Mirror intel/llvm commits #2796
Conversation
- Fix to urinfo to properly call release of adapters otherwise crucial teardown calls for adapters are missing. This was causing issues in UR only builds with adapters like L0 which have cleanup calls in the Adapter Release. Signed-off-by: Neil R. Spruit <[email protected]>
buffer read/write CTS tests aren't supported for the OpenCL adapter, because there is a gap in cl_khr_command_buffer that blocks implementing the required UR entry-points KhronosGroup/OpenCL-Docs#1281 is the cl_khr_command_buffer issue progressing this.
Adds support for using CUDA-Graph async alloc/free nodes in a `sycl_ext_codeplay_native_command` native-command object in a graph. This requires on CUDA 12.9 where child graphs with async alloc/free nodes can be added to a parent graph https://docs.nvidia.com/cuda/cuda-c-programming-guide/#memory-nodes-in-child-graphs
by using multiple in-order queues and round-robin strategy to dispatch work. With this approach we don't need to worry about events' lifetime. Since we are still using counter-based events, we don't need any special logic to handle cases where event release is called right after being passed as signal event or as part of a wait list.
- Adds support for the L0 extension ZE_extension_device_vector_sizes - Enables for querying one or more supported vector width preferrred and native vector widths for each supported data type. - By default, the max vector widths per datatype are reported based on the max values. - UR_L0_VECTOR_WIDTH_SIZE = <vector_width_size> is a hint to return the perferred and native sizes for a specific vector width. If it is not supported, then the defaults are returned. - If this extension is not supported, then the previous defaults are returned. Signed-off-by: Neil R. Spruit <[email protected]>
Signed-off-by: Neil R. Spruit <[email protected]>
github-actions
bot
force-pushed
the
mirror-commits-
branch
from
cb4d997 to
e6349bf
Compare
Unified Runtime -> intel/llvm Repo Move NoticeInformationThe source code of Unified Runtime has been moved to intel/llvm under the unified-runtime top-level directory, The code will be mirrored to oneapi-src/unified-runtime and the specification will continue to be hosted at oneapi-src.github.io/unified-runtime. The contribution guide will be updated with new instructions for contributing to Unified Runtime. PR MigrationAll open PRs including this one will be marked with the Should you wish to continue with your PR you will need to migrate it to intel/llvm. If your PR should remain open and not be closed automatically, you can remove the This is an automated comment. |
Automated changes by create-pull-request GitHub action