Re: Normal buildtime for a kernel?
Lennart Sorensen schrieb:
On Wed, Mar 23, 2005 at 02:53:16PM +0100, Oliver.Korpilla@gmx.de wrote:
That's the strange thing - my AMD64 3500+ in 64-bit mode, with 2 GB of
memory and a 250 Gb drive, takes about 30min. But this system _should_
have more oomph, not less!
Well the Athlon 2800+ is 2087Mhz so at least for raw clock speed it is
not much slower than your cpu. Given it is doing 64bit for every
pointer instead of 32bit you will have a bit bigger code, and hence a
bit more disk access time needed. I may also be wrong on 20 minutes.
It might be closer to 30, but I haven't timed it in a while. Given gcc
almost certainly does nothing to take advantage of the athlon64
instruction set, I wouldn't be surprised if the 3500+ A64 is not really
much faster than an Athlon 2800+ at compiling.
Yeah, but slower??
What speed is your ram and is it single or dual channel (depends on the
socket type in your system). This system is dual channel DDR333 (to
match the 333FSB of the Barton core), so it has nice memory bandwidth.
DDR400 PC3200 Dual-Channel
If your system is a socket 754 then you at least theoretically have less
Socket 939
memory bandwidth than this system which could make compiling slower
since it goes through a lot of memory accesses when compiling.
Well, but more than a half of the compile time up to 2/3rds account to
system time, which I think is somewhat excessive.
Well you could try doing a make -j 2 or whatever the current method to
tell the kernel build system to do multiple things at once since then
you might be able to have one task compiling in memory while the other
waits for the disk. Maybe, although not for certain.
I already do -j 3.
Maybe it's not processor related, but to the disk driver? Any way to
find out?
Well what does hdparm -t /dev/sda say your disk speed is able to do?
What kind of HD and what size?
250GB Seagate SATA.
# hdparm -v /dev/sda
/dev/sda:
IO_support = 0 (default 16-bit)
readonly = 0 (off)
readahead = 256 (on)
geometry = 30401/255/63, sectors = 250059350016, start = 0
With kind regards,
Oliver Korpilla
Reply to: