GitLab

* [CK Tile] Batched GEMM Example

* [CK Tile] Batched GEMM Example - minor refactor

* [CK Tile] Batched GEMM Example - README update

* [CK Tile] Batched Gemm Example - review changes

- Added tensor data layours as input parameters
- Changed structure of Host and Kernel args
- Removed bug with invalid vector read on non-contiguous memory

* [CK Tile] Batched Gemm Example - remove comment

* [CK Tile] Batched Gemm Example - Add GTests part1

* [CK Tile] Batched Gemm Example - GTests part2 + review changes

* [CK TILE] Batched GEMM post merge fixes

* [CK Tile] Batched GEMM Example - fix pad views

* [CK TILE] Add gemm compute pipeline v3

* Enable universal gemm compute pipeline.

* Rename example and add compute pipeline.

* Introduce ag bg cr pipeline impl base.

* Refactor to reuse code.

* Cleaning

* Formatting.

---------
Co-authored-by: Adam Osewski <19374865+aosewski@users.noreply.github.com>
Co-authored-by: Adam Osewski <Adam.Osewski@amd.com>

* add interwave scheduler for gemm mem pipeline

* Fix merge artifacts.

* Refactor unit tests.

* Switch to interwave scheduler for mem example

---------
Co-authored-by: Adam Osewski <19374865+aosewski@users.noreply.github.com>
Co-authored-by: Adam Osewski <Adam.Osewski@amd.com>

* Fix cmake example build

* Support max3 in smoothquant one pass

* support max3 in two pass

* support max3 in add_rmsnorm_rdquant

* Allow getting batch size from splitkv tile partitioner

* Fix wrong paged-kvcache impl for group mode

* Fix wrong example code for page-kvcache

* Undo changes in fmha_fwd.cpp

* Always use 2D block table

* Add is_gappy kernel argument for paged-kvcache

The is_gappy argument is used for differentiating seqstart_k_ptr usage
in flash-attention & xformers

* Remove out-of-date comments

* Remove no-longer used method

* Fix wrong # page-block calculation

* Fix wrong comment

---------
Co-authored-by: Qianfeng <qianfeng.zhang@amd.com>

* Block universal gemm.

* Universal block gemm with interwave scheduler - draft.

* Refactoring

* Move a/b_warp_tiles into BlockGemmImpl
* set BlockGemmImpl as a class member

* Change tile size for more suitable to memory bound cases.

* Introduce kKPerThread to WarpGemm

* Add documentation comment.

* Fix Interwave scheduler block gemm.

* Add compute/memory friendly tile configuration.

* Clean

* New tile configurations in gemm mem example.

* Add more static checks and fix loop order in block gemm.

* Add more static checks and use warp gemm mfma dispatcher.

* Add default scheduler block gemm.

* Remove logging in example.

* moe pipeline

* update code

* compile OK

* update

* update cpu reference

* update pipeline_gemm0

* compiler ok

* update pipeline

* rename to ex pipeline

* block-asm

* update

* update

* update first gemm ok

* compute correct

* update file structure

* update README

* update

* update

* update code

* update API

* return unsupport case

* add comment

* update readme

* update

* uncomment

* update

* fix build err

---------
Co-authored-by: valarLip <340077269@qq.com>

* Fix mis-matched tuple<> elem types

* Rename MakeKargs() as MakeKargsImpl()

---------
Co-authored-by: Qianfeng <qianfeng.zhang@amd.com>

* update MOCK_ID for moe-sorting

* add moe-smoothquant

* update a comment

* fix format

* hot fix

* update topk in overflow case

* update comments

* update bf16 cvt

---------
Co-authored-by: valarLip <340077269@qq.com>

* Change in fwd-splitkv kernel to support num_splits=1 case

* Update in codegen fwd-splitkv to make num_splits > 1 cases pass

* Specify instance traits in dispatch

* Fix link error for fp8 kernels

---------
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* Generate group mode paged-attn kernel

* Enable paged-kvcache + group mode support

* Add missing header: fused_moe.hpp

* Add comment to explain kernel arg usage

* Make error message more clear

* Add comment for confusing data member names

* Add more comment for confusing variable names

* Fix typo in option description

* Finished the feature

* Modified the test file

* Test case update

* addresss comment

* Addressed the review comment

* Fixed the CI error

* [CK_TILE] add more stride for layernorm to support un-continuous Tensor

* align CK coding style

* extend strides to layernrom expample

* clang-format...

* add moe_sorting & check ok

* fix comments & typo

* Run remod.py under include/ck_tile & example/ck_tile directories

* format codes

* fix output ci check bug

* fix moe sorting readme and error commit file

* use magiv div to accelerate compute

* add an loop unroll for moe lds ops

* add extblocksnel to set zeros for moebufs

* [Ck_tile] moe set zero run ok, add size check and fix ref check

* [Ck_tile]fix moe_sorting fuse set_zero remod

* [Ck_tile] change name style, fix zero buffer size err, change folder

* [Ck_tile] moe_sorting: fix name style

* [Ck_tile] moe_sorting, remove useless params in traits

* [Ck_tile] change outputtile cnt * unit_size; change output buf alloc

---------
Co-authored-by: dummycoderfe <noplydummmycoder@163.com>
Co-authored-by: Po Yen, Chen <PoYen.Chen@amd.com>
Co-authored-by: carlushuang <carlus.huang@amd.com>

* optimze small N case using vec io and using rcp div

* [Ck_tile] layernorm, add param to control fastdiv; change generate codes and test pass

* [Ck_tile] fix blockSize compute in Generic2dBlockShape

* [Ck_tile]fix kfastfdiv template style

* [Ck_tile] layernorm, fix stype in review

---------
Co-authored-by: dummycoderfe <noplydummmycoder@163.com>

Before, generate.py appended the list at the end of the output file.
When running the cmake configuration steps multiple times on the
examples, the blob list (such as fwd_blob_list.txt) would grow at every
configuration.
`library/src/tensor_operation_instance/gpu/mha/CMakeLists.txt` worked around
this issue by removing the output file if it exists.

Now, generate.py overrides the content of the output file.
There is no need for the workaround in the CMakeLists.txt;
and the issue is solved for the example projects too.

* more accurate residual

* modify comment

* Fix literal case in README.md

---------
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* fix compile error

* fix typo of padding

* Add smoothquant op

* Add smoothquant instance library

* refine type

* add test script

* Re-generate smoothquant.hpp

* Always use 'current year' in copyright

* use Generic2dBlockShape instead

* Add vector = 8 instance back

* Find exe path automatically

* Simplify the api condition

* Remove debugging code

* update year

* Add blank line between function declaration

* explicitly cast return value to dim3

* refine return value

* Fix default warmup and repeat value

* Add comment

* refactor sommthquant cmake

* Add README

* Fix typo

---------
Co-authored-by: Po Yen, Chen <PoYen.Chen@amd.com>

* hot fix ln

* some rename

* add prenorm/postnorm support, refactor using generate.py

* update README

* update README

* fix format

* update some description and fix format

* update format

* format

* use non-raw for loading

* format and update n4096

* dynamic-quant ready

* update readme

* support fused dynamic-quant

* update fused-quant, with smooth

* update README

* update args

* update some based on comment

* CK-Tile GEMM with memory bound pipeline.

* Memory bound gemm pipeline.

* Fix not closed namespace.

* Block gemm mem pipeline draft.

* Do not use ck_tile:: within ck_tile namespace.

* Refactoring & Move Layout info to pipeline problem.

* Get hot loop and TailNum information before lunching kernel.

* Fixes in pipeline.

* Add comment to load_tile_raw and change variable naming style.

* Few small changes & formatting.

* Do not use macro.

* Add gtests.

* Use AccDataType for Output of MFMA instruction.

* Formatting.

* Refactor gemm examples.

* Switch over to current block gemm.

* Use currently available pipeline policy.

* Refactoring and review comment.s

* Fixes after merge.

* Add missing include.

* Add load tile overload which accepts output tensor as parameter.

* This give 8% perf boost at the cost of using more registers.

* Rename example.

* Small changes.

* Fix compilation err and lower K.

* Support different layouts for A/B

* Fix vector size for different layouts.

* Rename Alignment into VectorSize

* Unblock tests.

* Add reduce2d new api

* Prevent user use cross warp reduction

* Fix bug of std caculation

* Add rmsnorm2d

* Add rmsnorm small example

* Remove static assert to prevent compile fail

* Add script to test performance and correctness

* Add missing cmake change

* refine naming

* refine example of rmsnorm

* Fix bug of rmsnorm

* Refine naming

* Fix cmake

* clang format

* Refine pipeline name

* Add add_rmsnorm2d_rdquant kernel

* Add reduce op

* host verification

* Fix bug of one pass pipeline

* Refine tile size

* Add two pass pipeline

* Rename two pass to three pass

* Fix bug of kSaveX == false

* Add instance library

* Add test script

* Fix bug of x verification

* Add save_x to trait

* Add README

* Move reduce2d into reduce folder

* Fix bug of welford when number of m warp > 1

* remove reduncant comment

* 1. move 06_rmsnorm2d to 10_rmsnorm2d
2. move 07_add_rmsnorm2d_rdquant to 11_add_rmsnorm2d_rdquant

* clang format and add missing header

* Add host validation of add + layernorm2d + rsquant

* Revert "Add host validation of add + layernorm2d + rsquant"

This reverts commit 936cb457978b928b90eff89a08fcdb7dc8bbed67.

* Remove deprecated flag

* Add ceil_to_qualified_tile_length()

* Rename kK0BlockLength to kQKHeaddim

* Add kSubQKHeaddim concept to support headdim96

* Fix in math.hpp to avoid using __half interfaces

* Add LdsBufferSequence instance for headdim96

* Update in fmha_fwd/fmha_fwd_splitkv codegen to support hd96 testing

* Disable hd96 instance generation in codegen fmha_fwd and fmha_fwd_splitkv to save compiling time

* Reformat one file

* Fix text alignment in fmha_fwd_splitkv.py

---------
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* topk_softmax

* remove some file

* fix atomix linear_offset

* address various comment, and change sfc get_index api to static(tuple)

* Use pre-defined constants for readability

* Use vector write for o_acc tensor

* Remove no-longer used policy method

* Deprecate no-longer used policy/pipeline

* Specify gemm0/gemm1 block warps separately in codegen

* Fix wrong ps_idx creation logic

* Add single-warp block gemm

* Supoprt single-warp gemm0

* Make MakeCBlockTile() as static method

* Use MakeCBlockTile() to get underlying tile distribution

* Use kNumGemm1Warps to compute # threads for gemm1

* Put normal case in the if clause

* Refine fmha splitkv block mapping

* Refine & fix the lse_acc/o_acc layout

* Fix wrong LDS size for K tile

* Use kK0=64 for hdim=128,256 fmha splitkv kernels

* Use kK1=64 for hdim=32,64,128 fmha splitkv kernels

* Undo kK0/kK1 changes

* Use more reasonable GetAlignmentV() computation

* Using store_tile() in fmha splitkv kernel epilogue

* port layernorm

* change warp_welford.hpp

* Update warpshuffle

* 1. Add save mean and save std back
2. Move construction of tensor_view and tile_window to operator()

* refine welford max count calculation

* unify layernorm api

* Rename file

* Remove save mean and inv std

* Revert "refine welford max count calculation"

This reverts commit 02236580

.

* Fix order of parameter

* refine welford max count calculation again

* Remove fp32 instances

* Fix bug of padding

* refactor api

* Support bf16

* Extract common function

* Refine arg of operator()

* Add kMThreadPerBlock to template parameter

* clang format

* Refine variable name

* Refine file name

* remove redundant line

* refactor layernorm2d pipeline and add block-per-block utility

* fix name

* rename more

* add more block-per-tile instance

* remove duplicated define

* update instance for 2048, 1024 case

* support up to 2048 now

* opt loading

* add n1536

* Add two pass pipeline

* format

* Fix incorrect type

* parallel compilation

* Use smaller N

* fix 2p pass

* Support Repeat_M in distribution

* Refine nameing

* Add reduce example

---------
Co-authored-by: letaoqin <letaoqin@amd.com>
Co-authored-by: aska-0096 <haocwang@amd.com>
Co-authored-by: rocking <ChunYu.Lai@amd.com>
Co-authored-by: carlushuang <carlus.huang@amd.com>

* Use smaller width for lse_accum dist tensor

* Update pipeline comment

* Fix wrong distribution for lse_accum

* Remove duplicate dim in lse_accum dist encoding

* Decide fmha splitkv combine kernel kBlockSize by kM0

* Remove assumption of MPerThread=1

* Add log<4> & log<8> specialization

* Enlarge occupancy array

* Fix vector size for small tile

* Add support for kMaxSplits=8

* Re-format gemm.hpp

* Use 16x16x16 warp gemm for fwd_splitkv

* Centralize policy code changes

* Leave fp8/bf8 tile settings unchanged

* [CK_TILE] Add block universal gemm pipeline policy

* Fixes

* fixes2

* Fixes3

* fixeS

* ake the cshuffle compilable

* modify Mhe reference on gpu and cpu. Correaccess of cshuffle

* fix the cpu reference code

* Complete the in tile shuffle logic

* restructure the kernel template input

* change the naming pattern of ck_tile gemm pipeline

* Re-format files using remod.py

* Solve the fmha conflict with gemm

* Comment Addressed from Carlus

---------
Co-authored-by: Po Yen, Chen <PoYen.Chen@amd.com>

* Fix text alignment of ArgParser::print()

* Update example README files

* Clarify make-ck-dev.sh <arch> usage

* Only keep some of the argument from '-?' output

* Undo command line output changes in README

* Only keep existing argument on doc and update description

* Fix text alignment

* Make cmake-ck-*.sh compatible with 'sh' command

* Simplify the codes in splitkv_combine pipeline

* Always set kPadSeqLenK=true for fmha splitkv kernels

* Change in Oacc Alignment and TileDistribution to be more adaptable to tile sizes

---------
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* Fix compile error

* Add one pass pipeline

* Extract creating tile_window to operator()

* clang format

* reduce duplicated code

* do not hardcode

* Support padding in layernorm

---------
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* Adding seed and offset pointer support to the philox random number generator.

* Separating seed and offset pointer checks with different condition statements.

* Changes include, adding support for device seed and offset pointers, union is used to store seed/offset values and device pointers to minimize device SGPRs.

* Correcting a typo in the readme file

* Re-format files using remod.py

* Use STL type for API parameters

* Use simpler struct design for drop_seed & drop_offset

* Undo unnecessary changes

* Sync kargs style for fmha_fwd.hpp/.cpp

* Use templated union to reduce code

* Use structured binding to make code more readable

---------
Co-authored-by: Sudhir Kylasa <sukylasa@amd.com>
Co-authored-by: Po Yen Chen <PoYen.Chen@amd.com>

* Use same layout for o_acc and o tensor

* Use better param names in partitioner

* Remove redundant kargs 'max_seqlen_q'

* Use better param names in splitkv kernel

* Add comment for additional kernel arguments

* Sync empty loop early return logics between pipelines

* Pass more arguments to cmake in scripts

* Align backslashes

* Fix wrong o_acc tensor view strides

* Change o_acc layout if o_perm=0

* Handle whole row masked via attn_bias

* Use use vector width = 1 for o_acc

* Use more even split sizes

* [CK_TILE] Image to Column kernel

* Fixes

* Vector loads and stores

* Fixes

* Fixes

* change test dir name

* add barriers

* tail bias barriers

* adjust bf16/hd256 tol

* continue adjust bf16/hd256 tol

* Support the N dimension padding

* Finished the padding feature for different dimension of K

* Finished the feature of gpu verification

* Add the ck_tile_gemm test in the CI CD

* add the include of tensor_layou in reference_gemm

* Comment Addressed

* split ck_tile fhma and gemm tests into separate stages

* restructure the reference gemm

* restructure a new reference_gemm api that could read the device mem

---------
Co-authored-by: carlushuang <carlus.huang@amd.com>
Co-authored-by: illsilin <Illia.Silin@amd.com>