Conversation
eeezio
requested review from
Snektron,
ThomasNing,
afagaj,
andriy-ca,
aosewski,
asleepzzz,
bartekxk,
carlushuang,
cgmillette,
coderfeli,
geyyer,
illsilin,
poyenc,
qianfengz,
shumway,
tenpercent,
vidyasagar-amd and
vpietila-amd
as code owners
Contributor
There was a problem hiding this comment.
Pull request overview
This PR adds support for MX-FP6 data type in the flatmm operations, including implementing dwordx3 (12-byte) synchronous and asynchronous load/store operations needed for this data type.
Changes:
- Added pk_fp6x16_t type support throughout the codebase with corresponding type traits and conversion utilities
- Implemented dwordx3 (12-byte) load/store operations for AMD buffer addressing
- Extended flatmm pipeline to handle FP6 data type with specialized tile distributions and memory layouts
Reviewed changes
Copilot reviewed 22 out of 22 changed files in this pull request and generated 4 comments.
Show a summary per file
| File | Description |
|---|---|
| test/ck_tile/memory_copy/test_copy.hpp | Added dwordx3 copy configuration and implementation for 12-byte data transfers |
| test/ck_tile/memory_copy/test_copy.cpp | Added test cases for pk_fp6x16_t memory copy operations |
| include/ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp | Added FP6 case to MFMA warp gemm attributes |
| include/ck_tile/ops/flatmm/pipeline/mx_flatmm_pipeline_agmem_bgmem_creg_v1_policy.hpp | Extended pipeline policy to support FP6 with dwordx3 operations and specialized tile distributions |
| include/ck_tile/ops/flatmm/pipeline/mx_flatmm_pipeline_agmem_bgmem_creg_v1.hpp | Updated pipeline implementation to handle FP6 data type with appropriate memory operations |
| include/ck_tile/ops/common/utils.hpp | Added DataTypeTraits for pk_fp6x16_t |
| include/ck_tile/host/reference/reference_gemm.hpp | Added FP6 support to reference GEMM implementation |
| include/ck_tile/host/check_err.hpp | Added error checking specialization for pk_fp6x16_t |
| include/ck_tile/core/tensor/buffer_view.hpp | Added support for 12-element int8_t buffer loads/stores in LDS |
| include/ck_tile/core/numeric/vector_type.hpp | Added int32x3_tt and int32x6_tt types for FP6 vector operations |
| include/ck_tile/core/numeric/type_convert.hpp | Added include for pk_fp6.hpp |
| include/ck_tile/core/numeric/pk_fp6.hpp | New file defining pk_fp6_t type with pack/unpack operations |
| include/ck_tile/core/arch/amd_buffer_addressing_builtins.hpp | Added dwordx3 load/store builtins and extended buffer operations |
| include/ck_tile/core/arch/amd_buffer_addressing.hpp | Added dwordx3 load support |
| include/ck_tile/core.hpp | Added include for pk_fp6.hpp |
| include/ck/utility/amd_xdlops.hpp | Updated comment for FP6 MFMA operation |
| example/ck_tile/18_flatmm/mxgemm/run_mx_flatmm.inc | Added FP6-specific initialization logic |
| example/ck_tile/18_flatmm/mxgemm/mx_flatmm_instance.cpp.in | Added FP6 type alias |
| example/ck_tile/18_flatmm/mxgemm/mx_flatmm_instance.cmake | Added FP6xFP6 configuration |
| example/ck_tile/18_flatmm/mxgemm/mx_flatmm.hpp | Added MXfp6_FlatmmConfig16 configuration |
| example/ck_tile/18_flatmm/mxgemm/mx_flatmm.cpp | Added FP6 case to example runner |
| example/ck_tile/18_flatmm/CMakeLists.txt | Removed blank line |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
asleepzzz
reviewed
Jan 27, 2026
asleepzzz
reviewed
Jan 27, 2026
asleepzzz
reviewed
Jan 27, 2026
asleepzzz
reviewed
Jan 27, 2026
asleepzzz
reviewed
Jan 27, 2026
DDEle
reviewed
Jan 27, 2026
added 15 commits
February 2, 2026 02:10
asleepzzz
approved these changes
Feb 2, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Proposed changes
new feature: support dwordx3 sync/async load/store and support mx-fp6 flatmm
Checklist
Please put an
xinto the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.clang-formaton all changed filesDiscussion
If this is a relatively large or complex change, feel free to start a discussion by explaining why you chose the solution you did and what alternatives you considered