Add flag ROC_SIGNAL_POOL_SIZE_PROFILE and update default value #71
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
The change is two-fold: - add a flag to configure the pool size used when profiling, this allows to clearly configure the two values (profile vs. not-profile). - change the size of the profile pool down from 4096 which was too large: the kernel only provides 4094 events for DGPUs, *and* using two command-queues in OpenCL results in the bug described here: ROCm/ROCR-Runtime#186 The new default of 1000 has this rationale: it allows up to 4 queues to fit within the 4094 events provided by the kernel (with a little margin).
|
Thanks for digging this up, I have an internal change that makes #signals not just for profiling but for everything to 64 per queue. That would fix your issue and also addresses the issue in ROCr, I would guess. |
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Outdated suggestions cannot be applied.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.
The change is two-fold:
add a flag to configure the pool size used when profiling, this allows to clearly configure the two values (profile vs. not-profile).
change the size of the profile pool down from 4096 which was too large: the kernel only provides 4094 events for DGPUs, and using two command-queues in OpenCL results in the bug described here:
OpenCL hot loop (100% one thread) when using two command queues with profiling ROCR-Runtime#186
The new default of 1000 has this rationale: it allows up to 4 queues to fit within the 4094 events provided by the kernel (with a little margin).