refactor: namespace csrc kernels and extract common helpers

- attention family -> astrai::attention; fp8 family -> astrai::fp8
- new common/reduce.cuh (warp/group reductions, atomic_max_float)
- new common/cp_async.cuh (predicated cp_async_16, commit/wait group)
- move MAX_SPLITS into attention/common.h; delete warp_utils.cuh
- .cu bindings and pure C tests open family namespaces via using
This commit is contained in:
2026-08-24 14:49:55 +08:00
parent 34471252ab
commit 31ca357c61
21 changed files with 243 additions and 139 deletions
@@ -26,6 +26,9 @@
#define DEVICE_FORCEINLINE static __device__ __forceinline__
#define HOST_DEV_FORCEINLINE static __host__ __device__ __forceinline__
namespace astrai {
namespace attention {
using bf16 = __nv_bfloat16;
// ============================================================================
@@ -253,3 +256,6 @@ struct PagedKV {
return kv_addr_from_token(p, c, token, d);
}
};
} // namespace attention
} // namespace astrai