[What] Fix data type for host reduction

[Why] F16 issue for host reduction has been fix in c1ef7319

example/19_gemm_softmax/gemm_softmax_xdl_fp16.cpp

@@ -42,10 +42,6 @@ using CDataType              = F16;
 using AccDataType            = F32;
 using EltwiseComputeDataType = F32;
 
-// CAUTION - host reduce_max will call numeric_limits<ck::half_t>::lowest()
-// However, numeric_limits<ck::half_t>::lowest() will return zero. So, used half_float::half instead
-using HostReduceDataType = half_float::half;
-
 using ALayout = ck::tensor_layout::gemm::RowMajor;
 using BLayout = ck::tensor_layout::gemm::ColumnMajor;
 using CLayout = ck::tensor_layout::gemm::RowMajor;
@@ -192,18 +188,18 @@ using DeviceElementwiseDivInstance =
 using HostGemmInstance = ck::tensor_operation::host::
     ReferenceGemm<ADataType, BDataType, CDataType, PassThrough, PassThrough, PassThrough>;
 
-using HostReduceMaxInstance = ReductionHost<HostReduceDataType,
-                                            HostReduceDataType,
-                                            HostReduceDataType,
+using HostReduceMaxInstance = ReductionHost<CDataType,
+                                            CDataType,
+                                            CDataType,
                                             ReduceMaxId,
                                             Rank,
                                             NumReduceDim,
                                             ReducePropagateNan,
                                             false>;
 
-using HostReduceSumInstance = ReductionHost<HostReduceDataType,
+using HostReduceSumInstance = ReductionHost<CDataType,
                                             AccDataType,
-                                            HostReduceDataType,
+                                            CDataType,
                                             ReduceSumId,
                                             Rank,
                                             NumReduceDim,
@@ -510,9 +506,9 @@ int main(int argc, char* argv[])
 
         host_gemm_invoker.Run(host_gemm_argument);
         host_reduce_max.Run(1, // alpha
-                            reinterpret_cast<const HostReduceDataType*>(c_m_n.mData.data()),
+                            c_m_n.mData.data(),
                             0, // beta
-                            reinterpret_cast<HostReduceDataType*>(host_c_n_max.mData.data()),
+                            host_c_n_max.mData.data(),
                             host_indices.mData.data());
 
         host_broadcast2D<Tensor<CDataType>,
@@ -523,9 +519,9 @@ int main(int argc, char* argv[])
                          0>(host_exp_m_n, c_m_n, c_n_max, M, N, SubExp{});
 
         host_reduce_sum.Run(1, // alpha
-                            reinterpret_cast<const HostReduceDataType*>(exp_m_n.mData.data()),
+                            exp_m_n.mData.data(),
                             0, // beta
-                            reinterpret_cast<HostReduceDataType*>(host_exp_n_sum.mData.data()),
+                            host_exp_n_sum.mData.data(),
                             host_indices.mData.data());
 
         host_broadcast2D<Tensor<CDataType>,