𝕏

Optimized app

Message boards : Number crunching : Optimized app
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1133 - Posted: 18 Jan 2017, 9:26:29 UTC

New app is "heavy" (over 4h in modern cpu).
Have you scheduled an optimized app??
ID: 1133 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile jcastro
Avatar

Send message
Joined: 16 Mar 15
Posts: 219
Credit: 14,859
RAC: 0
Message 1135 - Posted: 19 Jan 2017, 9:35:40 UTC

Hi!

New simulations are a bit longer. Also this new version calculates markers which could cost just a bit more.

We are working on the database and the optimization of the app is on our horizont but is not our priority. The part of the code which could be highly optimizable is the one that comes from the model, so we have to modify the tool we use to traduce CellML to C++ so there is not a trivial work.

Best regards, Joel.
PD: The code is open for everyone. If you know how to compile with a better compiler or with SSE additions you can do it! :)
ID: 1135 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1136 - Posted: 19 Jan 2017, 11:07:48 UTC - in response to Message 1135.  

PD: The code is open for everyone. If you know how to compile with a better compiler or with SSE additions you can do it! :)


Thank you!
Which is the app on github for "DENIS Stable version used to fill the Markers Database"? CellML2C??
ID: 1136 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile jcastro
Avatar

Send message
Joined: 16 Mar 15
Posts: 219
Credit: 14,859
RAC: 0
Message 1138 - Posted: 20 Jan 2017, 21:06:51 UTC - in response to Message 1136.  

The stable version of denis simulator is here: https://github.com/DENISproject/denis-boinc-baseapp
ID: 1138 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1170 - Posted: 24 Feb 2017, 12:58:06 UTC

No one want to try to compile an optimized app??
ID: 1170 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r

Send message
Joined: 9 Apr 15
Posts: 16
Credit: 20,823,669
RAC: 0
Message 1178 - Posted: 1 Apr 2017, 18:45:53 UTC - in response to Message 1170.  
Last modified: 1 Apr 2017, 19:32:57 UTC

No one want to try to compile an optimized app??


It's not just to grab the source code and "compile" and optimized app... there's are a 'few' things to be considered here (◔_◔) ...

Analyzing the code (models),rewriting the code and checking results...That's what needs to be done. It took me 2 days to get a working model for CRLP2011_EPI only.

You sound very ungrateful/disrespectful , to be quite frankly, demanding someone compile an optimized app for you.


However, i've got a new app that's ~10 to 14 times as fast as the current stock app.

Anyway, you're welcome to put in the work yourself and come up with a optimized app yourself.

The source code of the stock app is available at: https://github.com/DENISproject/denis-boinc-baseapp

I'm waiting for your apps to be released, till i make mine available for download...let's see what happens.
t(ಠ益ಠt)

meanwhile... i'll give you something to look at/compare your apps to...
i5-4200H -> http://denis.usj.es/denisathome/results.php?hostid=5205

Can't wait to see your apps compared to mine.(i'll give you a month).
ID: 1178 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
mmonnin

Send message
Joined: 18 Nov 16
Posts: 14
Credit: 10,059,426
RAC: 3,853
Message 1181 - Posted: 2 Apr 2017, 16:47:14 UTC - in response to Message 1178.  

[quote]You sound very ungrateful/disrespectful , to be quite frankly, demanding someone compile an optimized app for you.


Reading the last several posts here it seems like the pot is calling the kettle black.
ID: 1181 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1182 - Posted: 3 Apr 2017, 6:19:33 UTC - in response to Message 1178.  

You sound very ungrateful/disrespectful , to be quite frankly, demanding someone compile an optimized app for you.


Thank you.
You are not obliged to answer to my little question, but if you feel the duty to insult me, you did well.
ID: 1182 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1183 - Posted: 3 Apr 2017, 6:30:36 UTC - in response to Message 1181.  

Reading the last several posts here it seems like the pot is calling the kettle black.


Thank you too.
I hope your days are better after your posts.
ID: 1183 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BlackHeart64

Send message
Joined: 12 Jul 17
Posts: 3
Credit: 0
RAC: 0
Message 1185 - Posted: 3 Apr 2017, 12:29:46 UTC - in response to Message 1178.  

No one want to try to compile an optimized app??


It's not just to grab the source code and "compile" and optimized app... there's are a 'few' things to be considered here (◔_◔) ...

Analyzing the code (models),rewriting the code and checking results...That's what needs to be done. It took me 2 days to get a working model for CRLP2011_EPI only.

You sound very ungrateful/disrespectful , to be quite frankly, demanding someone compile an optimized app for you.


However, i've got a new app that's ~10 to 14 times as fast as the current stock app.

Anyway, you're welcome to put in the work yourself and come up with a optimized app yourself.

The source code of the stock app is available at: https://github.com/DENISproject/denis-boinc-baseapp

I'm waiting for your apps to be released, till i make mine available for download...let's see what happens.
t(ಠ益ಠt)

meanwhile... i'll give you something to look at/compare your apps to...
i5-4200H -> http://denis.usj.es/denisathome/results.php?hostid=5205

Can't wait to see your apps compared to mine.(i'll give you a month).


This is very interesting, I look forward to the release of the optimized app(if you do choose to release)
I've noticed with different CPUs, there is def a difference in processing time.

My i7 3770k beats my X5670 with completing task.
ID: 1185 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BlackHeart64

Send message
Joined: 12 Jul 17
Posts: 3
Credit: 0
RAC: 0
Message 1186 - Posted: 3 Apr 2017, 16:21:16 UTC - in response to Message 1185.  

I've noticed my task seem to complete the fastest on OSX... on older hardware. 1 Hour vs 4 hour.
ID: 1186 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
mmonnin

Send message
Joined: 18 Nov 16
Posts: 14
Credit: 10,059,426
RAC: 3,853
Message 1187 - Posted: 3 Apr 2017, 22:48:12 UTC - in response to Message 1183.  

Reading the last several posts here it seems like the pot is calling the kettle black.


Thank you too.
I hope your days are better after your posts.


I was agreeing with you...
ID: 1187 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile jcastro
Avatar

Send message
Joined: 16 Mar 15
Posts: 219
Credit: 14,859
RAC: 0
Message 1189 - Posted: 4 Apr 2017, 8:21:36 UTC

We know that optimize our application is really hard, because the main operations are inside the models. So when we add a new model to the project, the optimization have to be done on each separate model.

This is why we have try to optimize the way we convert from CellML to C, we have modify few things from the code of the last optimization and we are looking different approaches to make the app faster.

Also, in the other hand, we have to develop software as much stable as possible and it can be executed also in old machines, so we have to use old versions of compilers and libraries.

In addition I will thank a lot to the people that have released their optimized versions, because we have learned a lot reading his code.

Best regards, Joel.
ID: 1189 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Tom_unoduetre

Send message
Joined: 9 Apr 15
Posts: 3
Credit: 55,994
RAC: 0
Message 1269 - Posted: 29 Jun 2017, 8:50:07 UTC

Hi jcastro, please make the optimized application available to everyone.

If Crunch3r´s numbers are correct and an optimization can bring up to 10-14 times better speed it is something significant and should be made available to the whole DENIS community.
ID: 1269 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Crunch3r

Send message
Joined: 9 Apr 15
Posts: 16
Credit: 20,823,669
RAC: 0
Message 1287 - Posted: 7 Jul 2017, 21:49:51 UTC - in response to Message 1269.  
Last modified: 7 Jul 2017, 21:52:01 UTC

Hi jcastro, please make the optimized application available to everyone.

If Crunch3r´s numbers are correct and an optimization can bring up to 10-14 times better speed it is something significant and should be made available to the whole DENIS community.


It will be at least from my perspective.

However, there are still quite a few "stock" models that need to be "fixed".

Anyway, i'll be happy to upload the code for the ones that i've modified allready.

There's just one minor thing that i'd like to adress... I'd like to make sure that the "optimized" code is some sort of dedication (mentioning his name) to my dad who died recently on 17.05.2017 of a stoke.

Btw, the stock apps for windows and osx do not deliver the same results.
ID: 1287 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Beyond

Send message
Joined: 28 Jun 17
Posts: 1
Credit: 0
RAC: 0
Message 1288 - Posted: 7 Jul 2017, 22:44:53 UTC - in response to Message 1287.  

I'd like to make sure that the "optimized" code is some sort of dedication (mentioning his name) to my dad who died recently on 17.05.2017 of a stoke.

My condolences Crunch3r. Never an easy thing to deal with.
ID: 1288 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Tom_unoduetre

Send message
Joined: 9 Apr 15
Posts: 3
Credit: 55,994
RAC: 0
Message 1290 - Posted: 11 Jul 2017, 8:07:55 UTC - in response to Message 1287.  

I'd like to make sure that the "optimized" code is some sort of dedication (mentioning his name) to my dad who died recently on 17.05.2017 of a stoke.

Oh, I´m very sorry to hear that, so sorry for your loss :-( I wish you all the strength to get passed this bad time.
ID: 1290 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
[VENETO] boboviz

Send message
Joined: 9 Apr 15
Posts: 171
Credit: 1,371,098
RAC: 1,159
Message 1294 - Posted: 17 Jul 2017, 19:33:09 UTC - in response to Message 1287.  

I'd like to make sure that the "optimized" code is some sort of dedication (mentioning his name) to my dad who died recently on 17.05.2017 of a stoke.


Also, my condolences
ID: 1294 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Tex1954

Send message
Joined: 7 Jul 15
Posts: 28
Credit: 29,892,124
RAC: 23,958
Message 1379 - Posted: 23 Mar 2018, 18:58:25 UTC - in response to Message 1294.  

Likewise, my condolences...

Also, of particular note, someone mentioned OSX and Winderz results are different? I hope someone knows what the significance of the difference is...

In any case, FWIW, the tasks seem to average:

3.3-Hrs on the Linux 61XX (Currently AMD 6180) setup
3.85-Hrs on the Linux E3-1240V2 setup
2-Hrs on the Winderz 5930K setup
2.75-Hrs on Winderz 1090T setup

Anyways, seems the code is longer sometimes, hard to say at this point. So far, only "Download" errors seem to be popping up...

Good luck to all!

8-)
ID: 1379 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
mmonnin

Send message
Joined: 18 Nov 16
Posts: 14
Credit: 10,059,426
RAC: 3,853
Message 1383 - Posted: 23 Mar 2018, 19:58:56 UTC - in response to Message 1379.  
Last modified: 23 Mar 2018, 19:59:18 UTC

Likewise, my condolences...

Also, of particular note, someone mentioned OSX and Winderz results are different? I hope someone knows what the significance of the difference is...

In any case, FWIW, the tasks seem to average:

3.3-Hrs on the Linux 61XX (Currently AMD 6180) setup
3.85-Hrs on the Linux E3-1240V2 setup
2-Hrs on the Winderz 5930K setup
2.75-Hrs on Winderz 1090T setup

Anyways, seems the code is longer sometimes, hard to say at this point. So far, only "Download" errors seem to be popping up...

Good luck to all!

8-)


New app now but I can add some comparisons to that for the new app.
1:45 on 3570k @3.8
2:00 on 1950k @3.75
ID: 1383 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
1 · 2 · Next

Message boards : Number crunching : Optimized app