Arm backend: Add event profiling to VGF backend#19703
Open
wwwind wants to merge 2 commits into
Open
Conversation
Signed-off-by: Elena Zhelezina <elena.zhelezina@arm.com> Change-Id: I26c6c8857744e91911daa4cf52ce7260e452e72e
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/19703
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 1 New FailureAs of commit af840eb with merge base 1feb56c ( NEW FAILURE - The following job has failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
Bot
added
the
CLA Signed
wwwind
added
partner: arm
meta-cla
Bot
added
the
CLA Signed
zingo
approved these changes
May 21, 2026
digantdesai
reviewed
May 21, 2026
| } | ||
|
|
||
| bool VgfRepr::init_timestamp_queries() { | ||
| const char* enable = std::getenv("EXECUTORCH_VGF_ENABLE_TIMESTAMP_QUERIES"); |
Contributor
There was a problem hiding this comment.
did you consider backend options as opposed to envvar?
digantdesai
reviewed
May 21, 2026
|
|
||
| if (!valid_vgf) { | ||
| #ifdef ET_EVENT_TRACER_ENABLED | ||
| event_tracer_end_profiling_delegate(event_tracer, init_total_event); |
Contributor
There was a problem hiding this comment.
Do you see nested profiling events leading to some noise? I.e. the time spent in recording the inner profiling event measured in the outer one? I understand this is needed for chrometrace etc. Just want to make sure this record overhead isn't too much.
digantdesai
reviewed
May 21, 2026
|
|
||
| ## Profiling of VGF Backend | ||
|
|
||
| VGF profiling now emits both host-side ExecuTorch event tracer ranges and Vulkan timestamp-query measurements. The host ranges split init into `VGF_INIT_*` phases, including `VGF_INIT_CREATE_DATA_GRAPH_PIPELINE`, and split execute into `VGF_COPY_INPUTS`, `VGF_QUEUE_SUBMIT`, `VGF_QUEUE_WAIT_IDLE`, `VGF_TIMESTAMP_QUERY_READBACK`, `VGF_DISPATCH_AND_WAIT`, and `VGF_COPY_OUTPUTS`. Vulkan timestamp queries are inserted into the recorded VGF command buffer around `vkCmdDispatchDataGraphARM()`, producing `VGF_DATA_GRAPH_DEVICE_TIME`, which measures device-side elapsed time for the submitted data-graph command buffer region. To collect a profile, build the VGF runner with event tracing enabled, run the model with an ETDump path, then convert the ETDump to Chrome trace JSON: |
Contributor
There was a problem hiding this comment.
Are you planning to expand this to hardware performance counters like occupancy going beyond time measurements, and chrometrace.
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Add event profiling to VGF backend.
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani