[AMD] [GLM-5.3-Flash Day 0] Build the fused DSA k-pool top-k JIT kernel on HIP (#39339)

Co-authored-by: Raiden-Makoto <Raiden-Makoto@users.noreply.github.com>
Co-authored-by: Thomas Wang <thomawan@amd.com>
Co-authored-by: Kevin Mi <mikevin920@yahoo.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

python/sglang/kernels/jit/csrc/dsa/kpool_topk_transform.cuh

@@ -11,7 +11,11 @@
 #include <bit>
 #include <cstddef>
 #include <cstdint>
+#ifndef USE_ROCM
 #include <cuda_fp16.h>
+#else
+#include <hip/hip_fp16.h>
+#endif
 
 namespace sglang {
 namespace {

python/sglang/kernels/jit/include/sgl_kernel/utils.cuh

@@ -53,6 +53,8 @@ inline constexpr auto cudaSuccess = hipSuccess;
 #define cudaDeviceGetAttribute hipDeviceGetAttribute
 #define cudaDevAttrComputeCapabilityMajor hipDeviceAttributeComputeCapabilityMajor
 #define cudaDevAttrComputeCapabilityMinor hipDeviceAttributeComputeCapabilityMinor
+#define cudaFuncSetAttribute hipFuncSetAttribute
+#define cudaFuncAttributeMaxDynamicSharedMemorySize hipFuncAttributeMaxDynamicSharedMemorySize
 #endif
 
 namespace sglang {