[SYCL][Graph] Add support for CUDA backend #11133
Merged
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
Implements urCommandBufferAppendKernelLaunchExp function to append kernel to a command buffer. Implements urCommandBufferAppendKernelLaunchExp function to enqueue a command buffer to a cuda stream. Enable all tests supported by the current state of the Cuda backend. Sets a ScopedContext before enqueuing command buffer. * [SYCL][Graph] Adds missing exception catch
Implements CUDA backend support for: - memcpyUSM - memcpy Device to Device (1D and 2D) - memcpy Host to Device (Write 1D and 2D) - memcpy Device to Host (Read 1D and 2D) Enables all remaining e2e tests * [SYCL][Graph] Adds helper to register sync point
…kend (#309) * [SYCL][Graph] Updates design Doc to report CUDA as supported backend * [SYCL][Graph] Improves description and adds links to CUDA doc * Update sycl/doc/design/CommandGraph.md Co-authored-by: Ewan Crawford <ewan@codeplay.com> * Update sycl/doc/design/CommandGraph.md Co-authored-by: Julian Miller <julian.miller@intel.com> * Update sycl/doc/design/CommandGraph.md Co-authored-by: Julian Miller <julian.miller@intel.com> * Update sycl/doc/design/CommandGraph.md Co-authored-by: Julian Miller <julian.miller@intel.com> * Update sycl/doc/design/CommandGraph.md Co-authored-by: Julian Miller <julian.miller@intel.com> * Update sycl/doc/design/CommandGraph.md Co-authored-by: Julian Miller <julian.miller@intel.com> --------- Co-authored-by: Ewan Crawford <ewan@codeplay.com> Co-authored-by: Julian Miller <julian.miller@intel.com>
…end (#279) * [SYCL][Graph] Implement graph creation and finalization for CUDA backend Implements CommandBuffer creation, retain, release and finalize functions for the cuda backend. These functions rely on the cuda graph feature which is part of cuda runtime. Consequently, this PR adds the required linking dependencies. Fixes a commandbuffer bug in the cuda pluging initialization. Reports cuda backend as supported for the graph extension. Adds a tests that creates and finalizes an empty graph.
…ecessary e2e test
The offsets are taken into account after CUdeviceptr type casting.
This patch does not fix e2e test failure on AWS. This reverts commit cdc19b1.
…n the CUDA backend (#325)
mfrancepillois
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
mfrancepillois
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
Bensuo
force-pushed
the
command-buffer-cuda-backend
branch
from
71c681e
to
be7a830
Compare
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Bensuo
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
- Commands which enqueue to a command graph now correctly respect their dependencies by waiting on them. - Prevents issues where allocation commands with dependent copies could be delayed due to device being busy and execute in an incorrect order with regards to future command graph executions.
Bensuo
force-pushed
the
command-buffer-cuda-backend
branch
from
be7a830
to
9b984e3
Compare
This is a temporary workaround to unblock the UR merge queue. We will investigate and fix as follow-up work.
reble
reviewed
Nov 1, 2023
reble
reviewed
Nov 1, 2023
reble
reviewed
Nov 1, 2023
Bensuo
approved these changes
Nov 2, 2023
|
@intel/llvm-reviewers-runtime please review ASAP, this is blocking other UR changes. |
|
@intel/llvm-reviewers-runtime Could you give this PR a look please |
steffenlarsen
approved these changes
Nov 3, 2023
sommerlukas
pushed a commit
that referenced
this pull request
Apr 8, 2024
Enables buffer tests which were disabled on #11133 due to a failure which is no longer reproducible.
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.
Adds support for enabling CUDA backend with Sycl-Graph:
Companion PR with CUDA UR adapter changes is here: oneapi-src/unified-runtime#932
Authors
Co-authored-by: Maxime France-Pillois maxime.francepillois@codeplay.com
Co-authored-by: Ben Tracy ben.tracy@codeplay.com
Co-authored-by: Ewan Crawford ewan@codeplay.com