mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-14 02:02:46 +00:00
f5beedb2e9
For pure-decode workloads (sq=1) at d=128 the m128 tile wastes most of
its 128 query rows, capping CK below Triton on every batch size in our
sweep (4..256). Add two small-Q tiers that mirror the d=64 GQA-8 ladder:
* decode_d128_mha_m16 : kBlockM=16, 1 warp, 16x16 MFMA (tiny-decode)
* decode_d128_mha_m32 : kBlockM=32, 1 warp, 32x32 MFMA (tiny-decode)
select_config now ladders by (avg_q, max_q): m16 -> m32 -> m128 -> prefill.
d=128 MHA, hq=16/hk=16, sq=1, sk=120k, num_blocks=60k:
batch before after CK BW
4 ~0.95x 0.98x 4.76 TB/s
8 ~0.85x 1.29x 5.00 TB/s
32 ~0.85x 1.14x 5.29 TB/s
64 ~0.75x 0.93x 5.35 TB/s
128 ~1.00x 1.09x 5.39 TB/s
256 ~1.03x 1.02x 5.41 TB/s
Correctness suite stays at 241/245 (same 4 known int32-overflow
failures in the prefill path).
Co-authored-by: Cursor <cursoragent@cursor.com>
CK Tile Example Suite
This directory contains a comprehensive suite of examples demonstrating the CK Tile programming model for high-performance GPU kernels. Each example illustrates a key deep learning or HPC operation, implemented using tile-based parallelism, modular pipelines, and data movement policy.
What is CK Tile?
CK Tile is a composable GPU programming API that expresses kernels as a composition of "tiles"—rectangular blocks of computation and data movement. The pipeline & policy orchestrates data movement (global <-> LDS <-> registers), computation, and synchronization, enabling high efficiency and flexibility.
Example Index
| Example | Operation | Description |
|---|---|---|
| 01_fmha | Fused Multi-Head Attention | Tile-based FMHA with masking, quantization, and epilogue fusion |
| 02_layernorm2d | LayerNorm2D | Blockwise layer normalization with fusion and quantization |
| 03_gemm | GEMM | Matrix multiplication with tilewise parallelism |
| 04_img2col | im2col | Image-to-column transformation for GEMM-based convolution |
| 05_reduce | Reduction | Tilewise sum, max, mean reductions |
| 06_permute | Permute | Generic tensor permutation (up to rank-8) |
| 09_topk_softmax | TopK-Softmax | Rowwise softmax and top-k selection for MoE gating |
| 10_rmsnorm2d | RMSNorm2D | Root mean square normalization for LLMs |
| 11_add_rmsnorm2d_rdquant | Add + RMSNorm2D + RDQuant | Fused add, RMSNorm, and rowwise dynamic quantization |
| 12_smoothquant | SmoothQuant | Per-channel scaling and quantization for int8 inference |
| 13_moe_sorting | MoE Sorting | Token-to-expert rearrangement for MoE dispatch |
| 14_moe_smoothquant | MoE-SmoothQuant | Expert-dependent quantization fused with top-k selection |
| 15_fused_moe | Fused MoE | End-to-end fused MoE block: sorting, group-GEMM, activation, weighting |
| 16_batched_gemm | Batched GEMM | Parallel computation of multiple GEMMs |
| 17_grouped_gemm | Grouped GEMM | Multiple independent GEMMs with different shapes |
| 18_flatmm | FLATMM | Flattened matrix multiplication for packed layouts |
| 19_gemm_multi_d | Multi-D GEMM | GEMM with multiple side inputs (bias, residual, etc.) |
| 35_batched_transpose | Batched Transpose | NCHW <-> NHWC and other layout conversions |
| 36_copy | Copy | Minimal example for tile-based memory movement |
| 37_transpose | Block Transpose | High-performance tiled transpose for large tensors |
Technical Highlights
- Tile Distribution: See
include/ck_tile/tile_program/tile_distribution/for mapping tiles to thread blocks. - Block Tile Pipelines: See
include/ck_tile/tile_program/block_tile_pipeline/for memory/computation pipelines. - Policies and Utilities: Many examples use custom policies for tile/block size and memory access.
How to Build & Run
mkdir build && cd build
sh ../script/cmake-ck-dev.sh ../ <arch>
make -j
Each example produces its own executable in build/bin/.
Learning and Extending
- Start Simple: Try 03_gemm or 36_copy to learn tile basics.
- Explore Fusion: See 11_add_rmsnorm2d_rdquant, 15_fused_moe, or 14_moe_smoothquant for advanced fusion.
- Experiment: Modify tile sizes, layouts, or pipelines to explore performance and flexibility.