| 00:53 | dv__ | bencoh: what did you mean? |
| 00:54 | dv_ | uwah looks like I'm getting far behind meta-cubox rep |
| 00:55 | dv_ | note that my meta-cubox layer and naguirre's older one are completely different. mine is not a fork, it is done from scratch. |
| 01:00 | bencoh | hm, I cloned mine from naguirre's |
| 01:00 | bencoh | (as far as I remember) |
| 01:01 | bencoh | I guess I should yours ... someday(tm) :) |
| 01:01 | bencoh | +look at |
| 01:10 | dv_ | initially I tried out his |
| 01:10 | dv_ | but it didnt work properly with yocto, and many things were missing that I wanted |
| 01:11 | dv_ | anyway, mine works with both danny and dylan versions of yocto |
| 01:11 | dv_ | feedback welcome :) |
| 01:53 | dv_ | on another note, perhaps somebody knows what is going on? |
| 01:53 | dv_ | 1080p video with hardware decoder on an arm platform: 77% CPU usage. then I comment out the memcpy from the decoder's DMA buffer to the output frame's GstBuffer: 12%. |
| 01:53 | dv_ | (well, the arm platform being the cubox actually :) ) |
| 01:54 | dv_ | this is at 30 fps, so with UYVY pixels, thats about 240 MB/s. not that much, actually. is memory bandwidth very low with the cubox? |
| 12:19 | dv505 | hooray, one year of irc logging on this channel |
| 12:19 | dv505 | got 4MB of archived text |
| 12:19 | dv505 | any one even reading those logs? |
| 12:20 | Coolgeek | 12:20 (cubox~cubox) /home/cubox/.weechat/logs |
| 12:20 | Coolgeek | % ls -l *cubox* [0.03 0.04 - 20%] |
| 12:20 | Coolgeek | 12:20 (cubox~cubox) /home/cubox/.weechat/logs |
| 12:20 | Coolgeek | it's usefull if you search for something |
| 12:23 | dv505 | online too, that's what I was talking about: http://server.vijge.net/static/cubox/irclog/ |
| 18:54 | MMlosh | is it possible, that my cubox's sata does only 15MB/s when writing zeros over a HDD? |
| 18:59 | dv_ | sounds low |
| 18:59 | dv_ | check the hd settings with hdparm |
| 19:00 | dv_ | dd if=/dev/zero of=/path/to/hdd/mount/zeros-dump bs=32768 should give you more than just 15 MB/s |
| 19:02 | MMlosh | hmm.. bigger blocksize might help |
| 19:05 | MMlosh | dv_, it says: multcount=0 (default), IO_support=0 (default), readonly=0 (off), readahead=256 (on), geometry=30401/256/63 sectors=488397168 start=0 |
| 19:06 | MMlosh | not much settings.. did I miss a parameter? |
| 19:07 | MMlosh | the only difference is me writing to /dev/sdb.. if I wanted a file full of zeros, I would make a file with a hole with truncate |
| 19:09 | MMlosh | dv_, the blocksize seems to do the trick so far does that mean that the the driver sucks or IO scheduler is misconfigured? |
| 19:09 | MMlosh | you were right.. 68MB/s |
| 19:10 | MMlosh | not awesome, but tolerable... |
| 19:22 | MMlosh | oh wait... 68MBps = 544mbps > USB 2.0 |
| 19:23 | MMlosh | which does 480mbps |
| 22:48 | MMlos | 22:48 * MMlosh is disappoint.. Kernel doesn't seem to see UUIDs (it says 0000-0000-etc), so I cannot use them for "root=" |