View previous topic :: View next topic |
Author |
Message |
thechris Veteran
Joined: 12 Oct 2003 Posts: 1203
|
Posted: Sat Aug 14, 2004 11:26 pm Post subject: Idea -- Gentoo At Home or Acovea At Home |
|
|
i was reading on the acovea website that any benchmark that took 1 minute to compile and 1 minute to run would take over 5 days to complete.
one of the criticisms of acovea is that its fairly vague in the tests. what do these tests mean? they test some mathmatical stuff.
My proposal is an idea for a distributed processing for these long acovea runs. basically more elaborate tests could be done for more specific packages. information from acovea runs could be added to a database and then general optimizations and package specific optimizations could be identified. the general optimizations could be forwarded to the package developers as suggestion and then could be shown to gentoo users.
as i see it, if you did 4 hours of acovea per day for 100 users, you could accumulate 16 days of runs in one day. this means you could do testbenchs that take a total of 4 minutes per run and get same day results.
The issues as i see them are:
1.) acovea testbenches must be secured and stripped of dangerous options. i suggest that the acovea-at-home be implemented from inside a chrooted environment and that code be screened for possible malicious code before it is compiled and executed.
2.) it must run with minimal other things running. so checks need to be in place to assure some quality of testing. |
|
Back to top |
|
|
.:deadhead:. Advocate
Joined: 25 Nov 2003 Posts: 2963 Location: Milano, Italy
|
Posted: Sat Aug 21, 2004 9:30 pm Post subject: |
|
|
Being Acovea so hardware dependant it's quite useless.
Acovea stress your hardware, and the resulting flags [after a serius test set (5 times maybe)] can be used only on your machine or on identical machine.
Imagine all the variables present in your PC [processor, RAM, mobo]...
The only thing you can do is:
1) run it in init 1, zero daemons,, so you'll have clean results
2) run it multiple times
3) take a looooot of time
cya _________________ Proudly member of the Gentoo Documentation Project: the Italian Conspiracy ! |
|
Back to top |
|
|
ewan.paton Veteran
Joined: 29 Jul 2003 Posts: 1219 Location: glasgow, scotland
|
Posted: Sun Aug 22, 2004 1:22 am Post subject: |
|
|
i doubt amount ram{1} and motherboard chipsets would affect the performance of a binary leaving the cpu as the major variable. ignoring steppings there are really only six x86 cpu cores that could be variables gentoo users are likely to have pentium2,3 and 4, the athlon, athlon xp and amd64 so a pretty good idea
i may be wrong if different frequency cores have different performance but even then if enough people were involved it would solve this
other variables such as key software components like gcc would also have to be listed
{1} fsb maybe but still dubious _________________ Giay tay nam | Giay nam cao cap | Giay luoi |
|
Back to top |
|
|
.:deadhead:. Advocate
Joined: 25 Nov 2003 Posts: 2963 Location: Milano, Italy
|
Posted: Sun Aug 22, 2004 10:36 am Post subject: |
|
|
ewan.paton wrote: | i doubt amount ram{1} and motherboard chipsets would affect the performance of a binary |
http://www.coyotegulch.com/acovea/index.html wrote: | Memory architectures, pipelines, on- and off-chip caches -- all affect code performance in ways that are not obvious to programmers using a high-level language. An optimization that may seem to produce faster code may, in fact, create large code that causes more cache misses, thus degrading performance. |
2 thechris:
already seen this project? _________________ Proudly member of the Gentoo Documentation Project: the Italian Conspiracy ! |
|
Back to top |
|
|
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|