📊 feat: distinguish compute density and communication density in DensityObserver

- Add record_comm_density() call in select_blocks to track CPU block selection
- Add get_per_layer_comm_density() method for detailed analysis
- Update print_summary() to show both densities and H2D savings ratio
- Set DensityObserver mode (offload/gpu_only) in test_ruler.py
- Update get_summary() to return both density types

Key insight: Comm density can be 100% even when compute density is ~37%
because sparse BSA blocks are distributed across all CPU blocks.
Since CPU block granularity is 32x coarser (4096 vs 128 tokens),
any() aggregation across heads/Q-blocks results in all CPU blocks being needed.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

tests/test_ruler.py

@@ -386,8 +386,11 @@ def run_ruler_benchmark(
     if sparse_policy and sparse_policy.upper() == "XATTN_BSA":
         DensityObserver.enable()
         DensityObserver.complete_reset()
+        # Set mode for correct density interpretation
+        DensityObserver.set_mode("offload" if enable_cpu_offload else "gpu_only")
         if not json_output:
-            print("[DensityObserver] Enabled for XAttention BSA")
+            mode_str = "offload" if enable_cpu_offload else "gpu_only"
+            print(f"[DensityObserver] Enabled for XAttention BSA (mode: {mode_str})")
 
     # LLM initialization kwargs
     llm_kwargs = {