I just thought that I'd mention this, because it wasn't completely
obvious to me at first, and maybe someone else who happens to google for
this subject might come up with this result.
Using the stage3 gentoo arm-softfloat-linux-uclibc filesystem, I had for
some time been natively compiling packages for my target system, with
the root located over NFS.
Normally it wasn't so bad - sure it was quite slower than
cross-compiling. But one thing I hated about cross compiling was
randomly running into packages that use some sort of pkgconfig file that
I didn't have present, or trying to run some native program in configure
and coming up with an error.
When I got around to compiling the GNU Classpath (for use with jamvm,
and stripped of all but the necessities), I found that compiling this
app over an NFS root failed quite often because of out-of-memory errors.
The process was killed outright. Then I read somewhere about creating a
swap file on an NFS root, using losetup to create a loop device, and
using swapon to load that swap into the kernel's VM. That turned out to
be the worst idea I think I've ever had, because the overhead of using
RPC with the NFS root was probably of O^2 complexity at that point,
especially considering all of the swapping that had to be done with such
a huge compile, using jikes of course.
Then I thought - hey, i have 2 external hard disks right in front of me
that connect via USB - my board has USB - (lightbulb !). Now the jikes
process is running at a staggering 50-70% as opposed to the 1-10% it was
before. The overhead is considerably less with usb-storage, even with
the swap file located on the external USB hard disk.
~/Chris
--
[EMAIL PROTECTED] mailing list
