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

SWDEV-518209: GPU Metrics 1.8 (#177)

Browse Source 
- Updates:
    - Adding the following metrics to allow new calculations for violation status:
        - Per XCP metrics gfx_below_host_limit_ppt_acc
        - Per XCP metrics gfx_below_host_limit_thm_acc
        - Per XCP metrics gfx_low_utilization_acc
        - Per XCP metrics gfx_below_host_limit_total_acc
    - Increasing available JPEG engines to 40. Current ASICs may not support all 40. These will be indicated as UINT16_MAX or N/A in CLI.

Signed-off-by: Juan Castillo <juan.castillo@amd.com>
Co-authored-by: Charis Poag <Charis.Poag@amd.com>
This commit is contained in:
Castillo, Juan
2025-03-19 10:24:02 -05:00
committed by GitHub
parent fba8b3f6e1
commit 7c882b2f69
9 changed files with 857 additions and 690 deletions
Download Patch File Download Diff File Expand all files Collapse all files
py-interface/amdsmi_interface.py
+31 -2
View File
@@ -4200,6 +4200,10 @@ def amdsmi_get_gpu_metrics_info(
"xcp_stats.vcn_busy": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_busy_acc": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_below_host_limit_acc": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_below_host_limit_ppt_acc": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_below_host_limit_thm_acc": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_low_utilization_acc": list(gpu_metrics.xcp_stats),
"xcp_stats.gfx_below_host_limit_total_acc": list(gpu_metrics.xcp_stats),
"pcie_lc_perf_other_end_recovery": _validate_if_max_uint(gpu_metrics.pcie_lc_perf_other_end_recovery, MaxUIntegerTypes.UINT32_T),
"vram_max_bandwidth": _validate_if_max_uint(gpu_metrics.vram_max_bandwidth, MaxUIntegerTypes.UINT64_T),
"xgmi_link_status": _validate_if_max_uint(list(gpu_metrics.xgmi_link_status), MaxUIntegerTypes.UINT16_T),
@@ -4231,15 +4235,40 @@ def amdsmi_get_gpu_metrics_info(
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_busy_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_busy_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T, isActivity=True))
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output["xcp_stats.gfx_busy_acc"][xcp_index] = xcp_detail
if 'xcp_stats.gfx_below_host_limit_acc' in gpu_metrics_output:
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_below_host_limit_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_below_host_limit_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T, isActivity=True))
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output['xcp_stats.gfx_below_host_limit_acc'][xcp_index] = xcp_detail
# new for gpu metrics v1.8
if 'xcp_stats.gfx_below_host_limit_ppt_acc' in gpu_metrics_output:
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_below_host_limit_ppt_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_below_host_limit_ppt_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output['xcp_stats.gfx_below_host_limit_ppt_acc'][xcp_index] = xcp_detail
if 'xcp_stats.gfx_below_host_limit_thm_acc' in gpu_metrics_output:
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_below_host_limit_thm_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_below_host_limit_thm_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output['xcp_stats.gfx_below_host_limit_thm_acc'][xcp_index] = xcp_detail
if 'xcp_stats.gfx_low_utilization_acc' in gpu_metrics_output:
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_low_utilization_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_low_utilization_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output['xcp_stats.gfx_low_utilization_acc'][xcp_index] = xcp_detail
if 'xcp_stats.gfx_below_host_limit_total_acc' in gpu_metrics_output:
for xcp_index, xcp_metrics in enumerate(gpu_metrics_output['xcp_stats.gfx_below_host_limit_total_acc']):
xcp_detail = []
for val in xcp_metrics.gfx_below_host_limit_total_acc:
xcp_detail.append(_validate_if_max_uint(val, MaxUIntegerTypes.UINT64_T))
gpu_metrics_output['xcp_stats.gfx_below_host_limit_total_acc'][xcp_index] = xcp_detail
return gpu_metrics_output
py-interface/amdsmi_wrapper.py
+5 -1
View File
@@ -1849,10 +1849,14 @@ class struct_amdsmi_gpu_xcp_metrics_t(Structure):
struct_amdsmi_gpu_xcp_metrics_t._pack_ = 1 # source:False
struct_amdsmi_gpu_xcp_metrics_t._fields_ = [
('gfx_busy_inst', ctypes.c_uint32 * 8),
('jpeg_busy', ctypes.c_uint16 * 32),
('jpeg_busy', ctypes.c_uint16 * 40),
('vcn_busy', ctypes.c_uint16 * 4),
('gfx_busy_acc', ctypes.c_uint64 * 8),
('gfx_below_host_limit_acc', ctypes.c_uint64 * 8),
('gfx_below_host_limit_ppt_acc', ctypes.c_uint64 * 8),
('gfx_below_host_limit_thm_acc', ctypes.c_uint64 * 8),
('gfx_low_utilization_acc', ctypes.c_uint64 * 8),
('gfx_below_host_limit_total_acc', ctypes.c_uint64 * 8),
]
amdsmi_gpu_xcp_metrics_t = struct_amdsmi_gpu_xcp_metrics_t