[XPU] Enable torch.compile for XPU GDN attention

vllm/_xpu_ops.py

@@ -92,6 +92,72 @@ def _int4_gemm_w4a16_fake(
         return torch.empty((M, N), dtype=input.dtype, device=input.device)
 
 
+def _gdn_attention_core_xpu_impl(
+    core_attn_out: torch.Tensor,
+    z: torch.Tensor,
+    projected_states_qkvz: torch.Tensor,
+    projected_states_ba: torch.Tensor,
+    layer_name: str,
+) -> None:
+    """Custom op wrapping the XPU SYCL GDN kernel for torch.compile."""
+    from vllm.forward_context import get_forward_context
+    from vllm.v1.attention.backends.gdn_attn import GDNAttentionMetadata
+
+    forward_context = get_forward_context()
+    self = forward_context.no_compile_layers[layer_name]
+    attn_metadata = forward_context.attn_metadata
+
+    if attn_metadata is None:
+        return
+
+    assert isinstance(attn_metadata, dict)
+    attn_metadata = attn_metadata[self.prefix]
+    assert isinstance(attn_metadata, GDNAttentionMetadata)
+
+    # TODO: xpu does not support speculative decoding yet
+    assert attn_metadata.spec_sequence_masks is None
+
+    conv_weights = self.conv1d.weight.view(
+        self.conv1d.weight.size(0), self.conv1d.weight.size(2)
+    )
+
+    torch.ops._xpu_C.gdn_attention(
+        core_attn_out,
+        z,
+        projected_states_qkvz,
+        projected_states_ba,
+        self.num_k_heads,
+        self.num_v_heads,
+        self.head_k_dim,
+        self.head_v_dim,
+        conv_state=self.kv_cache[0],
+        ssm_state=self.kv_cache[1],
+        conv_weights=conv_weights,
+        conv_bias=self.conv1d.bias,
+        activation=self.activation,
+        A_log=self.A_log,
+        dt_bias=self.dt_bias,
+        num_prefills=attn_metadata.num_prefills,
+        num_decodes=attn_metadata.num_decodes,
+        has_initial_state=attn_metadata.has_initial_state,
+        non_spec_query_start_loc=attn_metadata.non_spec_query_start_loc,
+        non_spec_state_indices_tensor=attn_metadata.non_spec_state_indices_tensor,
+        num_actual_tokens=attn_metadata.num_actual_tokens,
+        tp_size=self.tp_size,
+        reorder_input=not self.gqa_interleaved_layout,
+    )
+
+
+def _gdn_attention_core_xpu_fake(
+    core_attn_out: torch.Tensor,
+    z: torch.Tensor,
+    projected_states_qkvz: torch.Tensor,
+    projected_states_ba: torch.Tensor,
+    layer_name: str,
+) -> None:
+    return
+
+
 def _xpu_ops_deepseek_scaling_rope_impl(
     positions: torch.Tensor,
     query: torch.Tensor,
@@ -618,6 +684,13 @@ def register_ops_once() -> None:
                 fake_impl=_xpu_mxfp4_quantize_fake,
             )
 
+            direct_register_custom_op(
+                op_name="gdn_attention_core_xpu",
+                op_func=_gdn_attention_core_xpu_impl,
+                mutates_args=["core_attn_out", "z"],
+                fake_impl=_gdn_attention_core_xpu_fake,
+            )
+
             _OPS_REGISTERED = True
 
 

vllm/config/compilation.py

@@ -742,6 +742,7 @@ class CompilationConfig:
         "vllm::linear_attention",
         "vllm::plamo2_mamba_mixer",
         "vllm::gdn_attention_core",
+        "vllm::gdn_attention_core_xpu",
         "vllm::olmo_hybrid_gdn_full_forward",
         "vllm::kda_attention",
         "vllm::sparse_attn_indexer",

vllm/model_executor/layers/mamba/gdn_linear_attn.py

@@ -620,53 +620,20 @@ def forward_xpu(
         # ============================================================
         # Part 2: Core Attention
         # ============================================================
-        forward_context = get_forward_context()
-        attn_metadata: AttentionMetadata = forward_context.attn_metadata
         core_attn_out = torch.zeros(
             (num_tokens, self.num_v_heads // self.tp_size, self.head_v_dim),
             dtype=hidden_states.dtype,
             device=hidden_states.device,
         )
         z = torch.empty_like(core_attn_out)
-        if attn_metadata is not None:
-            attn_metadata = attn_metadata[self.prefix]
-
-            # TODO: xpu does not support this param yet
-            spec_sequence_masks = attn_metadata.spec_sequence_masks
-            assert spec_sequence_masks is None
 
-            conv_weights = self.conv1d.weight.view(
-                self.conv1d.weight.size(0), self.conv1d.weight.size(2)
-            )
-
-            conv_state = self.kv_cache[0]
-            ssm_state = self.kv_cache[1]
-
-            torch.ops._xpu_C.gdn_attention(
-                core_attn_out,
-                z,
-                projected_states_qkvz,
-                projected_states_ba,
-                self.num_k_heads,
-                self.num_v_heads,
-                self.head_k_dim,
-                self.head_v_dim,
-                conv_state=conv_state,
-                ssm_state=ssm_state,
-                conv_weights=conv_weights,
-                conv_bias=self.conv1d.bias,
-                activation=self.activation,
-                A_log=self.A_log,
-                dt_bias=self.dt_bias,
-                num_prefills=attn_metadata.num_prefills,
-                num_decodes=attn_metadata.num_decodes,
-                has_initial_state=attn_metadata.has_initial_state,
-                non_spec_query_start_loc=attn_metadata.non_spec_query_start_loc,
-                non_spec_state_indices_tensor=attn_metadata.non_spec_state_indices_tensor,
-                num_actual_tokens=attn_metadata.num_actual_tokens,
-                tp_size=self.tp_size,
-                reorder_input=not self.gqa_interleaved_layout,
-            )
+        torch.ops.vllm.gdn_attention_core_xpu(
+            core_attn_out,
+            z,
+            projected_states_qkvz,
+            projected_states_ba,
+            self.prefix,
+        )
 
         # ============================================================
         # Part 3: Output Projection