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

Added memory information to Guest

Browse Source 
Signed-off-by: Maisam Arif <maisarif@amd.com>
Change-Id: I3e6b4bba3746c9498d34e6d2840461489eb3bc61
Signed-off-by: Maisam Arif <maisarif@amd.com>


[ROCm/amdsmi commit: 02b4b67ca0]
This commit is contained in:
Maisam Arif
2023-10-05 02:13:36 -05:00
committed by Galantsev, Dmitrii
parent 657d011f57
commit 8a9387cdec
3 changed files with 41 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/amdsmi/amdsmi_cli/amdsmi_commands.py
+37 -24
View File
@@ -167,6 +167,8 @@ class AMDSMICommands():
args.driver = driver
if numa:
args.numa = numa
if vram:
args.vram = vram
if self.helpers.is_linux() and self.helpers.is_baremetal():
if ras:
args.ras = ras
@@ -174,8 +176,6 @@ class AMDSMICommands():
args.limit = limit
if board:
args.board = board
if vram:
args.vram = vram
# Handle No GPU passed
if args.gpu == None:
@@ -192,8 +192,8 @@ class AMDSMICommands():
if not any([args.asic, args.bus, args.vbios, args.limit, args.driver, args.ras, args.board, args.numa, args.vram]):
args.asic = args.bus = args.vbios = args.limit = args.driver = args.ras = args.board = args.numa = args.vram = self.all_arguments = True
if self.helpers.is_linux() and self.helpers.is_virtual_os():
if not any([args.asic, args.bus, args.vbios, args.driver]):
args.asic = args.bus = args.vbios = args.driver = self.all_arguments = True
if not any([args.asic, args.bus, args.vbios, args.driver, args.vram]):
args.asic = args.bus = args.vbios = args.driver = args.vram = self.all_arguments = True
static_dict = {}
@@ -401,6 +401,39 @@ class AMDSMICommands():
logging.debug("Failed to get driver info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['driver'] = driver_info
if args.vram:
try:
vram_info = amdsmi_interface.amdsmi_get_gpu_vram_info(args.gpu)
# Get vram type string
vram_type_enum = vram_info['vram_type']
if vram_type_enum == amdsmi_interface.amdsmi_wrapper.VRAM_TYPE_GDDR6:
vram_type = "GDDR6"
else:
vram_type = amdsmi_interface.amdsmi_wrapper.amdsmi_vram_type_t__enumvalues[vram_type_enum]
# Remove amdsmi enum prefix
vram_type = vram_type.replace('VRAM_TYPE_', '').replace('_', '')
# Get vram vendor string
vram_vendor_enum = vram_info['vram_vendor']
vram_vendor = amdsmi_interface.amdsmi_wrapper.amdsmi_vram_vendor_type_t__enumvalues[vram_vendor_enum]
if "PLACEHOLDER" in vram_vendor:
vram_vendor = "N/A"
else:
# Remove amdsmi enum prefix
vram_vendor = vram_vendor.replace('AMDSMI_VRAM_VENDOR__', '')
vram_info['vram_type'] = vram_type
vram_info['vram_vendor'] = vram_vendor
if self.logger.is_human_readable_format():
vram_info['vram_size_mb'] = f"{vram_info['vram_size_mb']} MB"
except amdsmi_exception.AmdSmiLibraryException as e:
vram_info = "N/A"
logging.debug("Failed to get vram info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['vram'] = vram_info
if self.helpers.is_hypervisor() or self.helpers.is_baremetal():
if args.ras:
try:
@@ -408,26 +441,6 @@ class AMDSMICommands():
except amdsmi_exception.AmdSmiLibraryException as e:
static_dict['ras'] = "N/A"
logging.debug("Failed to get ras block features for gpu %s | %s", gpu_id, e.get_error_info())
if args.vram:
try:
vram_info = amdsmi_interface.amdsmi_get_gpu_vram_info(args.gpu)
vram_type_enum = vram_info['vram_type']
vram_vendor_enum = vram_info['vram_vendor']
vram_type = amdsmi_interface.amdsmi_wrapper.amdsmi_vram_type_t__enumvalues[vram_type_enum]
vram_vendor = amdsmi_interface.amdsmi_wrapper.amdsmi_vram_vendor_type_t__enumvalues[vram_vendor_enum]
# Remove amdsmi enum prefix
vram_info['vram_type'] = vram_type.replace('VRAM_TYPE_', '').replace('_', '')
vram_info['vram_vendor'] = vram_vendor.replace('AMDSMI_VRAM_VENDOR__', '')
if self.logger.is_human_readable_format():
vram_info['vram_size_mb'] = f"{vram_info['vram_size_mb']} MB"
except amdsmi_exception.AmdSmiLibraryException as e:
vram_info = "N/A"
logging.debug("Failed to get vram info for gpu %s | %s", gpu_id, e.get_error_info())
static_dict['vram'] = vram_info
if self.helpers.is_linux() and self.helpers.is_baremetal():
if args.numa:
try:
projects/amdsmi/amdsmi_cli/amdsmi_parser.py
+1 -1
View File
@@ -320,11 +320,11 @@ class AMDSMIParser(argparse.ArgumentParser):
static_parser.add_argument('-b', '--bus', action='store_true', required=False, help=bus_help)
static_parser.add_argument('-V', '--vbios', action='store_true', required=False, help=vbios_help)
static_parser.add_argument('-d', '--driver', action='store_true', required=False, help=driver_help)
static_parser.add_argument('-v', '--vram', action='store_true', required=False, help=vram_help)
# Options to display on Hypervisors and Baremetal
if self.helpers.is_hypervisor() or self.helpers.is_baremetal():
static_parser.add_argument('-r', '--ras', action='store_true', required=False, help=ras_help)
static_parser.add_argument('-v', '--vram', action='store_true', required=False, help=vram_help)
if self.helpers.is_linux():
static_parser.add_argument('-B', '--board', action='store_true', required=False, help=board_help)
static_parser.add_argument('-l', '--limit', action='store_true', required=False, help=limit_help)
projects/amdsmi/src/amd_smi/amd_smi.cc
+3 -1
View File
@@ -805,7 +805,9 @@ amdsmi_status_t amdsmi_get_gpu_vram_info(
uint64_t total = 0;
r = rsmi_wrapper(rsmi_dev_memory_total_get, processor_handle,
RSMI_MEM_TYPE_VRAM, &total);
info->vram_size_mb = total / (1024 * 1024);
if (r == AMDSMI_STATUS_SUCCESS) {
info->vram_size_mb = total / (1024 * 1024);
}
return AMDSMI_STATUS_SUCCESS;
}