[Fusion] [Graph]Add Matmul Allreduce Rmsnorm fusion Pass (#5034)

This PR add `MatmulAllreduceRmsnorm` operator and introduces a graph
fusion pass for `matmul_allreduce_rmsnorm` operations. The
implementation includes a new configuration flag, a pattern matching
pass using `torch._inductor.pattern_matcher`.

Co-authored-by: Trunrain [270250579@qq.com](mailto:270250579@qq.com)

- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: wxsIcey <1790571317@qq.com>
Signed-off-by: tongrunze <t00574058@china.huawei.com>

vllm_ascend/ascend_config.py

@@ -169,7 +169,9 @@ class AscendCompilationConfig:
     deployed on Ascend platforms.
     """
 
-    def __init__(self, fuse_norm_quant: bool = True, fuse_qknorm_rope: bool = False, **kwargs):
+    def __init__(
+        self, fuse_norm_quant: bool = True, fuse_qknorm_rope: bool = False, fuse_allreduce_rms: bool = False, **kwargs
+    ):
         """
         Initialize the configuration.
 
@@ -179,10 +181,13 @@ class AscendCompilationConfig:
                 Default: True
             fuse_qknorm_rope (bool): Whether to enable qknorm and rope fusion optimization.
                 Default: False
+            fuse_allreduce_rms (bool): Whether to enable allreduce and addrmsnorm fusion optimization.
+                Default: False
             **kwargs: Additional optional parameters for forward compatibility and configuration extension.
         """
         self.fuse_norm_quant = fuse_norm_quant
         self.fuse_qknorm_rope = HAS_TRITON or fuse_qknorm_rope
+        self.fuse_allreduce_rms = fuse_allreduce_rms
 
 
 class XliteGraphConfig: