Differences
This shows you the differences between two versions of the page.
|
soc:2010:cooldavid:journal:weekb1 [2010/05/26 09:51] cooldavid created |
soc:2010:cooldavid:journal:weekb1 [2010/05/26 09:57] cooldavid |
||
|---|---|---|---|
| Line 1: | Line 1: | ||
| === Week -1: Modify JMicron driver, RX offloading === | === Week -1: Modify JMicron driver, RX offloading === | ||
| - | === May 2nd to May 8th === | + | === May 2nd to May 12th === |
| == Chanloading with undionly image == | == Chanloading with undionly image == | ||
| Line 16: | Line 16: | ||
| == Implement RX checksum offloading == | == Implement RX checksum offloading == | ||
| Testing results can be found in this note: [[soc:2010:cooldavid:notes:benchmark|TCP Window Size and RX Csum Offload Benchmark]] | Testing results can be found in this note: [[soc:2010:cooldavid:notes:benchmark|TCP Window Size and RX Csum Offload Benchmark]] | ||
| + | |||
| + | == Modify JMicron driver according to stefanha suggestions == | ||
| + | - jme_process_receive() isn't using a read memory barrier. | ||
| + | - Allocate/Free Ring buffers on open/close. | ||
| + | - Using jme_free_?x_resources() for allocation fail recover. | ||
| + | - Simply use struct io_buffer, instead of jme_buffer_info. | ||
| + | - There is no need to discard pending TX buffers. | ||
| + | - extra memset in probe. | ||
| + | - Ring size tune, performance test. | ||
| + | |||
| + | Detailed modifications can reference the [[http://git.etherboot.org/?p=people/cooldavid/gpxe.git;a=shortlog;h=refs/heads/jme_driver_bak|jme_driver_bak branch of my git tree]]. | ||
