mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-20 12:59:49 +00:00
393ebc1a50
Backup commit grouping all in-progress local work so nothing is lost: - Modified CK-UA kernel + example sources (unified_attention.cpp, unified_attention_kernel.hpp) and CMake/build files. - Updated dispatcher README and ctypes_utils.py. - New unified_attention example notes: PARAMETERS.md, VARIABLES.md. - New unified_attention instances for d128 fp16/bf16 (mask/nmask, gqa6). - New 99_toy_tutorial/ collection: bank-conflict investigations (test_*.cpp, *.js, *.gdb, *.asm, *.md), tile distribution / row reduction / calling_gemm / thread_buffer tutorials. - Slide decks and supporting assets (bank_conflict_slides.qmd/.html, tile_distribution_slides.qmd, assets/, *_files/, step1_reshape_only, xor_full_steps_simple). - GDB helper script (break_on_ds_read.gdb). Not intended for upstream review; pure WIP snapshot.
39 lines
1.1 KiB
CMake
39 lines
1.1 KiB
CMake
# Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
|
# SPDX-License-Identifier: MIT
|
|
|
|
include_directories(AFTER
|
|
${CMAKE_CURRENT_LIST_DIR}
|
|
)
|
|
|
|
add_subdirectory(01_fmha)
|
|
add_subdirectory(02_layernorm2d)
|
|
add_subdirectory(03_gemm)
|
|
add_subdirectory(04_img2col)
|
|
add_subdirectory(05_reduce)
|
|
add_subdirectory(06_permute)
|
|
add_subdirectory(09_topk_softmax)
|
|
add_subdirectory(10_rmsnorm2d)
|
|
add_subdirectory(11_add_rmsnorm2d_rdquant)
|
|
add_subdirectory(12_smoothquant)
|
|
add_subdirectory(13_moe_sorting)
|
|
add_subdirectory(14_moe_smoothquant)
|
|
add_subdirectory(15_fused_moe)
|
|
add_subdirectory(16_batched_gemm)
|
|
add_subdirectory(17_grouped_gemm)
|
|
add_subdirectory(18_flatmm)
|
|
add_subdirectory(19_gemm_multi_d)
|
|
add_subdirectory(20_grouped_convolution)
|
|
add_subdirectory(21_elementwise)
|
|
add_subdirectory(22_gemm_multi_abd)
|
|
add_subdirectory(35_batched_transpose)
|
|
add_subdirectory(36_pooling)
|
|
add_subdirectory(38_block_scale_gemm)
|
|
add_subdirectory(40_streamk_gemm)
|
|
add_subdirectory(41_batched_contraction)
|
|
add_subdirectory(42_mx_gemm)
|
|
add_subdirectory(42_unified_attention)
|
|
add_subdirectory(50_sparse_attn)
|
|
add_subdirectory(51_tile_distr_enc_reg_map)
|
|
add_subdirectory(99_toy_tutorial)
|
|
|