9b21a13119
Pipeline: 1. extract.py: Parses all 26 CCCL tuning_*.cuh → 26 YAML schemas in muh/schema/ 2. parse.py: .muh file parser with extends-inheritance + schema validation 3. gen_yaml.py: .muh → computility-run.yaml (verified: matches competition reference) 4. gen_patch.py: .muh → vllm kernel unified diff patches (6 algorithm mappings) 5. baseline.muh: Competition reference config, all tuning values pending BI-V100 benchmarks Schemas extracted: 26 algorithms, 8-19 params each, SM75/80/90/100 reference tunings Priority mapping: reduce→attention, topk→sampling, scan→paged_attention, transform→activations, batch_memcpy→KV_cache, for→RoPE Tested: extract→parse→validate→gen_yaml→gen_patch full pipeline passes
33 lines
572 B
YAML
33 lines
572 B
YAML
# muh schema index — all extracted CCCL tuning algorithms
|
|
|
|
algorithms:
|
|
- adjacent_difference
|
|
- batch_memcpy
|
|
- batched_topk
|
|
- find
|
|
- find_bound_sorted_values
|
|
- for
|
|
- histogram
|
|
- merge
|
|
- merge_sort
|
|
- radix_sort
|
|
- reduce
|
|
- reduce_by_key
|
|
- rle_encode
|
|
- rle_non_trivial_runs
|
|
- scan
|
|
- scan_by_key
|
|
- segmented_radix_sort
|
|
- segmented_reduce
|
|
- segmented_scan
|
|
- segmented_sort
|
|
- select_if
|
|
- three_way_partition
|
|
- topk
|
|
- transform
|
|
- transform_tile
|
|
- unique_by_key
|
|
|
|
total: 26
|
|
source: cccl_upstream/cub/cub/device/dispatch/tuning/
|