Posts by [boinc@poland]goofyx
log in
1) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1619)
Posted 3 days ago by Profile [boinc@poland]goofyx
First test CPU WU was from sub-project.. but server died <- to the end of this week I wnt to prepare CPU and NCI on dedicated server and official start it on next week.
I have to prepare new scripts, make somehow badges ect

Sorry, but I have some delayed about spliting GoofyxGrid@Home because I am working on some changes on DrugDiscovery@Home (where I am admin too).
At this moment everythink stays on GG@H without spliting.
2) Message boards : Number crunching : Spliting GoofyxGrid - common pages, forum (Message 1617)
Posted 8 days ago by Profile [boinc@poland]goofyx
I did think about it, but the "main" project will be closed and repleaced by nci.goofyxgridathome.net
3) Message boards : Number crunching : Spliting GoofyxGrid - common pages, forum (Message 1615)
Posted 9 days ago by Profile [boinc@poland]goofyx
1. What do you think about making common forum for every project on GoofyxGrid@Home but without using built-in forum, but somethink like: phpForum, bbForum, Simple Machines.
2. I have some idea for common part of web site, news ect


I don't know what do you think about 2 diffrent forum ect.. but for me it's rather uncomfortable solution.
4) Message boards : Number crunching : gg@home cpu test - intel vs arm (Message 1614)
Posted 9 days ago by Profile [boinc@poland]goofyx
If you compare intel with linux and arm with linux, it is not so different.
Intel with Win is much faster than intel with linux.

Not in everythink.
I tested cpu app on win 64bits and lin 64bit, diffrent in processing time was about 5-7sec. Win 64b need 10-12sec on my i7, and lin 64bit on my i7 need 15-19sec <- it's not much... but this differences will be BIGGER when I start generating WU for about 1-1,5h work.


ps.: I try to install some strand library on my odroid-xu4 and recompiling kernel, but without any success. I am a little confused about it <- or I in big shock... that this arm's cpu don't like random operation.
At this moment I have to make some tests with my neural network which I were planning to place on 25x Odroid-XU4 <- if my 4-5x XU4 will be so slow even with machine learning I will have to re-plan NN.
5) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1612)
Posted 9 days ago by Profile [boinc@poland]goofyx
First test CPU WU was from sub-project.. but server died <- to the end of this week I wnt to prepare CPU and NCI on dedicated server and official start it on next week.
I have to prepare new scripts, make somehow badges ect
6) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1610)
Posted 10 days ago by Profile [boinc@poland]goofyx
I think the server crashed again?

502 Bad Gateway. Can't report workunits nor enter the website

Test server is officially dead :(
I will set up everythink on dedicated server.
7) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1608)
Posted 10 days ago by Profile [boinc@poland]goofyx
I think more time is needed to prepare WU to start and send to server it's why TT7 serie don't load CPU too much.

ps.: Wait for TT8
8) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1606)
Posted 10 days ago by Profile [boinc@poland]goofyx
First test enviroment can't handle server, even right now when I have only 51 hosts.


EDITED at 09:45 CET:
This envirometnt is:
- odroid-xu4 as server
- 10 shared folders from local NAS to speed up work <- I try NSFS and CIFS/SMB
it's look like Odroid-XU4 has to low RAM memory even for apache2 <- it's why server dieds durring big apache2 load

Before moved project to second hardware I will try diffrent params for apache2 and mysql <- but 2GB(1,8GB ram after system start) Ram is too low public boinc server.
What if second enviroment screw up?
I will set up both project on acctuall server which can handle it without any problem

Last info about TT series:
- TT1-TT6 has difrent deadline time and fpops in template
- TT7 has workout with 36,000,000 random words <- 10-15second on my i7-4790 and almost 2h on odroid-xu4
- TT8 has 36,000,000,000 random words <- iI hope it will gave about 0,75-1 hour on i7 and I won't test it on odroid-xu4 (doesn't make any sense)
9) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1605)
Posted 11 days ago by Profile [boinc@poland]goofyx
Are you going to have this added to BAM! and Free-DC?
Thanks

Finnaly yes.
But after I decided this enviroment is good enought to became production version.. if not I will have moved it to diffrent hardware/server
10) Message boards : Number crunching : gg@home cpu test - intel vs arm (Message 1602)
Posted 12 days ago by Profile [boinc@poland]goofyx
ps.: Even on DrugDiscovery@Home I don't have that kind of diffrent time against intel/amd.
In that case there is no sense to crunching monkeys cpu on arm... ?
11) Message boards : Number crunching : gg@home cpu test - intel vs arm (Message 1601)
Posted 12 days ago by Profile [boinc@poland]goofyx
I observe strange think even now. On intel cpu WU was crunching in 10 second, and the same WU on ODROID-XU4 on arm was crunched in 45minutes.... maybe it's because ARM does'n support hardware random. What do you thin about it?

I know that ARM is weak instead of intel but this is terrible BIGtime diffrent.
12) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1600)
Posted 12 days ago by Profile [boinc@poland]goofyx
Yes... We kill test server :(

I set up too short WU and too small shmem for feeder <- I read in system logs that first died feeder after it system died.
I will set up bigger shmem and feeder cache today.

ps.: and of course WU will be much longer <- I will try to set it up on about 1hour...
It will be difficult because it's depends on CPU power but 10-15second it definitely to short :D
13) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1591)
Posted 13 days ago by Profile [boinc@poland]goofyx
I have some cores I can through at it.. I added the project and attempted to login but kept getting rejected.

Edit -- Project added.

thx.
I am preparing apps right now and would like to create some test series tonight.
I have to check if my solutions won't die with 100 cores
14) Message boards : Number crunching : Spliting GoofyxGrid - CPU sub-project (Message 1589)
Posted 14 days ago by Profile [boinc@poland]goofyx
ok, I am setting up new project for testing: http://cpu.goofyxgridathome.net.
It will be avaibleable today or tommorow <- I think so.

After I start it will be able for testing purpose.
I have to check somethink I change in original source code of boinc server.
I will need about (in summary) min.200-250 core/thread or more to really test a stability for that enviroment
15) Message boards : Number crunching : Spliting GoofyxGrid ?? (Message 1587)
Posted 24 days ago by Profile [boinc@poland]goofyx
I had considered offering a similar suggestion a couple of weeks ago.
Set up a second CPU only project, call it Monkeys@home.....

If I split it it will be not Monkeys@home, but somethink like that:
- GoofyxGrid.NCI@home
- GoofyxGrid.CPU@home


- GoofyxGrid@home NCI
- GoofyxGrid@home CPU

GoofyxGrid@home - (tm), brand, it is better not to change

ok, good idea.
And server www adress will be:
- nci.goofyxgridathome.net
- cpu.goofyxgridathome.net
That's should be ok <- ofcourse after split I will have write to Willy from boincstats to change name and adres for acctual GoofyxGrid@home <- acctual badges and statistic should be continued in nci.goofyxgridathome.net
16) Message boards : Number crunching : Spliting GoofyxGrid ?? (Message 1585)
Posted 25 days ago by Profile [boinc@poland]goofyx
I had considered offering a similar suggestion a couple of weeks ago.
Set up a second CPU only project, call it Monkeys@home.....

If I split it it will be not Monkeys@home, but somethink like that:
- GoofyxGrid.NCI@home
- GoofyxGrid.CPU@home
17) Message boards : Science : Impact Earth (Message 1583)
Posted 27 days ago by Profile [boinc@poland]goofyx
The beginning is simple <- rest is a black magic :D

We have 6 input params:
1) L0 - size of object

2) Pi - dentisity of object:
a) ice - 500-1500 kg/3
b) rock - 2000-3000 kg/m3
c) iron - 8000 kg/m3

3) V0 - velocity of object

4) θ Theta - angle of hits

5) Pt - target type:
a) solid rock - 2500 kg/m3
b) crystal rock - 2750 kg/m3
c) water - 1000 kg/m3

5c) for Pt = water we need Dw as water deep

6) r distanse from hit
18) Message boards : Number crunching : Monkeys - CPU (Message 1582)
Posted 27 days ago by Profile [boinc@poland]goofyx
Last night I try to change some source code of boinc server (scheduler) to use nci and cpu <- but without success.
After every change I had to repcompile all sources, update server and restart project <- it's why you propably has some problem with communication with project
19) Message boards : Number crunching : Monkeys - CPU (Message 1581)
Posted 27 days ago by Profile [boinc@poland]goofyx

1. 2 servers
http://quakecatcher.net/sensor/
http://quakecatcher.net/continual/

If good undestanding it It 2 separete project with 2 separete database (2 forums, diffrent user ) etc. <- it's very simple to split GoofyxGrid in that way
20) Message boards : Number crunching : What's up with v4? (Message 1580)
Posted 27 days ago by Profile [boinc@poland]goofyx
We have to change the name of the theory from
Infinite monkey theorem

to
Infinite boinc-client with GoofyxGrid apps theorem


:)


Next 20

Main page · Your account · Message boards


Copyright © 2017 Goofyx

Generated 24 Mar 2017, 2:08:02 UTC