From 83bd77c9833110bca8b7e37e63e92de480200059 Mon Sep 17 00:00:00 2001
From: Shaoxu Cheng <2906339855@qq.com>
Date: Tue, 24 Mar 2026 09:00:11 +0800
Subject: [PATCH] [310p]: add rmsnorm gated fallback and unit test (#7424)

### What this PR does / why we need it?
RFC #7394
310P cannot use the fused `rmsnormgated` operator and must fall back to
the native implementation.

### Does this PR introduce _any_ user-facing change?
NO
### How was this patch tested?
ut
- vLLM version: v0.17.0
- vLLM main:
https://github.com/vllm-project/vllm/commit/4497431df654e46fb1fb5e64bf8611e762ae5d87

---------

Signed-off-by: Tflowers-0129 <2906339855@qq.com>
---
 tests/ut/_310p/ops/test_layernorm_310.py | 41 ++++++++++++++++++++++++
 vllm_ascend/_310p/ops/layernorm.py       | 12 +++++++
 vllm_ascend/utils.py                     |  7 +++-
 3 files changed, 59 insertions(+), 1 deletion(-)
 create mode 100644 tests/ut/_310p/ops/test_layernorm_310.py

diff --git a/tests/ut/_310p/ops/test_layernorm_310.py b/tests/ut/_310p/ops/test_layernorm_310.py
new file mode 100644
index 00000000..92bdf0a7
--- /dev/null
+++ b/tests/ut/_310p/ops/test_layernorm_310.py
@@ -0,0 +1,41 @@
+from unittest.mock import MagicMock, patch
+
+import pytest
+import torch
+from vllm.config import set_current_vllm_config
+from vllm.model_executor.layers.layernorm import RMSNormGated
+
+from vllm_ascend._310p.ops.layernorm import AscendRMSNormGated310
+
+
+@pytest.fixture(autouse=True)
+def default_vllm_config():
+    mock_config = MagicMock()
+    mock_config.compilation_config.custom_ops = ["all"]
+    with set_current_vllm_config(mock_config):
+        yield mock_config
+
+
+def test_rmsnorm_gated_310_forward_oot_uses_forward_native():
+    layer = AscendRMSNormGated310(hidden_size=8, eps=1e-5)
+    x = torch.randn(2, 8, dtype=torch.float32)
+    z = torch.randn(2, 8, dtype=torch.float32)
+    expected = torch.randn(2, 8, dtype=torch.float32)
+
+    with patch.object(RMSNormGated, "forward_native", autospec=True, return_value=expected) as mock_forward_native:
+        out = layer.forward_oot(x, z)
+
+    mock_forward_native.assert_called_once_with(layer, x, z)
+    assert out is expected
+
+
+def test_rmsnorm_gated_310_forward_oot_uses_forward_native_without_gate():
+    layer = AscendRMSNormGated310(hidden_size=8, eps=1e-5)
+    x = torch.randn(2, 8, dtype=torch.float32)
+    expected = torch.randn(2, 8, dtype=torch.float32)
+
+    with patch.object(RMSNormGated, "forward_native", autospec=True, return_value=expected) as mock_forward_native:
+        out = layer.forward_oot(x, None)
+
+    mock_forward_native.assert_called_once_with(layer, x, None)
+    assert out is expected
diff --git a/vllm_ascend/_310p/ops/layernorm.py b/vllm_ascend/_310p/ops/layernorm.py
index fea8b903..ca0e8c04 100644
--- a/vllm_ascend/_310p/ops/layernorm.py
+++ b/vllm_ascend/_310p/ops/layernorm.py
@@ -1,5 +1,6 @@
 import torch
 import torch_npu
+from vllm.model_executor.layers.layernorm import RMSNormGated
 
 from vllm_ascend.ops.layernorm import AscendGemmaRMSNorm, AscendRMSNorm
 
@@ -37,3 +38,14 @@ class AscendGemmaRMSNorm310(AscendGemmaRMSNorm):
 
         x, _ = torch_npu.npu_rms_norm(x, 1.0 + self.weight, self.variance_epsilon)
         return x
+
+
+class AscendRMSNormGated310(RMSNormGated):
+    def forward_oot(
+        self,
+        x: torch.Tensor,
+        z: torch.Tensor | None = None,
+    ) -> torch.Tensor:
+        # 310P should not depend on the Triton-gated layernorm path.
+        # Reuse the upstream native implementation directly.
+        return super().forward_native(x, z)
diff --git a/vllm_ascend/utils.py b/vllm_ascend/utils.py
index 841e8562..c1e1db42 100644
--- a/vllm_ascend/utils.py
+++ b/vllm_ascend/utils.py
@@ -661,7 +661,11 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
     if is_310p():
         from vllm_ascend._310p.fused_moe.fused_moe import AscendFusedMoE310, AscendSharedFusedMoE310
         from vllm_ascend._310p.ops.activation import AscendSiluAndMul310
-        from vllm_ascend._310p.ops.layernorm import AscendGemmaRMSNorm310, AscendRMSNorm310
+        from vllm_ascend._310p.ops.layernorm import (
+            AscendGemmaRMSNorm310,
+            AscendRMSNorm310,
+            AscendRMSNormGated310,
+        )
         from vllm_ascend._310p.ops.mm_encoder_attention import AscendMMEncoderAttention310
         from vllm_ascend._310p.ops.rotary_embedding import AscendRotaryEmbedding310
         from vllm_ascend._310p.ops.vocab_parallel_embedding import (
@@ -675,6 +679,7 @@ def register_ascend_customop(vllm_config: VllmConfig | None = None):
                 "RotaryEmbedding": AscendRotaryEmbedding310,
                 "RMSNorm": AscendRMSNorm310,
                 "GemmaRMSNorm": AscendGemmaRMSNorm310,
+                "RMSNormGated": AscendRMSNormGated310,
                 "FusedMoE": AscendFusedMoE310,
                 "SharedFusedMoE": AscendSharedFusedMoE310,
                 "ParallelLMHead": AscendParallelLMHead310,