Bug#1117561: ITP: composable-kernel -- Performance Portable Programming Model for Machine Learning Tensor Operators

[Christian Bayle <bayle@debian.org>]

Package: wnpp
Severity: wishlist
Owner: Christian Bayle <bayle@debian.org>
X-Debbugs-Cc: debian-devel@lists.debian.org

* Package name    : composable-kernel
  Version         : 7.0.1
  Upstream Contact: https://github.com/orgs/ROCm/people
* URL             : https://github.com/ROCm/composable_kernel
* License         : (MIT)
  Programming Lang: (C++, CMake, Python)
  Description     : Performance Portable Programming Model for Machine Learning
Tensor Operators

The Composable Kernel (CK) library provides a programming model for writing
performance-critical kernels for machine learning workloads across multiple
architectures (GPUs, CPUs, etc.). The CK library uses general purpose kernel
languages, such as HIP C++.

CK uses two concepts to achieve performance portability and code
maintainability:

    * A tile-based programming model
    * Algorithm complexity reduction for complex machine learning (ML)
operators. This uses an innovative technique called Tensor Coordinate
Transformation.

The current CK library is structured into four layers:

    * Templated Tile Operators
    * Templated Kernel and Invoker
    * Instantiated Kernel and Invoker
    * Client API

This package will be developped with the ROCm Team
https://wiki.debian.org/Teams/ROCm