Contact
Subscribe

Building a kernel on the eeepc.

Jan 7, 2009
Uncategorized

Don’t try this at home.

For the record: It takes an eeepc 901 just under nine hours to build a Fedora kernel.

Yes, dumb idea, but I just had to know.

Related posts:

  1. Not attending kernel summit. Tomorrow, I should have been getting on a plane, and...
  2. An annoying kernel packaging bug. The kernel rpm package creates an initramfs file that gets...

Related posts brought to you by Yet Another Related Posts Plugin.

11 Comments
Tags ,

11 Comments

  1. Gawain Lynch  •  Jan 7, 2009 @22:00

    If you can do without your 901 for a week or so, enquiring minds want to know how long openoffice.org takes to compile. :-P

  2. davej  •  Jan 8, 2009 @00:13

    Yeah ‘a week or so’ is my guess too. I think the ssd might die before it completes :)

  3. kmike  •  Jan 8, 2009 @02:14

    It certainly took much much less on Acer Aspire One w/512MB RAM and 8GB SSD. Less than 2 hours I’d think, but that was with the concurrency level 2.

  4. davej  •  Jan 8, 2009 @07:15

    single core celeron + crap first gen SSD = fail.

  5. Valerie Aurora  •  Jan 8, 2009 @10:18

    I am either aghast or in awe. No, no, I’m aghast.

  6. Bryan  •  Jan 8, 2009 @11:47

    It must be the SSD, because my AMD Geode GX + 4200 rpm HDD does the compile in about an hour (for the stock kernel).

    What does hdparm give for performance values? It sounds to me like the 901 is a waste of money.

  7. davej  •  Jan 8, 2009 @11:52

    hdparm only benchmarks reads, which aren’t too shabby. Where this ssd sucks, is writes. They take forever.

  8. Bryan  •  Jan 8, 2009 @15:18

    How did you test writes? dd?

    I don’t suppose you can have the Makefile write object code to a tmpfs mount?

    Maybe gcc -pipe will help a little. Also, out of curiosity, is the SSD replacable? It’s not a hardwired NAND flash chip (like in the OLPC) is it? Of course then it would be super awesome because you could use jffs2 and writes wouldn’t suck because you wouldn’t have lousy hardware wear leveling or whatever goofiness they use that is slowing down the write speed.

  9. davej  •  Jan 8, 2009 @23:46

    tmpfs wouldn’t help, because it’s crippled with RAM too. Only 1GB iirc.
    (It’s off right now, and I’m too lazy to power it up to check).

    -pipe .. I’m pretty sure the kernel build does that by default. Again, I’m too lazy to check right now.

    And removable SSD.. I think the 700 series was soldered to the board. not sure at all about the 900 series.

  10. kmike  •  Jan 9, 2009 @03:28

    In fact the SSD on the Aspire One is extremely slow when writing, too, so it must be something else. Maybe the Celeron fails here, or maybe me setting dirty_writeback_centisecs to 15000 has helped?

    On the related note:
    http://kernelslacker.livejournal.com/132087.html?thread=680439#t680439

  11. Artem S. Tashkinov  •  Jan 9, 2009 @14:49

    Almost certainly you used GCC 4.3.x for compiling. If you had used GCC 3.4.x it would have probably taken three times less time.



  • huaglahglah huaglahglah