Merge remote-tracking branch 'origin/develop' into cpu_avx2

include/ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer_v6r1.hpp

@@ -102,8 +102,13 @@ struct ThreadwiseTensorSliceTransfer_v6r1
 
             // apply pointwise operation
             static_for<0, ScalarPerVector, 1>{}([&](auto i) {
-                element_op_(dst_vector_container.template AsType<DstData>()(i),
-                            src_vector_container.template AsType<SrcData>()[i]);
+                SrcData v;
+
+                // apply element-wise operation
+                element_op_(v, src_vector_container.template AsType<SrcData>()[i]);
+
+                // apply type convert
+                dst_vector_container.template AsType<DstData>()(i) = type_convert<DstData>(v);
             });
 
             const bool is_dst_valid =