[NFC][SYCL] Cleanup device_impl's properties caching
#18450
Closed
Conversation
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
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
* Cache extensions
* Pre-initialize all in the ctor instead of using `std::call_once`
* Minor code cleanups
For
```
bool a = true;
for (int i = 0; i < 10; ++i) {
SimpleTimer t;
a &= d.has(sycl::aspect::fp16);
}
```
this results in 600ns -> 42..50ns improvement. Caching fp16 aspect directly (and
avoiding `std::find_if` over extensions vector) can bring it further down to
about 35..45ns, but, IMO, it isn't worth extra complexity.
aelovikov-intel
force-pushed
the
get-info-impl
branch
from
14e6b26 to
334f690
Compare
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
aelovikov-intel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
slawekptak
reviewed
May 14, 2025
| mutable ext::oneapi::experimental::architecture MDeviceArch{}; | ||
| mutable std::once_flag MDeviceArchFlag; | ||
|
|
||
| // TODO: Does this have a race? |
There was a problem hiding this comment.
Is there a race condition here? Could you please clarify what it is exactly?
|
|
||
| // Pre-compute some often used properties. | ||
|
|
||
| // Is used during submission. |
There was a problem hiding this comment.
This is about kernel submission to a queue, right? I think it would be helpful to clarify this in the comment.
|
Alternative approach: #18476 |
aelovikov-intel
added a commit
that referenced
this pull request
May 19, 2025
I think this is better than a simpler #18450 as it allows much easier experiments/tuning on what should be cached and how.
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.
std::find_ifover extensions vector) can bring it further down to about 35..45ns, but, IMO, it isn't worth extra complexity.std::call_oncedevice_impl::has_extensionto acceptstd::string_view