Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
Next revision Both sides next revision
soc:2010:andreif:journal:week2 [2010/06/04 12:56]
andreif
soc:2010:andreif:journal:week2 [2010/06/07 09:07]
andreif
Line 41: Line 41:
 ==== Day 5 [ 4 Jun 2010 ] ==== ==== Day 5 [ 4 Jun 2010 ] ====
  
-Git commit: ​[[ These are not the commits you are looking for :waves hand: ]]+Git commit: These are not the commits you are looking for :waves hand:
  
 Today didn't yield much in terms of code but I did manage to do some testing. First I set up QEmu and ran this small test: Today didn't yield much in terms of code but I did manage to do some testing. First I set up QEmu and ran this small test:
Line 61: Line 61:
  
 Unfortunately,​ I didn't get to go through the feedback on the driver or start working on the link state code. Slow day :( Unfortunately,​ I didn't get to go through the feedback on the driver or start working on the link state code. Slow day :(
 +
 +==== Day 6 [ 5 Jun 2010 ] ====
 +
 +Git commit: [[http://​git.etherboot.org/?​p=people/​andreif/​gpxe.git;​a=commit;​h=551178b108b99374d6918ec7fdc92b8eee8bd47e|551178b108b99374d6918ec7fdc92b8eee8bd47e]]
 +
 +I started today by looking at the link state code in the Linux driver. The problem with it is that it depends on the number of PHYs and the duplex mode of the NIC. The size of the code is comparable to that of some of my other main routines, but I find that its impact is minimal. I decided that for now I would leave link state on in .probe and implement it if someone specifically requests it.
 +
 +Testing followed. Mdc and stefanha both told me to try and download some images larger than 100MB to see what speeds the driver reaches and how it performs. The old driver got an 100MB image in 178s my driver got it 145s. Something was wrong. Thanks to mdc, meteger and stefanha the bug eventually surfaced and got fixed. I was writing some wrong values into the CSR0 register. After this, the driver got the 100MB image in 37s. The following days I'll try to improve this. I'll post some comparisons tomorrow.
 +
 +After a wonderful meeting with stefanha and mdc, I got back to fixing the stuff people told me about in the email thread. In the process, I managed to read (skipping some sections) this [[http://​www.rdrop.com/​users/​paulmck/​scalability/​paper/​whymb.2009.04.05a.pdf|paper]]. I'll go through it tomorrow too to make sure I fully understand the concepts.
 +
 +==== Day 7 [ 31 May 2010 ] ====
 +
 +me.rest()

QR Code
QR Code soc:2010:andreif:journal:week2 (generated for current page)