I did a few quick performance tests this afternoon on the WiShield, comparing Arduino and Maple.
Here are the results:
I used a 1030 byte file, and test runs of about 5 minutes.
Arduino, stock stack uIP buffer set to default 400 bytes:
17 transactions per second
Maple, stock iperry port, uIP buffer set to default 400 bytes:
18.4 transactions per second
Maple, iperry port, uIP buffer set to 1500 bytes:
30.4 transactions per second
. . . .
For the Arduino, I used the stock stack from AsyncLabs:
http://github.com/asynclabs/WiShield
For the Maple, I used Perry Hung's port of the AsyncLabs stack:
http://github.com/iperry/maple-library/tree/wishield-dev
The performance test setup was Grinder (http://grinder.sourceforge.net/) using an adapted http.py script that did not save the downloaded pages.
I would like the stack to be even faster, but some initial experiments with increasing the number of connections did not result in faster speeds.