support HMM profiling event #96
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Install LICENSE.txt to share/doc/smi-lib Change-Id: Idcbb70db8808111203e8e4a4c3ab4d1e070ac79d
Add rpm License header for cpack Change-Id: I2f4a89015b6389cfde801f41d4f6e0f59e7087aa
pop_back() was causing a seg fault when pp_dpm_pcie file is empty and returns whitespace. Signed-off-by: Divya Shikre <DivyaUday.Shikre@amd.com> Change-Id: I888f1f79751cd456e43751a5b96d08560a039677 (cherry picked from commit ec71380)
SMI lib function rsmi_event_notification_get read events from all GPUs, each event returned with device dv_idx. Currently we create read thread for each GPU, it is not necessary because each thread reads same events, and each thread display events from other GPUs with incorrect GPU index. Create one read thread for multiple GPUs, and display event with correct GPU index received from data.dv_idx. Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Update kfd_ioctl.h from KFD to add HMM migration and recoverable page fault, queue eviction and restore event, and event triggers defines. Update rocm_smi.h to add new SMI notification events and triggers defines, with the same enum value as kfd_ioctl.h, to avoid value translation in smi lib. Change fscanf %63s format to %MAX_EVENT_NOTIFICATION_MSG_SIZE[^\n] to read entire line as one message. Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Use SMI_EVENT_ALL_PROCESS to receive event from all processes because HMM migration events are per process event, KFD requires this flag plus super user premission to receive events from other process, so showevents to relaunchAsSudo if arguments are in new event list. User can specify event name in short format, for example "--showevents migrate" will show MIGRATE_START, MIGRATE_END events. Define message size using macro from rocm_smi.h Signed-off-by: Philip Yang <Philip.Yang@amd.com>
Add new event names defines, set event mask RSMI_EVT_NOTIF_ALL_PROCESS to receive events from all processes. Add protection check in case new event type returns from KFD, to avoid out of range access segmentation fault. Signed-off-by: Philip Yang <Philip.Yang@amd.com>
bill-shuzhou-liu
force-pushed
the
master
branch
from
66e101a
to
bd95425
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Changes:
rocm smi lib: Define HMM profiling events
rocm_smi.py: showevents support HMM profiling events
unit test: TestEvtNotifReadWrite support HMM profiling events