The bug was reproduced like this.
In terminal #1, run command:
sudo amd-smi ras --cper --gpu 6 --severity all --folder /tmp/cper_dump --follow
In terminal #2, inject errors:
while true; do sudo amdgpuras -b 7 -s 1 -m 6 -t 2; sleep 2; done
The terminal #1 starts dumping cper entry information that it captures. After 20 entries have been captured, open terminal #3 and run same command as terminal #1:
sudo amd-smi ras --cper --gpu 6 --severity all --folder /tmp/cper_dump --follow
From terminal #3, there will be no output, even when terminal #1 continues capturing and printing information.
The fix:
Since we already have more than 20 CPER entries available in the GPU buffer, when we run the command from terminal #3 to start capturing from the beginning and pass 20 buffers to copy entries to, the C++ API returns a code saying there is more data available.
The Python CLI should not treat this as an error, but should continue to print what the API returned.
---------
Signed-off-by: Oosman Saeed <oossaeed@amd.com>
[ROCm/amdsmi commit: 5b95d227bc]
Topology numa_bw checks for non-xgmi links to set as N/A.
The recent change in link_type enum mapping caused this
condition to check for PCIE instead of XGMI. Corrected
the same.
Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>
[ROCm/amdsmi commit: fa9ca21520]
The topology method calls amdsmi_topo_get_p2p_status repeatedly
for the same GPU pairs across different table sections,
significantly impacting performance with 60+ GPUs. Reduced this
by implemeting result caching.
Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>
[ROCm/amdsmi commit: c3453f7c97]
New:
- gpu_cache_read.h and gpu_cache_read.cc
- Test reads GPU cache info and asserts valid structure
Updated:
- integration_test.py
- Added test_gpu_cache_info() and asserts valid structure
- test_get_gpu_compute_partition() to loop through all devices when test fail/pass
Added:
- test_get_gpu_compute_partition_returns_string() to integration_test.py
- This test displays the current compute partition for each bdf
---------
Signed-off-by: Juan Castillo <juan.castillo@amd.com>
Signed-off-by: Castillo, Juan <Juan.Castillo@amd.com>
Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
[ROCm/amdsmi commit: 470c62f887]
* Made corrections to reading of pp_od_clk_voltage
* Added fall back to pp_dpm files if pp_od_clk_voltage doesn't exist
---------
Signed-off-by: gabrpham_amdeng <Gabriel.Pham@amd.com>
Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
[ROCm/amdsmi commit: 4262aee8f5]
The custom_dump function was not printing list's key
and so static numa was not displaying list keys
CPU affinity and Socket affinity. Updated custom_dump
to print the keys.
Signed-off-by: Bindhiya Kanangot Balakrishnan <Bindhiya.KanangotBalakrishnan@amd.com>
[ROCm/amdsmi commit: 6fbda16098]
* Added copyrights
* Fixed type hinting for processor_handle in python_interface
* Fixed Incorrect type hinting to actual return types
---------
Signed-off-by: Arif, Maisam <Maisam.Arif@amd.com>
Change-Id: Ie2a09acf628ed0c43eacc8ec78c159d125acbcdb
[ROCm/amdsmi commit: 23b9da656c]
* CMAKE - Remove example build from src/CMakeLists.txt
For some reason it was building examples every time even when not
necessary...
* CMAKE - Format
* Fix drm_example broken PRIu32
* CMAKE - Do NOT create lib64 when building examples
* CMAKE - Examples should only install C and CMake files
---------
Change-Id: I6274b72a085a41b5bd5ae698af798f60a8a092a0
Signed-off-by: Galantsev, Dmitrii <dmitrii.galantsev@amd.com>
[ROCm/amdsmi commit: f9b8066c26]
* Adjusted help text
* Adjusted --afid to run only with --cper-file
* Fixed interface return error
Signed-off-by: Maisam Arif <Maisam.Arif@amd.com>
Change-Id: I2b96f4515c85f3b9dd84ba5c2d819729a997141b
[ROCm/amdsmi commit: ac63f410c2]
Narlo, Joseph
Saeed, Oosman
Cheruvally, Aravindan
gabrpham_amdeng
Kanangot Balakrishnan, Bindhiya
Justin Williams
Jeremy Newton
Maisam Arif
Galantsev, Dmitrii
Castillo, Juan
Justin Williams
Deepak Mewar
dependabot[bot]