Hi,

Latetely I've been trying to optimize the time it takes to rebuild the 
beaglebone factory image, which is delivered as a 4GiB SD card image people can 
dd to their SD cards. I tried and failed to replace genext2fs with mkfs.ext4 
since uboot can only read from a ext2 -> tune2fs -> ext4 partition, not from a 
mkfs.ext4'ed partition. So the next step was to tackle the compression step. 

All tests done on a intel i7 with 16GiB  RAM

Regular xz:

$ time xz -k -e 4gb.img

real    7m26.158s
user    6m38.345s
sys     0m2.076s

$ du -hs *xz
143M 4gb.img.xz

With https://github.com/vasi/pixz:

$ time pixz -t 4gb.img 4gb.img.xz
real    2m9.997s
user    9m41.085s
sys     0m6.990s

$ du -hs *xz
151M 4gb.img.xz

After posting those stats on G+ Brian Marete pointed me to the alpha releases 
of xz that have pthread support:

With xz git master and -T8:
real    1m43.144s
user    10m29.579s
sys     0m5.456s

Based on that I'd like to update xz(-native) to the alpha release, but I'm not 
familiar enough with the xz project to say if that's a good idea or not. So, 
any comments from xz users?

regards,

koen


_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.linuxtogo.org/cgi-bin/mailman/listinfo/openembedded-core

Reply via email to