Re: Breaking /emul/ia32-linux for squeeze
Steve Langasek <vorlon@debian.org> writes:
> On Tue, Mar 17, 2009 at 06:29:08PM +0100, Simon Richter wrote:
>
>> On Mon, Mar 16, 2009 at 03:21:44PM +0100, Marco d'Itri wrote:
>
>> > I think it would be very helpful if somebody could summarize why a
>> > multiarch system is useful, except for the obvious case of installing
>> > proprietary i386 software on amd64 systems.
>
>> Most applications don't really need the extended address space, and run a
>> bit faster and need less memory in 32 bit mode (it's less pronounced on
>> i386 as the 32 bit ISA has less registers).
>
> If by "less pronounced" you mean "not true at all". Otherwise, show
> benchmarks.
mrvn@frosties-32:~% time gzip -9 <debian-testing-i386-netinst.iso >/dev/null
gzip -9 < debian-testing-i386-netinst.iso > /dev/null 13.26s user 0.13s system 92% cpu 14.518 total
mrvn 30604 84.2 0.0 1720 640 pts/13 R+ 21:23 0:03 gzip -9
mrvn@frosties:~% time gzip -9 <debian-testing-i386-netinst.iso >/dev/null
gzip -9 < debian-testing-i386-netinst.iso > /dev/null 11.31s user 0.10s system 99% cpu 11.470 total
mrvn 30608 117 0.0 4016 736 pts/13 R+ 21:23 0:02 gzip -9
but
mrvn@frosties-32:~% time bzip2 <debian-testing-i386-netinst.iso >/dev/null
bzip2 < debian-testing-i386-netinst.iso > /dev/null 73.17s user 0.23s system 99% cpu 1:13.89 total
mrvn 30596 95.7 0.6 8888 6928 pts/13 R+ 21:23 0:03 bzip2
mrvn@frosties:~% time bzip2 <debian-testing-i386-netinst.iso >/dev/null
bzip2 < debian-testing-i386-netinst.iso > /dev/null 75.30s user 0.25s system 99% cpu 1:16.15 total
mrvn 30546 94.2 0.6 13212 7016 pts/13 R+ 21:22 0:09 bzip2
It can go either way on speed (or the noise is just bigger than the
speedup) but the memory footprint is bigger in 64bit.
MfG
Goswin
Reply to: