[AMD] Dsv4/pr2 compressor opt (#26208)

Co-authored-by: wunhuang <wunhuang@amd.com>
Co-authored-by: Thomas Wang <1am9trash@gmail.com>
Co-authored-by: Xinyi Song <86638975+RolaoDenthu@users.noreply.github.com>
Co-authored-by: HaiShaw <hixiao@gmail.com>
Co-authored-by: amd-danli103 <danli103@amd.com>
Co-authored-by: Lin, Soga <soga.lin@amd.com>
Co-authored-by: Raiden-Makoto <Raiden-Makoto@users.noreply.github.com>
Co-authored-by: Hubert Lu <55214931+hubertlu-tw@users.noreply.github.com>
Co-authored-by: yichiche@amd.com <jacky.cheng>
Co-authored-by: yctseng0211 <yctseng@amd.com>
Co-authored-by: Bingxu Chen <bingxche@amd.com>

sgl-kernel/python/sgl_kernel/__init__.py

@@ -36,6 +36,9 @@ else:
         concat_mla_absorb_q,
         concat_mla_k,
         copy_to_gpu_no_ce,
+        dsv4_fused_k_norm_rope_flashmla,
+        dsv4_fused_q_indexer_rope_hadamard_quant,
+        dsv4_fused_q_norm_rope,
         fused_add_rmsnorm,
         gelu_and_mul,
         gelu_tanh_and_mul,
@@ -125,6 +128,7 @@ else:
 
     if torch.version.hip is not None:
         from sgl_kernel.elementwise import gelu_quick
+        from sgl_kernel.top_k import deepseek_v4_topk_transform_512
 
     if hasattr(torch.version, "musa") and torch.version.musa is not None:
         from sgl_kernel.musa import (
@@ -152,6 +156,9 @@ else:
         "cutlass_mla_get_workspace_size",
         "dsv3_fused_a_gemm",
         "dsv3_router_gemm",
+        "dsv4_fused_k_norm_rope_flashmla",
+        "dsv4_fused_q_indexer_rope_hadamard_quant",
+        "dsv4_fused_q_norm_rope",
         "es_fp8_blockwise_scaled_grouped_mm",
         "es_sm100_mxfp8_blockscaled_grouped_mm",
         "es_sm100_mxfp8_blockscaled_grouped_quant",
@@ -205,6 +212,7 @@ else:
 
     if torch.version.hip is not None:
         _DEBUG_EXPORT_NAMES.append("gelu_quick")
+        _DEBUG_EXPORT_NAMES.append("deepseek_v4_topk_transform_512")
 
     for _name in _DEBUG_EXPORT_NAMES:
         if _name in globals():