[RadixTree][3/N Refactor]：Support unified insert/evict params (#17401)

python/sglang/srt/mem_cache/common.py

@@ -7,7 +7,7 @@ import torch
 import triton
 import triton.language as tl
 
-from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache
+from sglang.srt.mem_cache.base_prefix_cache import BasePrefixCache, EvictParams
 from sglang.srt.mem_cache.memory_pool import HybridReqToTokenPool, ReqToTokenPool
 from sglang.srt.mem_cache.swa_memory_pool import SWATokenToKVPoolAllocator
 from sglang.srt.server_args import get_global_server_args
@@ -244,11 +244,13 @@ def evict_from_tree_cache(tree_cache: BasePrefixCache | None, num_tokens: int):
         if full_available_size < num_tokens or swa_available_size < num_tokens:
             full_num_tokens = max(0, num_tokens - full_available_size)
             swa_num_tokens = max(0, num_tokens - swa_available_size)
-            tree_cache.evict(full_num_tokens, swa_num_tokens)
+            tree_cache.evict(
+                EvictParams(num_tokens=full_num_tokens, swa_num_tokens=swa_num_tokens)
+            )
     else:
         # Standard allocator
         if allocator.available_size() < num_tokens:
-            tree_cache.evict(num_tokens)
+            tree_cache.evict(EvictParams(num_tokens=num_tokens))
 
 
 def alloc_paged_token_slots_extend(
@@ -311,7 +313,7 @@ def alloc_req_slots(
         if mamba_available_size < mamba_state_needed:
             if tree_cache is not None and tree_cache.supports_mamba():
                 mamba_num = max(0, mamba_state_needed - mamba_available_size)
-                tree_cache.evict_mamba(mamba_num)
+                tree_cache.evict(EvictParams(num_tokens=0, mamba_num=mamba_num))
         req_pool_indices = req_to_token_pool.alloc(num_reqs, reqs)
     else:
         req_pool_indices = req_to_token_pool.alloc(num_reqs)