Discrepancy in Handling Large Data Writes: tokio::net::TcpStream vs std::net::TcpStream`

[yingjun-wu]

When I use tokio::net::tcpstream and std::net::tcpstream to build communication with the runner in the enclave, I find that when the length of the transmitted data exceeds 16384, the tokio::net::tcpstreammethod will divide the write into several times, while the std::net::tcpstream method will not. Why is this?

Is there any way to change this method, because when splitting multiple transmissions, the overall delay will be much larger than one transmission time?

rust-sgx/intel-sgx/enclave-runner/src/usercalls/interface.rs

Lines 73 to 77 in b6f0262

	fn write(
	self,
	fd: Fd,
	buf: *const u8,
	len: usize,

I mainly understand the size of len through the L77 line of code.

Thank You!!