Files
Enrico Degregori eb033ef208 [rocm-libraries] ROCm/rocm-libraries#4964 (commit 3271d9a)
[CK Tile] Eight Waves pipeline GEMM

## Motivation

Eight waves pipeline was added for ABQuant. The goal of this PR is to
enable it also for GEMM

## Technical Details

Summary:
 - Block:
- Create block struct for GEMM using eight warps specific distribution
encodings
   - Use this block struct in ABQuant for encodings
 - Pipeline:
- Create impl pipeline for eight waves which can be used by GEMM and
ABQuant as base (and for AQuant and BQuant in the future)
- Create eight waves pipeline for GEMM (this can not be easily
integrated in the existing async pipeline)
 - Pipeline policy:
- Extract GEMM specific parts in the ABQuant policy to define GEMM
policy (then ABQuant use it as base and add Quant specific methods)
- Minor: naming was inconsistent between warp/wave, everything is now
referred to as eight waves

So overall we have:
- block struct directly used by GEMM -> ABQuant derived struct to
implement operator
- Impl base pipeline with general implementation -> GEMM and ABQuant
pipelines use it to avoid code duplication but still define their own
pipelines
- pipeline policy struct directly used by GEMM -> ABQuant derived policy
struct for Quant specific parts

## Test Plan

Added new tests for GEMM pipeline:
`test_ck_tile_gemm_pipeline_comp_async_eight_waves` (only gfx950
supports it).

Note: K padding test is disabled for this pipeline because it's not
implemented yet

## Submission Checklist

- [x] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
2026-03-16 08:31:56 +00:00

92 lines
6.2 KiB
C++

// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
// SPDX-License-Identifier: MIT
#pragma once
#include "ck_tile/ops/gemm/block/block_gemm_areg_bgmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bgmem_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_eight_waves_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v2.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_breg_creg_v2_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_one_warp_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v2.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v2_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v2_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_areg_bsmem_creg_v2r1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_breg_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_breg_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_breg_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_bsmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_bsmem_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_bsmem_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_mx_areg_bsmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_mx_areg_bsmem_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_problem.hpp"
#include "ck_tile/ops/gemm/block/block_universal_gemm_as_bs_cr.hpp"
#include "ck_tile/ops/gemm/block/block_wp_asmem_breg_creg.hpp"
#include "ck_tile/ops/gemm/block/block_wp_asmem_bsmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/block/block_wp_asmem_bsmem_creg_v1_custom_policy.hpp"
#include "ck_tile/ops/gemm/kernel/batched_gemm_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/gemm_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/gemm_multi_abd_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/gemm_multi_d_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/gemm_tile_partitioner.hpp"
#include "ck_tile/ops/gemm/kernel/grouped_gemm_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/streamk_gemm/streamk_gemm_coherency.hpp"
#include "ck_tile/ops/gemm/kernel/streamk_gemm/streamk_gemm_kernel.hpp"
#include "ck_tile/ops/gemm/kernel/streamk_gemm/streamk_gemm_tile_partitioner.hpp"
#include "ck_tile/ops/gemm/kernel/streamk_gemm/streamk_gemm_tile_partitioner_impl.hpp"
#include "ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_base.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_async.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_async_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_async_eight_waves.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_async_eight_waves_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v3.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v4_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v5.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v5_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v6.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_comp_v6_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_eight_waves_base.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_mem.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_scheduler.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_async_v1.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v1.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v2.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_agmem_bgmem_creg_v2_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_problem.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipelines.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_universal_pipeline_ag_bg_cr_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/tile_gemm_shape.hpp"
#include "ck_tile/ops/gemm/pipeline/tile_gemm_traits.hpp"
#include "ck_tile/ops/gemm/pipeline/wp_pipeline_agmem_bgmem_creg_base_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/wp_pipeline_agmem_bgmem_creg_v2.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_mfma_impl.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_smfmac.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_smfmac_impl.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_wmma.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_wmma_impl.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_wmma_impl_16bit_traits.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_wmma_impl_8bit_traits.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_attribute_wmma_impl_base_traits.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_dispatcher.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_impl.hpp"
#include "ck_tile/ops/gemm/warp/warp_gemm_smfmac_impl.hpp"
#include "ck_tile/ops/gemm/warp/warp_wmma_gemm.hpp"
#include "ck_tile/ops/gemm/warp/warp_wmma_gemm_gfx11_utils.hpp"
#include "ck_tile/ops/common/generic_2d_block_shape.hpp"
#include "ck_tile/ops/common/load_and_convert_tile.hpp"
#include "ck_tile/ops/common/streamk_common.hpp"
#include "ck_tile/ops/common/tensor_layout.hpp"
#include "ck_tile/ops/common/utils.hpp"