composable_kernel

composable_kernel copied to clipboard

## Proposed changes Cleaned-up test code for the forward convolution instance creation tests for the CK Builder. Addressed some leftover comments from PR #3138. Main changes - Refactored tests such...

vpietila-amd

## Proposed changes This PR adds factory support for the remaining forward convolution device operations in CK Builder. **Jira Ticket:** https://amd-hub.atlassian.net/jira/software/c/projects/ALMIOPEN/boards/319/backlog?selectedIssue=ALMIOPEN-350 **Task 350** - Add remaining forward convolution device operations:...

JH-Leon-KIM-AMD

## Proposed changes 1. Allow specifying a replacement partition index for the `tile_window_with_static_distribution` template, with the resulting offset applied as an immediate offset in the instructions. 2. Allow specifying an...

poyenc

Add CK Tile Tutorials Folder with GEMM and COPY Kernel

1

## Proposed changes In this PR we bring a simple GEMM example demonstrating key CK Tile concepts such as - Problem & Policy struct - Shape & Kernel struct -...

AviralGoelAMD

## Proposed changes This adds a series of documentation on CK Tile internals. ## Checklist Please put an `x` into the boxes that apply. You can also fill these out...

vidyasagar-amd

## Proposed changes Implementation of multi reduce ops, both in a threadwise or multiblock (aka blockwise) fashion. It migrate most of the feature already present in old CK with a...

damien-lejeune

## Proposed changes Summary: - Add support for `AK1 != BK1` - Add support for `AK1, BK1 > 8` - Introduce KInner template parameter for pipelines when loading multiple tiles...

EnricoDeg

Remove "basic" and universal GEMM tests, and incorporate their test cases into the GEMM pipeline tests

1

## Proposed changes The "basic" GEMM tests do not test the basic GEMM (pipelines V1 and V2), but rather the V3 pipeline, which is tested by the universal GEMM tests....

SamiAario-AMD

## Proposed changes Extends ck-tile batched contraction kernel to support arbitrary multi-dimensional non-contiguous tensor layouts using descriptors. Extends example to cover testing this new added feature, user can pass any...

msaffari-amd

In CK Tile, DataTypeTraits is used to print out the datatype name when given a datatype in examples and test files. ``` template struct DataTypeTraits; template struct DataTypeTraits { static...

arai713