Skip to content

Update intro.rst #662

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.

Sign up for GitHub

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

Closed
wants to merge 1 commit into from
Closed

Update intro.rst #662

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
27 changes: 12 additions & 15 deletions docs/docfiles/intro.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,17 +1,14 @@
Welcome to Data-parallel Control (dpctl)'s documentation!
=========================================================
Data-Parallel Control - The Library That Controls XPUs
=======================================================

The data-parallel control (dpctl) library provides C and Python bindings for
`SYCL 2020 <https://www.khronos.org/registry/SYCL/specs/sycl-2020/html/sycl-2020.html>`_.
The SYCL 2020 features supported by dpctl are limited to those included by
Intel's DPCPP compiler and specifically cover the SYCL runtime classes described
in `Section 4.6 <https://www.khronos.org/registry/SYCL/specs/sycl-2020/html/sycl-2020.html#_sycl_runtime_classes>`_
of the SYCL 2020 specification. Apart from the bindings for these runtime
classes, dpctl includes bindings for SYCL USM memory allocators and
deallocators. Dpctl's Python API provides classes that implement
`Python buffer protocol <https://docs.python.org/3/c-api/buffer.html>`_
using SYCL USM memory; making it possible to create Python objects that are
backed by SYCL USM memory.
Data Parallel Control ``dpctl`` is the Python library that controls multiple devices of a platform, features classes for
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this statement is misleading. dpctl does not control devices. It allows user to specify the XPU which executes user's data-parallel code (including a call to oneMKL). It is even more fine-grained than that, but articulating that takes exposing the reader to some concepts of SYCL.

unified shared memory (USM) management, and implements tensor array API on top of it. It is a foundational part of
a larger family of libraries and tools for Data Parallel Python (DPPY) aimed to program XPUs the same way as CPUs.

Dpctl also supports the DPCPP ``ONEAPI::filter_selector`` extension and has
experimental support for SYCL's ``kernel`` and ``program`` classes.
The ``dpctl`` library is built upon `SYCL standard`<https://www.khronos.org/sycl/> and implements a subset of
`runtime classes specifications`<https://www.khronos.org/registry/SYCL/specs/sycl-2020/html/sycl-2020.html#_sycl_runtime_classes>,
which allow users to query SYCL platforms, discover and represent SYCL devices, and construct SYCL queues for execution
of data-parallel code.

The library also assists authors of Python native extensions written in C, Cython, or pybind11 to access objects
representing devices, queues, memory, and tensor array APIs.