {
return static_cast<amdsmi_status_t>(ret);
}
// libdrm is supported
if (amd_smi_status == AMDSMI_STATUS_SUCCESS) {
amd::smi::RocmSMI::getInstance().DiscoverAmdgpuDevices();
uint32_t device_count = amd::smi::RocmSMI::getInstance().devices().size();
for (uint32_t i=0; i < device_count; i++) {
std::stringstream ss;
//values for socked id are harcoded
ss << std::setfill('0') << std::uppercase << std::hex
<< std::setw(4) << drm_.get_bdfs()[i].domain_number << ":"
<< std::setw(2) << drm_.get_bdfs()[i].bus_number << ":"
<< std::setw(2) << drm_.get_bdfs()[i].device_number << "."
<< std::setw(2) << drm_.get_bdfs()[i].function_number;
for (uint32_t i=0; i < device_count; i++) {
uint64_t bdfid = 0;
ret = rsmi_dev_pci_id_get(i, &bdfid);
// Multiple devices may share the same socket
auto socket_id = ss.str();
AMDSmiSocket* socket = nullptr;
for (unsigned int j=0; j < sockets_.size(); j++) {
if (sockets_[j]->get_socket_id() == socket_id) {
socket = sockets_[j];
break;
}
}
if (socket == nullptr) {
socket = new AMDSmiSocket(ss.str());
sockets_.push_back(socket);
}
AMDSmiDevice* device = new AMDSmiGPUDevice(i, drm_);
socket->add_device(device);
devices_.insert(device);
}
}
else {
uint32_t device_count = 0;
ret = rsmi_num_monitor_devices(&device_count);
if (ret != RSMI_STATUS_SUCCESS) {
return static_cast<amdsmi_status_t>(ret);
}
uint64_t domain = (bdfid >> 32) & 0xffffffff;
uint64_t bus = (bdfid >> 8) & 0xff;
uint64_t device_id = (bdfid >> 3) & 0x1f;
uint64_t function = bdfid & 0x7;
std::stringstream ss;
ss << std::setfill('0') << std::uppercase << std::hex
<< std::setw(4) << domain << ":" << std::setw(2) << bus << ":"
<< std::setw(2) << device_id << "." << std::setw(2) << function;
// Multiple devices may share the same socket
auto socket_id = ss.str();
AMDSmiSocket* socket = nullptr;
for (unsigned int j=0; j < sockets_.size(); j++) {
if (sockets_[j]->get_socket_id() == socket_id) {
socket = sockets_[j];
break;
for (uint32_t i=0; i < device_count; i++) {
uint64_t bdfid = 0;
ret = rsmi_dev_pci_id_get(i, &bdfid);
if (ret != RSMI_STATUS_SUCCESS) {
return static_cast<amdsmi_status_t>(ret);
}
}
if (socket == nullptr) {
socket = new AMDSmiSocket(ss.str());
sockets_.push_back(socket);
}
AMDSmiDevice* device = new AMDSmiGPUDevice(i, drm_);
socket->add_device(device);
devices_.insert(device);
uint64_t domain = (bdfid >> 32) & 0xffffffff;
uint64_t bus = (bdfid >> 8) & 0xff;
uint64_t device_id = (bdfid >> 3) & 0x1f;
uint64_t function = bdfid & 0x7;
std::stringstream ss;
ss << std::setfill('0') << std::uppercase << std::hex
<< std::setw(4) << domain << ":" << std::setw(2) << bus << ":"
<< std::setw(2) << device_id << "." << std::setw(2) << function;
// Multiple devices may share the same socket
auto socket_id = ss.str();
AMDSmiSocket* socket = nullptr;
for (unsigned int j=0; j < sockets_.size(); j++) {
if (sockets_[j]->get_socket_id() == socket_id) {
socket = sockets_[j];
break;
}
}
if (socket == nullptr) {
socket = new AMDSmiSocket(ss.str());
sockets_.push_back(socket);
}
AMDSmiDevice* device = new AMDSmiGPUDevice(i, drm_);
socket->add_device(device);
devices_.insert(device);
}
}
} else {
return AMDSMI_STATUS_NOT_SUPPORTED;
+382
View File
@@ -0,0 +1,382 @@
/* * Copyright (C) 2022 Advanced Micro Devices. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy of
* this software and associated documentation files (the "Software"), to deal in
* the Software without restriction, including without limitation the rights to
* use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
* the Software, and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
* FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
* COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
* IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#include <errno.h>
#include <fcntl.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/mman.h>
#include <time.h>
#include <unistd.h>
#include <xf86drm.h>
#include <xf86drmMode.h>
#include <dirent.h>
#include <sys/types.h>
#include <memory>
#include <random>
#include <fstream>
#include <iostream>
#include <sstream>
#include <sys/ioctl.h>
#include <algorithm>
#include <string.h>
#include <limits.h>
#include "amd_smi/impl/amd_smi_utils.h"
#include "shared_mutex.h" // NOLINT
static const uint32_t kAmdGpuId = 0x1002;
static bool isAMDGPU(std::string dev_path) {
std::string vend_path = dev_path + "/device/vendor";
std::string vbios_v_path = dev_path + "/device/vbios_version";
if (!amd::smi::FileExists(vend_path.c_str())) {
return false;
}
if (!amd::smi::FileExists(vbios_v_path.c_str())) {
return false;
}
std::ifstream fs;
fs.open(vend_path);
if (!fs.is_open()) {
return false;
}
uint32_t vendor_id;
fs >> std::hex >> vendor_id;
fs.close();
if (vendor_id == kAmdGpuId) {
return true;
}
return false;
}
amdsmi_status_t smi_amdgpu_find_hwmon_dir(amd::smi::AMDSmiGPUDevice *device, std::string* full_path)
{
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
if (full_path == nullptr) {
return AMDSMI_STATUS_API_FAILED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
DIR *dh;
struct dirent * contents;
std::string device_path = "/sys/class/drm/" + device->get_gpu_path();
std::string directory_path = device_path + "/device/hwmon/";
if (!isAMDGPU(device_path)) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
dh = opendir(directory_path.c_str());
if (!dh) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
/*
First directory is '.', second directory is '..' and third directory is
valid directory for reading sysfs node
*/
while ((contents = readdir(dh)) != NULL) {
std::string name = contents->d_name;
if (name.find("hwmon", 0) != std::string::npos)
*full_path = directory_path + name;
}
closedir(dh);
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_board_info(amd::smi::AMDSmiGPUDevice* device, amdsmi_board_info_t *info) {
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
std::string product_name_path = "/sys/class/drm/" + device->get_gpu_path() + std::string("/device/product_name");
std::string product_number_path = "/sys/class/drm/" + device->get_gpu_path() + std::string("/device/product_number");
std::string serial_number_path = "/sys/class/drm/" + device->get_gpu_path() + std::string("/device/serial_number");
FILE *fp;
fp = fopen(product_name_path.c_str(), "rb");
if (!fp) {
fgets(info->product_name, sizeof(info->product_name), fp);
fclose(fp);
}
fp = fopen(product_number_path.c_str(), "rb");
if (!fp) {
fgets(info->model_number, sizeof(info->model_number), fp);
fclose(fp);
}
fp = fopen(serial_number_path.c_str(), "rb");
if (!fp) {
fscanf(fp, "%lx", &info->serial_number);
fclose(fp);
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_power_cap(amd::smi::AMDSmiGPUDevice* device, int *cap)
{
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
constexpr int DATA_SIZE = 10;
char val[DATA_SIZE];
std::string fullpath;
amdsmi_status_t ret = AMDSMI_STATUS_SUCCESS;
ret = smi_amdgpu_find_hwmon_dir(device, &fullpath);
SMIGPUDEVICE_MUTEX(device->get_mutex())
if (ret)
return ret;
fullpath += "/power1_cap_max";
std::ifstream file(fullpath.c_str(), std::ifstream::in);
if (!file.is_open()) {
printf("Failed to open file: %s \n", fullpath.c_str());
return AMDSMI_STATUS_API_FAILED;
}
file.getline(val, DATA_SIZE);
if (sscanf(val, "%d", cap) < 0) {
return AMDSMI_STATUS_API_FAILED;
}
// Dividing by 1000000 to get measurement in Watts
*cap /= 1000000;
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_ranges(amd::smi::AMDSmiGPUDevice* device, amdsmi_clk_type_t domain,
int *max_freq, int *min_freq, int *num_dpm)
{
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
std::string fullpath = "/sys/class/drm/" + device->get_gpu_path() + "/device";
char str[10];
unsigned int max, min, dpm;
switch (domain) {
case CLOCK_TYPE_GFX:
fullpath += "/pp_dpm_sclk";
break;
case CLOCK_TYPE_MEM:
fullpath += "/pp_dpm_mclk";
break;
case CLOCK_TYPE_VCLK0:
fullpath += "/pp_dpm_vclk";
break;
case CLOCK_TYPE_VCLK1:
fullpath += "/pp_dpm_vclk1";
break;
default:
return AMDSMI_STATUS_INVAL;
}
std::ifstream ranges(fullpath.c_str());
if (ranges.fail()) {
printf("Failed to open file: %s \n", fullpath.c_str());
return AMDSMI_STATUS_API_FAILED;
}
max = 0;
min = -1;
dpm = 0;
for (std::string line; getline(ranges, line);) {
unsigned int d, freq;
if (sscanf(line.c_str(), "%u: %d%s", &d, &freq, str) <= 2){
ranges.close();
return AMDSMI_STATUS_IO;
}
max = freq > max ? freq : max;
min = freq < min ? freq: min;
dpm = d > dpm ? d : dpm;
}
if (num_dpm)
*num_dpm = dpm;
if (max_freq)
*max_freq = max;
if (min_freq)
*min_freq = min;
ranges.close();
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_enabled_blocks(amd::smi::AMDSmiGPUDevice* device, uint64_t *enabled_blocks) {
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
std::string fullpath = "/sys/class/drm/" + device->get_gpu_path() + "/device/ras/features";
std::ifstream f(fullpath.c_str());
std::string tmp_str;
if (f.fail()) {
printf("Failed to open file: %s \n", fullpath.c_str());
return AMDSMI_STATUS_API_FAILED;
}
std::string line;
getline(f, line);
std::istringstream f1(line);
f1 >> tmp_str; // ignore
f1 >> tmp_str; // ignore
f1 >> tmp_str;
*enabled_blocks = strtoul(tmp_str.c_str(), nullptr, 16);
f.close();
if (*enabled_blocks == 0 || *enabled_blocks == ULONG_MAX) {
return AMDSMI_STATUS_API_FAILED;
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_bad_page_info(amd::smi::AMDSmiGPUDevice* device, uint32_t *num_pages, amdsmi_retired_page_record_t *info) {
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
std::string line;
std::vector<std::string> badPagesVec;
std::string fullpath = "/sys/class/drm/" + device->get_gpu_path() + std::string("/device/ras/gpu_vram_bad_pages");
std::ifstream fs(fullpath.c_str());
if (fs.fail()) {
printf("Failed to open file: %s \n", fullpath.c_str());
return AMDSMI_STATUS_NOT_SUPPORTED;
}
while (std::getline(fs, line)) {
badPagesVec.push_back(line);
}
if (badPagesVec.size() == 0) {
num_pages = 0;
return AMDSMI_STATUS_SUCCESS;
}
// Remove any *trailing* empty (whitespace) lines
while (badPagesVec.size() != 0 &&
badPagesVec.back().find_first_not_of(" \t\n\v\f\r") == std::string::npos) {
badPagesVec.pop_back();
}
*num_pages = static_cast<uint32_t>(badPagesVec.size());
if (info == nullptr) {
return AMDSMI_STATUS_SUCCESS;
}
char status_code;
amdsmi_memory_page_status_t tmp_stat;
std::string junk;
for (uint32_t i = 0; i < *num_pages; ++i) {
std::istringstream fs1(badPagesVec[i]);
fs1 >> std::hex >> info[i].page_address;
fs1 >> junk;
fs1 >> std::hex >> info[i].page_size;
fs1 >> junk;
fs1 >> status_code;
switch (status_code) {
case 'P':
tmp_stat = AMDSMI_MEM_PAGE_STATUS_PENDING;
break;
case 'F':
tmp_stat = AMDSMI_MEM_PAGE_STATUS_UNRESERVABLE;
break;
case 'R':
tmp_stat = AMDSMI_MEM_PAGE_STATUS_RESERVED;
break;
default:
return AMDSMI_STATUS_API_FAILED;
}
info[i].status = tmp_stat;
}
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t smi_amdgpu_get_ecc_error_count(amd::smi::AMDSmiGPUDevice* device, amdsmi_error_count_t *err_cnt) {
if (!device->check_if_drm_is_supported()) {
return AMDSMI_STATUS_NOT_SUPPORTED;
}
SMIGPUDEVICE_MUTEX(device->get_mutex())
char str[10];
std::string fullpath = "/sys/class/drm/" + device->get_gpu_path() + std::string("/device/ras/umc_err_count");
std::ifstream f(fullpath.c_str());
if (f.fail()) {
printf("Failed to open file: %s \n", fullpath.c_str());
return AMDSMI_STATUS_NOT_SUPPORTED;
}
std::string line;
getline(f, line);
sscanf(line.c_str(), "%s%ld", str, &(err_cnt->uncorrectable_count));
getline(f, line);
sscanf(line.c_str(), "%s%ld", str, &(err_cnt->correctable_count));
f.close();
return AMDSMI_STATUS_SUCCESS;
}
+264
View File
@@ -0,0 +1,264 @@
/* * Copyright (C) 2022 Advanced Micro Devices. All rights reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a copy of
* this software and associated documentation files (the "Software"), to deal in
* the Software without restriction, including without limitation the rights to
* use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of
* the Software, and to permit persons to whom the Software is furnished to do so,
* subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS
* FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR
* COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER
* IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
#include <sys/types.h>
#include <dirent.h>
#include <unistd.h>
#include <memory>
#include <vector>
#include <cstdlib>
#include <fstream>
#include <algorithm>
#include <string.h>
#include "amd_smi/amd_smi.h"
#include "amd_smi/impl/amd_smi_utils.h"
extern "C" {
amdsmi_status_t gpuvsmi_pid_is_gpu(const std::string &path, const char *bdf)
{
DIR *d;
struct dirent *dir;
d = opendir(path.c_str());
if (!d)
return AMDSMI_STATUS_NO_PERM;
/* iterate through all the fds, try to find
* a match for the GPU bdf
*/
while ((dir = readdir(d)) != NULL) {
std::string file = path + dir->d_name;
std::ifstream fdinfo(file.c_str());
for (std::string line; std::getline(fdinfo, line);) {
if (line.find(bdf) != std::string::npos) {
closedir(d);
return AMDSMI_STATUS_SUCCESS;
}
}
}
closedir(d);
return AMDSMI_STATUS_NOT_FOUND;
}
amdsmi_status_t gpuvsmi_get_pids(const amdsmi_bdf_t &bdf, std::vector<long int> &pids, uint64_t *size)
{
char bdf_str[13];
DIR *d;
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);
d = opendir("/proc");
if (!d)
return AMDSMI_STATUS_NO_PERM;
pids.clear();
/* Find the pid folders in /proc/ that we have access to */
while ((dir = readdir(d)) != NULL) {
if (dir->d_type == DT_DIR) {
/* Try to cast the name of the folder to a
* number, if it fails, it is not */
char *p;
long int pid;
pid = strtol(dir->d_name, &p, 10);
if (*p != 0)
continue;
/* Check if fdinfo is accesible */
std::string path = "/proc/" + std::string(dir->d_name) + "/fdinfo/";
if (access(path.c_str(), R_OK))
continue;
/* check if GPU is present */
if (gpuvsmi_pid_is_gpu(path, bdf_str))
continue;
pids.push_back(pid);
}
}
closedir(d);
*size = pids.size();
return AMDSMI_STATUS_SUCCESS;
}
amdsmi_status_t gpuvsmi_get_pid_info(const amdsmi_bdf_t &bdf, long int pid,
amdsmi_proc_info_t &info)
{
char bdf_str[13];
DIR *d;
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);
std::string path = "/proc/" + std::to_string(pid) + "/fdinfo/";
std::string name_path = "/proc/" + std::to_string(pid) + "/comm";
std::string cgroup_path = "/proc/" + std::to_string(pid) + "/cgroup";
if (gpuvsmi_pid_is_gpu(path.c_str(), bdf_str)) {
return AMDSMI_STATUS_INVAL;
}
d = opendir(path.c_str());
if (!d)
return AMDSMI_STATUS_NO_PERM;
/* Vectors to check if repated fd pasid */
std::vector<int> pasids;
memset(&info, 0, sizeof(info));
/* Iterate through all fdinfos */
while ((dir = readdir(d)) != NULL) {
std::string file = path + dir->d_name;
std::ifstream fdinfo(file.c_str());
for (std::string line; getline(fdinfo, line);) {
if (line.find("pasid:") != std::string::npos) {
int pasid;
if (sscanf(line.c_str(), "pasid: %d", &pasid) != 1)
continue;
auto it = std::find(pasids.begin(), pasids.end(), pasid);
if (it == pasids.end())
pasids.push_back(pasid);
} else if (line.find("gtt mem:") != std::string::npos) {
unsigned long mem;
if (sscanf(line.c_str(), "gtt mem: %lu", &mem) != 1)
continue;
info.mem += mem * 1024;
info.memory_usage.gtt_mem += mem * 1024;
} else if (line.find("cpu mem:") != std::string::npos) {
unsigned long mem;
if (sscanf(line.c_str(), "cpu mem: %lu", &mem) != 1)
continue;
info.mem += mem * 1024;
info.memory_usage.cpu_mem += mem * 1024;
} else if (line.find("vram mem:") != std::string::npos) {
unsigned long mem;
if (sscanf(line.c_str(), "vram mem: %lu", &mem) != 1)
continue;
info.mem += mem * 1024;
info.memory_usage.vram_mem += mem * 1024;
} else if (line.find("gfx") != std::string::npos) {
float usage;
int ring;
if (sscanf(line.c_str(), "gfx%d: %f%%", &ring, &usage) != 2)
continue;
if (ring >= AMDSMI_MAX_MM_IP_COUNT)
continue;
info.engine_usage.gfx[ring] += (uint16_t)(usage * 100);
} else if (line.find("compute") != std::string::npos) {
float usage;
int ring;
if (sscanf(line.c_str(), "compute%d: %f%%", &ring, &usage) != 2)
continue;
if (ring >= AMDSMI_MAX_MM_IP_COUNT)
continue;
info.engine_usage.compute[ring] += (uint16_t)(usage * 100);
} else if (line.find("dma") != std::string::npos) {
float usage;
int ring;
if (sscanf(line.c_str(), "dma%d: %f%%", &ring, &usage) != 2)
continue;
if (ring >= AMDSMI_MAX_MM_IP_COUNT)
continue;
info.engine_usage.sdma[ring] += (uint16_t)(usage * 100);
} else if (line.find("enc") != std::string::npos) {
float usage;
int ring;
if (sscanf(line.c_str(), "enc%d: %f%%", &ring, &usage) != 2)
continue;
if (ring >= AMDSMI_MAX_MM_IP_COUNT)
continue;
info.engine_usage.enc[ring] += (uint16_t)(usage * 100);
} else if (line.find("dec") != std::string::npos) {
float usage;
int ring;
if (sscanf(line.c_str(), "dec%d: %f%%", &ring, &usage) != 2)
continue;
if (ring >= AMDSMI_MAX_MM_IP_COUNT)
continue;
info.engine_usage.dec[ring] += (uint16_t)(usage * 100);
}
}
}
closedir(d);
if (!pasids.size())
return AMDSMI_STATUS_NOT_FOUND;
std::ifstream filename(name_path.c_str());
std::string name;
getline(filename, name);
if (name.empty())
return AMDSMI_STATUS_API_FAILED;
strncpy(info.name, name.c_str(), std::min(
(unsigned long) AMDSMI_NORMAL_STRING_LENGTH,
name.length()));
info.pid = (uint32_t)pid;
return AMDSMI_STATUS_SUCCESS;
}
} // extern "C"
@@ -160,7 +160,7 @@ void TestErrCntRead::Run(void) {
<< std::endl;
std::cout << "\t\tCorrectable errors: " << ec.correctable_err
<< std::endl;
std::cout << "\t\tUncorrectable errors: " << ec.uncorrectable_err
std::cout << "\t\tUncorrectable errors: " << ec.uncorrectable_count
<< std::endl;
}
// Verify api support checking functionality is working