When define __CONSTRAINED__ , client get message 12ms slower.

[x44yz]

Hello. Please forgive me for my poor English.

I create a simple ping-pong test project ( after host and client create connect, client send a empty message to host, when host got message then send a empty message back to client). If undefine CONSTRAINED, the duration is 3ms. When define CONSTRAINED, the duration is 15-17ms.

I am wondering why the delay is so big when __CONSTRAINED. Any help is appreciated, Thanks.

[lodicolo]

I am also seeing this, but based on the code that gets compiled when __CONSTRAINED__ is specified I do not understand how it can be so much slower.

Running the Win32 variant vs __CONSTRAINED__ variants of the platform-specific NetUtility/NetTime functions through a tight-loop with a stopwatch, I would expect (if it actually the difference) Win32 to be much faster, but it's actually orders of magnitude slower on half the functions.

None of the functions show up as big red flags in a CPU profile either.

Edit: So just by copying code from PlatformConstrained to PlatformWin32 I discovered that it seems like Environment.TickCount is what is slowing it down. I do not notice this in tight-loop timing tests (it's actually twice as fast as Stopwatch.GetTimestamp()) but for whatever reason it does appear when actually sending messages. I do not know why Stopwatch.GetTimestamp() isn't being used, but in my fork I changed the Environment.TickCount (which is actually milliseconds) to DateTime.UtcNow.Ticks (which is actually ticks, so I also had to convert it to the desired units with the TimeSpan.TicksPer* constants) and it appears to have mostly resolved the time issue.