llvm
llvm copied to clipboard
intel
Intel staging area for llvm.org contribution. Home for Intel LLVM-based projects.
Buffers mapped with `UR_MAP_FLAG_WRITE_INVALIDATE_REGION` need to copy back to the buffer's device memory when unmapped. The HIP adapter was already doing this, seems like an accidental omission in the CUDA...
Cherry-pick commits that reached the internal branch between intel/llvm cutoff and release branch pulldown.
Optimizes the `enqueue()` function of sycl graphs to bypass the scheduler whenever possible and avoid creating events when not needed. * Refactors the executable graph `enqueue()` to have different paths...
We selected up to 4 work items in each work group to do detection, the whole number of selected work items no more than kThreadSlotCount.
- Using [sycl_ext_codeplay_async_memory_alloc](https://github.com/intel/llvm/pull/14800) extension (PR currently in review) - Spec wording for graph support of the feature - Usage guide guidance for library authors - Usage guide examples of explicit...
This PR fixes problems of group algorithm tests in https://github.com/intel/llvm/issues/18390#issuecomment-2867168076 SYCL :: GroupAlgorithm/exclusive_scan_sycl2020.cpp SYCL :: GroupAlgorithm/inclusive_scan_sycl2020.cpp SYCL :: GroupAlgorithm/reduce_sycl2020.cpp
- Bump compute-benchmarks git_hash - Add definition for FinalizeGraph benchmark in compute.py
sycl::range and sycl::id perform validity checks every time setting them. Use std::array instead as dimensions should already be valid. In addition, remove explicitly padding dimensions smaller than 3 and get...
