| log in |
Message boards : News : latest change to ll-dc app
1 · 2 · Next
| Author | Message |
|---|---|
|
We have removed the fixed input -v -use NO_ASM in latest WU batch --passthrough_child -use NO_ASM in app_config.xml to run this on your GPU. For other options like -log xxx you need to change that to --passthrough_child -log xxx //xxx for any value see FAQ to avoid computation errors. | |
| ID: 11982 · Rating: 0 · rate:
| |
|
Is this change for Linux only ? | |
| ID: 11986 · Rating: 0 · rate:
| |
Is this change for Linux only ? For both OS. | |
| ID: 11987 · Rating: 0 · rate:
| |
|
Why did you remove cmdline -log. I was running another project, came back to this and had errors because you broke a function that was OK and now does not. | |
| ID: 11992 · Rating: 0 · rate:
| |
|
To be sure:
<app_version>
<app_name>LL-dc</app_name>
<cmdline>--passthrough_child -use NO_ASM</cmdline>
</app_version>
is that correct? | |
| ID: 11999 · Rating: 0 · rate:
| |
To be sure: All of this needs to be in the app_config.xml file <app_config>
<app_version>
<app_name>LL-dc</app_name>
<plan_class>opencl_nvidia_200</plan_class>
<cmdline>--passthrough_child -use -NO_ASM</cmdline>
</app_version>
</app_config> | |
| ID: 12000 · Rating: 0 · rate:
| |
To be sure: There was a change in the latest boinc wrapper command as discussed with boinc dev. You need the new --passthrough_child command to send the command line to boinc client which does not have the older wrapper. We had this included in input template but removed it. | |
| ID: 12001 · Rating: 0 · rate:
| |
To be sure: Just found out because of errors: the line should read <cmdline>--passthrough_child -use NO_ASM</cmdline> without the "-" in front of the NO_ASM. Just to inform everyone who might need that too (i cannot afford a new gfx card, and my 1080ti does still work smoothly) | |
| ID: 12010 · Rating: 0 · rate:
| |
|
Thx for testing, the FAQ had a mistake with "-". I have corrected it. | |
| ID: 12011 · Rating: 0 · rate:
| |
|
For low Tech volunteers where is the app_config.xml file located ? | |
| ID: 12012 · Rating: 0 · rate:
| |
For low Tech volunteers where is the app_config.xml file located ? boinc\data\projects\srbase.my-firewall.org_sr5 If there is none you must create an app_config.xml | |
| ID: 12013 · Rating: 0 · rate:
| |
For low Tech volunteers where is the app_config.xml file located ? This is correct: C:\ProgramData\BOINC\projects\srbase.my-firewall.org_sr5 | |
| ID: 12023 · Rating: 0 · rate:
| |
|
The latest app is using FFT 4M and doubled the runtime, the change is running on FFT 2M -fft 4:1K:4:256 Update: Here is a sample after run -tune (but crashed on latest app on AMD) from a RX5500XT, before 68h and after tuning 55h, bug reported to dev | |
| ID: 12118 · Rating: 0 · rate:
| |
|
I have noticed that since app version 0.12 went in my runtimes for a Nvidia 5060ti went from approx 15 hours to 18 hours and the ETA for the currently running task is 21.45 hours. I saw your post from today but I'm not understanding what you are telling me. Did they change the FFT size with the latest version? The statement: "The latest app is using FFT 4M and doubled the runtime, the change is running FFT 2M" What is that telling me? Isn't the latest app and change the same thing? I'm totally confused. | |
| ID: 12120 · Rating: 0 · rate:
| |
|
The standard setting was using FFT 2M now 4M. With the change you will get back your old runtimes. Also you can run -tune see FAQ to get it faster. It is similar to prst, higher FFT need more time to finish a task. I will ask the dev about the change to be sure. | |
| ID: 12121 · Rating: 0 · rate:
| |
|
So are you saying there is going to be a 0.13 version that will change the FFT back to 2M? | |
| ID: 12122 · Rating: 0 · rate:
| |
So are you saying there is going to be a 0.13 version that will change the FFT back to 2M? no, it is standard but can be manually changed. The dev is informed. Maybe we can fix it next time with the older wrapper and input file or a better app. | |
| ID: 12123 · Rating: 0 · rate:
| |
|
I read the FAQ concerning the app_config update and the option to run -tune. What is the process to run -tune? What is meant by a "standalone environment"? | |
| ID: 12124 · Rating: 0 · rate:
| |
I read the FAQ concerning the app_config update and the option to run -tune. What is the process to run -tune? What is meant by a "standalone environment"? You must run outside BOINC in a separate folder. You will find a tune.txt and a config.txt file. In the config.txt you can see a line # New settings based on a -tune run. -use LOADS=1000,STORES=1,MODM31=2 //sample which you need to copy into your app_config.xml. The result is a faster runtime, an user tested it on different cards with around 20-40%, I got around 37% faster after tuning. You need to do it only one time for one card. | |
| ID: 12125 · Rating: 0 · rate:
| |
|
Did they change to FFT 4M due to FFT 2M not being big enough? Would you happen to know at what point the FFT 2M will stop working? | |
| ID: 12126 · Rating: 0 · rate:
| |
Message boards :
News :
latest change to ll-dc app