-
Notifications
You must be signed in to change notification settings - Fork 787
[UR][L0] Memory interop support given external buffer #17458
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
[UR][L0] Memory interop support given external buffer #17458
Conversation
e8a3e8c
to
5006867
Compare
Enable L0 to directly import the memory allocated for NPU and let SYCL kernel running on iGPU be able to directly access the data it will reduce the overhead of data movement between NPU and iGPU. Signed-off-by: Zhang, Winston <[email protected]>
77ec555
to
536f165
Compare
Signed-off-by: Zhang, Winston <[email protected]>
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Updates look good to me!
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This shouldn't be specific to the NPU, right? It should enable more general memory sharing.
Co-authored-by: Peter Žužek <[email protected]>
Correct, and this would work for iGPU as well - as requested in the ticket. |
@intel/llvm-gatekeepers Ready for merge. |
@intel/llvm-gatekeepers please merge this change. |
Enable L0 to directly import the memory allocated for NPU/iGPU and let SYCL kernel running on iGPU be able to directly access the data it will reduce the overhead of data movement between NPU/iGPU and iGPU. --------- Signed-off-by: Zhang, Winston <[email protected]> Co-authored-by: Peter Žužek <[email protected]>
Enable L0 to directly import the memory allocated for NPU/iGPU and let SYCL kernel running on iGPU be able to directly access the data it will reduce the overhead of data movement between NPU/iGPU and iGPU. --------- Signed-off-by: Zhang, Winston <[email protected]> Co-authored-by: Peter Žužek <[email protected]>
Cherry-pick commits that reached the internal branch between intel/llvm cutoff and release branch pulldown. Patches included: --- [Coverity][UR] Remove noexcept keywords from pool_manager methods (#17678) This should fix coverity issues in `ur_pool_manager.hpp`, `ur_util.hpp` and `ur_adapter_registry.hpp` files reported in: https://coverity.devtools.intel.com/prod6/#/project-view/26752/10020 Patch-by: Krzysztof Święcicki <[email protected]> --- [UR][L0] Enable support for using L0 spec External Semaphores (#17671) - Implements the spec version of the L0 Extension External Semaphores - Implements handling the translation of handles if the driver only supports the EXP functions. Patch-by: Neil R. Spruit <[email protected]> --- [UR][L0] Memory interop support given external buffer (#17458) Enable L0 to directly import the memory allocated for NPU/iGPU and let SYCL kernel running on iGPU be able to directly access the data it will reduce the overhead of data movement between NPU/iGPU and iGPU. Patch-by: Zhang, Winston <[email protected]> Co-authored-by: Peter Žužek <[email protected]> --- [UR] Fix Loader Negative Prefilters (#17528) This commit fixes #17086. Patch-by: Isaac Ault <[email protected]> --- [UR] Remove unnecessary null pointer check in OpenCL command buffer code, and update declaration of filter lamba to avoid copies (#17615) Patch-by: Martin Grant <[email protected]>
Enable L0 to directly import the memory allocated for NPU/iGPU and let SYCL kernel running on iGPU be able to directly access the data it will reduce the overhead of data movement between NPU/iGPU and iGPU.