-
Notifications
You must be signed in to change notification settings - Fork 734
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[SYCL] Support sycl::kernel_bundle for multi-device scenario #15546
Open
againull
wants to merge
3
commits into
intel:sycl
Choose a base branch
from
againull:sycl_bundle_multi_device
base: sycl
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
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
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
September 28, 2024 00:47
464b6dc
to
02a5333
Compare
againull
had a problem deploying
to
WindowsCILock
September 28, 2024 00:47
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
September 30, 2024 05:13
02a5333
to
5b5df33
Compare
againull
had a problem deploying
to
WindowsCILock
September 30, 2024 05:14
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
September 30, 2024 05:36
5b5df33
to
ac8444b
Compare
againull
had a problem deploying
to
WindowsCILock
September 30, 2024 05:36
— with
GitHub Actions
Failure
againull
had a problem deploying
to
WindowsCILock
September 30, 2024 05:53
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
2 times, most recently
from
September 30, 2024 20:18
3cfe9b7
to
5e4b86e
Compare
againull
had a problem deploying
to
WindowsCILock
September 30, 2024 20:18
— with
GitHub Actions
Error
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
September 30, 2024 20:22
5e4b86e
to
2b387e4
Compare
againull
temporarily deployed
to
WindowsCILock
September 30, 2024 20:23
— with
GitHub Actions
Inactive
againull
had a problem deploying
to
WindowsCILock
September 30, 2024 20:53
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
October 1, 2024 07:26
2b387e4
to
b8e0652
Compare
againull
had a problem deploying
to
WindowsCILock
October 1, 2024 07:26
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
October 1, 2024 07:28
b8e0652
to
080bd06
Compare
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
October 1, 2024 07:37
080bd06
to
9f17f3f
Compare
againull
had a problem deploying
to
WindowsCILock
October 1, 2024 07:38
— with
GitHub Actions
Failure
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
October 1, 2024 08:05
9f17f3f
to
687d688
Compare
againull
had a problem deploying
to
WindowsCILock
October 1, 2024 08:05
— with
GitHub Actions
Failure
This PR includes: * Changes in the program manager methods to be able to properly create/build UR program for multiple devices. So far, we were mostly using only the first device in the vector to create/build UR program which made UR program unusable on other devices. * UR tag update brings the version of urProgramCreateWithBinary which allows to create UR program from multiple device binaries. * Our program cache key allowed only a single device. I have changed it to contain a set of devices. If UR program is created and built for a set of devices then the same UR program is usable whenver we have any subset of this set. That's why if we have a program built for a set of devices then add all subsets to the cache. Before we were adding a record to the cache for each device from the set which is incorrect. For example, if someone requests a UR program for {dev2, dev3} from the cache then it is expected that this UR progam must be usable to submit a kernel to dev3. But we could get a program for {dev1, dev2} from the cache which is unusable on dev3.
againull
force-pushed
the
sycl_bundle_multi_device
branch
from
October 1, 2024 08:50
687d688
to
cce501a
Compare
cperkinsintel
approved these changes
Oct 16, 2024
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.
This PR includes:
Changes in the program manager methods to be able to properly create/build UR program for multiple devices. So far, we were mostly using only the first device in the vector to create/build UR program which made UR program unusable on other devices.
UR tag update brings the version of urProgramCreateWithBinary which allows
to create UR program from multiple device binaries.
Our program cache key allowed only a single device. I have changed it to contain a set of devices. If UR program is created and built for a set of devices then the same UR program is usable whenver we have any subset of this set. That's why if we have a program built for a set of devices then add all subsets to the cache. Before we were adding a record to the cache for each device from the set which is incorrect. For example, if someone requests a UR program for {dev2, dev3} from the cache then it is expected that this UR progam must be usable to submit a kernel to dev3. But we could get a program for {dev1, dev2} from the cache which is unusable on dev3.