[ENet-discuss] New goodies in ENet CVS

[Intripoon]

Hi !

I upgraded from the previous version without these features to this one. I
noticefd that sending is a lot faster than before. Instead of 6MB/s in my
10MB sending test, I now get to the limit of my 100MBit/s network. However,
the effect was, that my older computer sometimes just drops the connection
instead of receiving the data. Setting the outgoing bandwidth of the sender
to 1000000 instead of 0 made it work again. However, my bandwidth monitor
tells me the sender sends with 8MB/s instead of 1MB/s. The value of 1000000
seems to have slown it down but not to the specified value. 

Apart from that, the rtt variable still makes me wonder what it actually
represents.

I noticed usage of GetTickCount() in enet on Win32. Using
QueryPerformanceCounter() gives higher accuracy. GetTickCount() is not
accurate in ms. =>
http://blogs.msdn.com/oldnewthing/archive/2005/09/02/459952.aspx

Best regards
	Marc


-----Ursprüngliche Nachricht-----
Von: enet-discuss-bounces at cubik.org [mailto:enet-discuss-bounces at cubik.org]
Im Auftrag von Lee Salzman
Gesendet: Donnerstag, 6. Juli 2006 23:35
An: enet-discuss at cubik.org
Betreff: [ENet-discuss] New goodies in ENet CVS

I put some stuff that was sitting around in Sauerbraten into ENet CVS, and
one entirely new thing that's basically untested:

Sauerbraten-tested stuff:

1) ENET_PACKET_FLAG_NO_ALLOCATE
	
	You can set the "data" field of ENetPacket's which your own data,
instead of having ENet allocate it for you and having to copy it in. The
data argument of enet_packet_create() just sets this field with the pointer
you passed in instead of copying it over when you use this flag.

2) Protocol space overhead is half what it was.

	This is only an issue if you are sending mind-boggling amounts of
small packets, but we were, so it was, and so it's implemented. Some "cute" 
tricks went into this.

Totally untested stuff:

1) enet_disconnect_later()

	This is for all you people who wanted a synchronous disconnect. Once
requested, no more received packets will be delivered, not even ones that
were sitting in the incoming queue. It will wait until all packets in the
outgoing queue have been sent and acknowledged, then do a normal disconnect.
enet_peer_send() attempts will be ignored during this time.

Just a warning that CVS is no longer in sync with the last release. Use at
your peril/pleasure.

Lee
_______________________________________________
ENet-discuss mailing list
ENet-discuss at cubik.org
http://lists.cubik.org/mailman/listinfo/enet-discuss