Add Composable Kernel examples

[j-stephan]

Motivation

This PR adds Composable Kernel's ck_tile examples to this repository.

Technical Details

This PR only targets ROCm + Linux; Windows and CUDA are not supported by Composable Kernel.

Test Plan

Test Result

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[j-stephan]

The failing markdown linter will be resolved once ROCm/rocm-docs-core#1449 is merged.

[j-stephan]

This PR requires ROCm 7.1. Once #341 is merged the build errors should disappear.

[Copilot]

Pull Request Overview

This PR adds Composable Kernel's ck_tile examples to the ROCm Examples repository, focusing exclusively on ROCm and Linux platforms (CUDA and Windows are not supported). The examples demonstrate various GPU operations using CK Tile's programming model, including GEMM operations, convolutions, and basic tensor operations.

Key Changes

• Added comprehensive examples for GEMM operations (batched, block-scale, flatmm, multi-d, grouped)
• Introduced grouped convolution examples (forward and backward weight)
• Implemented basic operations (elementwise, reduce, permute, img2col)
• Provided build infrastructure through CMake and Makefiles with architecture-specific support checks

Reviewed Changes

Copilot reviewed 111 out of 281 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
Libraries/ComposableKernel/gemm/flatmm/flatmm_basic.cpp	Implements FLATMM GEMM kernel with tile partitioning and pipeline configuration
Libraries/ComposableKernel/gemm/block_scale_gemm/gemm_aquant_basic.cpp	Implements block-scale quantized GEMM with group quantization support
Libraries/ComposableKernel/gemm/batched_gemm/batched_gemm.cpp	Implements batched GEMM operations with configurable pipeline strategies
Libraries/ComposableKernel/convolution/grouped_convolution/grouped_convolution_forward.cpp	Implements grouped convolution forward pass
Libraries/ComposableKernel/basic/reduce/reduce.cpp	Demonstrates 2D reduction operations with block tiling
Libraries/ComposableKernel/basic/permute/permute.cpp	Generic tensor permutation with matrix-core optimized alternative
CMakeLists.txt and Makefile files	Build configuration with architecture checks for gfx908/gfx90a/gfx942/gfx950

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[zichguan-amd]

LGTM

[vidyasagar-amd]

Thanks for the addition

[adeljo-amd]

LGTM