[MUH] Bootstrap muh toolchain — extract/parse/gen_yaml/gen_patch + baseline.muh

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

muh/schema/transform.yaml

@@ -0,0 +1,107 @@
+# muh schema for transform
+# Auto-extracted from cub/cub/device/dispatch/tuning/tuning_transform.cuh
+# Generated by muh/extract.py
+
+algorithm: transform
+source: cub/cub/device/dispatch/tuning/tuning_transform.cuh
+parameters:
+  threads_per_block:
+    type: int
+    range:
+      - 32
+      - 1024
+    step: 32
+  items_per_thread_no_input:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  min_items_per_thread:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  max_items_per_thread:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  prefetch_byte_stride:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  unroll_factor:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  items_per_thread:
+    type: int
+    range:
+      - 1
+      - 32
+    step: 1
+  vec_size:
+    type: int
+    range:
+      - 1
+      - 8
+    step: 1
+  store_vec_size:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  min_bytes_in_flight:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  copy_alignment:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  smem_size:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  tile_padding:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+  max_alignment:
+    type: int
+    range:
+      - 1
+      - 1024
+    step: 1
+    note: unknown_range
+bi_v100:
+  status: pending_benchmark
+  note: Run muh benchmark on Iluvatar BI-V100 to fill these values
+  threads_per_block: TBD
+  items_per_thread: TBD