AI/rocm-systems
2
0
Forka 0
Codice Problemi Pull Request Actions Pacchetti Progetti Rilasci Wiki Attività

GPU Usage/activity update

Sfoglia il codice sorgente 
CLI:
Every usage field is notated by "activity"
gfx_usage -> gfx_activity
umc_usage -> umc_activity
vcn_activities -> vcn_activity
jpeg_activities[AID#] -> jpeg_activity

Wrapper: fixed metric output, misalignment
with generator

update_wrapper.sh:
DOCKER_BUILDKIT to 0 (if unset)

API:
amdsmi_get_gpu_metrics_info:
1.3: Removed commenting out avg socket power

Signed-off-by: Charis Poag <Charis.Poag@amd.com>
Change-Id: Id3fcc20aef420c7b7a90ba22fa3bc643b2716333
This commit is contained in:
Charis Poag
2024-01-15 23:14:44 -06:00
parent 5a6b5d2a0a
commit 4575990ae7
6 ha cambiato i file con 23 aggiunte e 29 eliminazioni
Scarica il file Patch Scarica il file Diff Expand all files Collapse all files
py-interface/README.md
+3 -3
Vedi File
@@ -2126,8 +2126,8 @@ Output: Dictionary with fields
`temperature_vrsoc` | vrsoc temperature value | Celsius (C)
`temperature_vrmem` | vrmem temperature value | Celsius (C)
`average_gfx_activity` | Average gfx activity | %
`average_umc_activity` | Average umc activity | %
`average_mm_activity` | Average mm activity | %
`average_umc_activity` | Average umc (Universal Memory Controller) activity | %
`average_mm_activity` | Average mm (multimedia) engine activity | %
`average_socket_power` | Average socket power | W
`energy_accumulator` | Energy accumulated with a 15.3 uJ resolution over 1ns | uJ
`system_clock_counter` | System clock counter | ns
@@ -2178,7 +2178,7 @@ Output: Dictionary with fields
`mem_max_bandwidth` | Maximum memory bandwidth usage accumulated | GB/s
`pcie_nak_sent_count_acc` | PCIe NAC sent count accumulated |
`pcie_nak_rcvd_count_acc` | PCIe NAC received count accumulated |
`jpeg_activitys[AID<X>]` | List of JPEG engine activity for each AID (X=0-3) | %
`jpeg_activity` | List of JPEG engine activity | %
Exceptions that can be thrown by `amdsmi_get_gpu_metrics_info` function:
py-interface/amdsmi_interface.py
+1 -4
Vedi File
@@ -3242,10 +3242,7 @@ def amdsmi_get_gpu_metrics_info(
"mem_max_bandwidth": gpu_metrics.mem_max_bandwidth,
"pcie_nak_sent_count_acc": gpu_metrics.pcie_nak_sent_count_acc,
"pcie_nak_rcvd_count_acc": gpu_metrics.pcie_nak_rcvd_count_acc,
"jpeg_activities[AID0]": list(gpu_metrics.jpeg_activities)[:8],
"jpeg_activities[AID1]": list(gpu_metrics.jpeg_activities)[8:16],
"jpeg_activities[AID2]": list(gpu_metrics.jpeg_activities)[16:24],
"jpeg_activities[AID3]": list(gpu_metrics.jpeg_activities)[24:32],
"jpeg_activity": list(gpu_metrics.jpeg_activity),
}
py-interface/amdsmi_wrapper.py
+12 -12
Vedi File
@@ -906,7 +906,6 @@ amdsmi_clk_info_t = struct_amdsmi_clk_info_t
class struct_amdsmi_engine_usage_t(Structure):
pass
struct_amdsmi_engine_usage_t._pack_ = 1 # source:False
struct_amdsmi_engine_usage_t._fields_ = [
('gfx_activity', ctypes.c_uint32),
@@ -920,6 +919,16 @@ amdsmi_process_handle_t = ctypes.c_uint32
class struct_amdsmi_proc_info_t(Structure):
pass
class struct_engine_usage_(Structure):
pass
struct_engine_usage_._pack_ = 1 # source:False
struct_engine_usage_._fields_ = [
('gfx', ctypes.c_uint64),
('enc', ctypes.c_uint64),
('reserved', ctypes.c_uint32 * 12),
]
class struct_memory_usage_(Structure):
pass
@@ -931,16 +940,6 @@ struct_memory_usage_._fields_ = [
('reserved', ctypes.c_uint32 * 10),
]
class struct_engine_usage_(Structure):
pass
struct_engine_usage_._pack_ = 1 # source:False
struct_engine_usage_._fields_ = [
('gfx', ctypes.c_uint64),
('enc', ctypes.c_uint64),
('reserved', ctypes.c_uint32 * 12),
]
struct_amdsmi_proc_info_t._pack_ = 1 # source:False
struct_amdsmi_proc_info_t._fields_ = [
('name', ctypes.c_char * 32),
@@ -1527,11 +1526,12 @@ struct_amdsmi_gpu_metrics_t._fields_ = [
('current_socclks', ctypes.c_uint16 * 4),
('current_vclk0s', ctypes.c_uint16 * 4),
('current_dclk0s', ctypes.c_uint16 * 4),
('jpeg_activity', ctypes.c_uint16 * 32),
('mem_bandwidth_acc', ctypes.c_uint64),
('mem_max_bandwidth', ctypes.c_uint32),
('pcie_nak_sent_count_acc', ctypes.c_uint32),
('pcie_nak_rcvd_count_acc', ctypes.c_uint32),
('jpeg_activities', ctypes.c_uint16 * 32),
('PADDING_4', ctypes.c_ubyte * 4),
]
amdsmi_gpu_metrics_t = struct_amdsmi_gpu_metrics_t