derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
20 lines
1.8 KiB
ReStructuredText
20 lines
1.8 KiB
ReStructuredText
.. meta::
|
|
:description: rocThrust documentation and API reference
|
|
:keywords: rocThrust, ROCm, API, reference, execution policy
|
|
|
|
.. _hip-execution-policies:
|
|
|
|
*********************************************************
|
|
Avoiding synchronization barriers
|
|
*********************************************************
|
|
|
|
The ``hip_rocprim::par_nosync`` execution policy provides a way to avoid synchronization barriers when running algorithms.
|
|
|
|
``hip_rocprim::par_nosync`` and ``hip_rocprim::par`` are both parallel non-deterministic policies. ``hip_rocprim::par`` is :ref:`synchronous and blocking with respect to the host <synchronization-and-blocking>`. Under ``hip_rocprim::par``, algorithms are launched in parallel on the device, but the host blocks on each algorithm. The next algorithm won't be launched until each algorithm finishes.
|
|
|
|
The ``hip_rocprim::par_nosync`` policy can be used to avoid this synchronization barrier. Synchronization can be skipped when possible under the ``hip_rocprim::par_nosync`` policy. Under this policy, the host has the possibility of not blocking on the algorithms running on the GPU. The CPU can then perform other work while waiting for the GPU to finish running the algorithms. The host and device should be explicitly synchronized before accessing results.
|
|
|
|
.. note::
|
|
|
|
rocThrust doesn't support `hipGraphs <https://rocm.docs.amd.com/projects/HIP/en/latest/how-to/hip_runtime_api/hipgraph.html>`_. The operations inside a hipGraph must be asynchronous and the rocThrust API is synchronous by default. While there are asynchronous versions of the algorithms in the ``thrust::async`` namespace, these algorithms operate asynchronously by returning futures, which is different from the form of asynchronous execution required within hipGraphs. It is currently impossible to guarantee that synchronization doesn't occur within any rocThrust algorithm.
|