Ethernet testing of ubuntu-14.04-headless-z7020-20150130.1
Posted: Sun Mar 29, 2015 7:37 pmHi All,
I am very pleased to report that ethernet performance is significantly improved versus the jun2014 HDMI release. On the Jun2014 release the best I was able to achieve was 35MBytes/s (where a Mbyte is 1024*1024 bytes). I am now getting 69.5MBytes/s sustained (30+ minutes). Now if I could only get it up to 85MBytes/s
Both tests were performed using a QT based multi-threaded server on a MAC laptop, with the Parallella running as a client with 2 threads and a shared memory pool (1 producer thread which presently just manipulates semaphores, until I get DMA working for my FPGA project, and 1 consumer thread which writes to the TCP socket when data is ready.)
TOP reports 110% CPU usage for this scenario. (1.1 cores used).
64kBytes payload per write. Application on Parallella built using QT 5.2.1.
Tested both direct connected to a dedicated subnet, and through an external switch D-LINK DGS-1005D (still on a dedicated subnet).
Best Regards,
Mike
I am very pleased to report that ethernet performance is significantly improved versus the jun2014 HDMI release. On the Jun2014 release the best I was able to achieve was 35MBytes/s (where a Mbyte is 1024*1024 bytes). I am now getting 69.5MBytes/s sustained (30+ minutes). Now if I could only get it up to 85MBytes/s
Both tests were performed using a QT based multi-threaded server on a MAC laptop, with the Parallella running as a client with 2 threads and a shared memory pool (1 producer thread which presently just manipulates semaphores, until I get DMA working for my FPGA project, and 1 consumer thread which writes to the TCP socket when data is ready.)
TOP reports 110% CPU usage for this scenario. (1.1 cores used).
64kBytes payload per write. Application on Parallella built using QT 5.2.1.
Tested both direct connected to a dedicated subnet, and through an external switch D-LINK DGS-1005D (still on a dedicated subnet).
Best Regards,
Mike