[Piuparts-devel] faster-rdep-calc

[Andreas Beckmann]

On 2013-03-09 02:41, Dave Steele wrote:
> On Fri, Mar 8, 2013 at 7:21 AM, Andreas Beckmann <anbe at debian.org> wrote:
>>
>> ... the change in time is not significant (and
>> may depend on overall system load) or whatever
> 
> Hmmm. Speed improvement is the point of the branch.

I just wanted to say that my measurements (/usr/bin/time
piuparts-report) are probably not that reliable in terms of speed since
the background load (on cpu and disk) is not taken into account at all.
And it is rather hard (and time-consuming) to do reliable speed
benchmarks in my setup.

I just grepped for all the timings I had in my mailbox ... while memory
usage is pretty constant (since it dropped from 5.6 GB to 1.2 GB),
wallclock time is jumping wild between 2:45 h and 3:30 h. User and
system time are jumping around too, and seem not correlated to wallclock
time.

> Perhaps we
> shouldn't be in a hurry to merge it.

Do you expect a speed improvement if the rdep count is needed for *all*
packages (as done by p-report)?. I'd expect an improvement for p-master
if you only need the values for waiting-to-be-tested (and their rdeps),
and that usually amounts to much less than all packages.

I also merged rdep-chain-len for local testing :-)
Do you use "full" depth or only "blocked" depth?

786.00user 674.64system 3:06:11elapsed 13%CPU (0avgtext+0avgdata
1146848maxresident)k
508520inputs+19310688outputs (243major+15487557minor)pagefaults 0swaps

No influence on memory usage ,too.


Andreas