Am 06.07.2014 21:47, schrieb Lennart Poettering: > BTW, have you checked whether reuseing the XZ context might make the XZ > more competitive?
On Sun Jul 6 15:01:11 PDT 2014 Reindl Harald wrote: > please try a simple test compress 50 MB with XZ and GZ, LZO, LZ4 > or BZIP2 - XZ is *magnitudes* slower in any case Actually it is not, or rather, it is only that if you want it to be that. I did a quick comparison using the systemd 214 tar (30 MiB): xz -0 was 28% slower than lz4c -hc, but the result was 19% smaller. xz -0 was even 44% faster than gzip -9, and the result was still 5% smaller. > XZ is the wrong compression for anything where user feedback in time matters No, xz -6 is the wrong compression *setting* for that use-case, but at a lower setting xz is quite suitable even for that. > even with no care about memory usage which also is part of the game finally xz -0 needs about 3 MiB for compression, hardly a prohibitive requirement. --- jon@ilyena /tmp $ time lz4c -hc systemd-214.tar Compressed filename will be : systemd-214.tar.lz4 Compressed 31436800 bytes into 5120576 bytes ==> 16.29% real 0m2.751s user 0m2.696s sys 0m0.052s jon@ilyena /tmp $ time gzip -9 -k systemd-214.tar real 0m6.247s user 0m6.164s sys 0m0.032s jon@ilyena /tmp $ time xz -0 -k systemd-214.tar real 0m3.715s user 0m3.464s sys 0m0.032s jon@ilyena /tmp $ ls -s systemd-214.tar* 30700 systemd-214.tar 5004 systemd-214.tar.lz4 4288 systemd-214.tar.gz 4072 systemd-214.tar.xz _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
