CPU/SMP 高獎勵得分 選購/設定/除錯...集中綜合討論串

[Raziel]

忽然驚醒夢中人.....
現在即時已將-advmethods參數全刪以防「中招」~~

怎麼說咧...能開釋一下嗎?

 

[A4GTRS]

處女首試P6903 Bigbeta暫時配置和設定只屬初階，三日後若能完成並交出成績便會
實行進一步計劃，積極一點希望能趕及年終前再完成多一兩個此類WU。

SB-E i7-3930K 4257MHz (43x99) : 因43x100時偶爾會由43跳回32效能不穩
CPU power phase control: 此板最高級別為 Level 5，其他Voltage全Auto
Thermalright Truecopper with 2600prm 12 cm fan : 溫度23-62°C
2 x DDR3-2000 @ 1600 CL8-8-8-16 : 只跑Dual CH，因Triple只能CL9更慢

Ubuntu 10.10 Desktop 64-bit
Corefix.sh not running
-smp 12 -bigadv

3930K 看來後勁很強
穩跑4.8G~5.0G, P6903要衝到TPF 33mins.以內猶如桌上拿柑 ;)

 

[xs910203]

怎麼說咧...能開釋一下嗎?

怕收到新雷U吧

 

[OverFold]

照目前形勢來看，在Celeron G530雙核芯跑SMP2若加上-advmethods參數
必然會get到那些TPF1小時以上卻只有1400PPD的P10xxx WU，若去掉重跑
必然會抓到P6944、P7150甚至P11xxx那類較窩心的WU回復2500-3000PPD了~~

 

[Abcb]

前幾天就算不加-advmethods旗標, 也會收到P10XXX這個難搞的project....

因為我的機器都不加-advmethods旗標的, 前幾天幾乎都收到這project=_=

而且不只是SMP, 連單CPU client也照樣中獎, PDD只剩下幾百XD

 

[Raziel]

怕收到新雷U吧

照目前形勢來看，在Celeron G530雙核芯跑SMP2若加上-advmethods參數
必然會get到那些TPF1小時以上卻只有1400PPD的P10xxx WU，若去掉重跑
必然會抓到P6944、P7150甚至P11xxx那類較窩心的WU回復2500-3000PPD了~~

真的這麼厲害喔.....不過, 我不算雷U, 誰來算雷U.....還是平常心繼續擺著. 多謝指導了.

 

[kaoh08]

-advmethods flag 到底是什麼意思還是搞不太清楚，官網是寫進階的演算法?

如果是對整體運算有貢獻的就算PPD低一點也無妨。

 

[OverFold]

好不容易才等到i7-3930K@4257MHz跑完6903交功課的日子....
4天了...偏偏在今早10:00這個時候Server就連不上

[01:42:20] Unit 1 finished with 77 percent of time to deadline remaining.
[01:42:20] Updated performance fraction: 0.766979
[01:42:20] Sending work to server

[01:42:20] Project: 6903 (Run 0, Clone 3, Gen 37)

[01:42:20] + Attempting to send results [December 27 01:42:20 UTC]
[01:42:20] - Reading file work/wuresults_01.dat from core
[01:42:20] Connecting to http://130.237.232.237:8080/
[01:42:21] - Couldn't send HTTP request to server
[01:42:21] + Could not connect to Work Server (results)

[01:42:21] + Retrying using alternative port
[01:42:21] Connecting to http://130.237.232.237:80/
[01:42:21] - Couldn't send HTTP request to server
[01:42:21] + Could not connect to Work Server (results)
[01:42:21] - Error: Could not transmit unit 01 (completed December 27) to work server.
[01:42:21] - 1 failed uploads of this unit.
[
[01:42:21] - 2 failed uploads of this unit.

[01:42:21] + Attempting to send results [December 27 01:42:21 UTC]
[01:42:21] - Reading file work/wuresults_01.dat from core
[01:42:21] (Read 222416106 bytes from disk)
[01:42:21] Connecting to http://130.237.165.141:8080/
[01:42:21] - Couldn't send HTTP request to server
[01:42:21] + Could not connect to Work Server (results)

[01:42:21] + Retrying using alternative port
[01:42:21] Connecting to http://130.237.165.141:80/
[01:42:21] - Couldn't send HTTP request to server
[01:42:21] + Could not connect to Work Server (results)

[01:42:21] Could not transmit unit 01 to Collection server; keeping in queue.
[01:42:21] + Sent 0 of 1 completed units to the server

[01:42:21] - Preparing to get new work unit...

[01:42:21] Cleaning up work directory ;smash; (不是說keeping in queue嗎？)

[01:42:21] + Attempting to get work packet ;sorry;

[01:44:09] - 3 failed uploads of this unit.

[01:44:09] Working on queue slot 02 [December 27 01:44:09 UTC]


自動重試三次後就把我的P6903 wuresults_01所有檔案從work folder去掉
並抓到P6904進行新的queue slot 02工作 (順帶一提P6904 TPF為53:13s)。

[01:44:13] Project: 6904 (Run 0, Clone 27, Gen 40)
[01:44:20] Mapping NT from 12 to 12

[01:44:25] Completed 0 out of 250000 steps (0%)
[02:38:44] Completed 2500 out of 250000 steps (1%)
[03:31:57] Completed 5000 out of 250000 steps (2%)

過去亦曾出現過類似情況所以我有預先Backup以防萬一，在該WU完成95%時我有把整個FAH
Folder備份，只是不清楚如果我把當前的6904移除並把先前未完成的6903補上，是否能繼續
先前的進度呢？我計算過目前仍未過Preferred Deadline，以TPF為40分鐘左右的速度絕對可
完成多一個Bigbeta，未知各位師兄有無遇過類似情況需要動用備份這種非常的補救方法呢？

 

[OverFold]

補救方法似乎不行，始終說Connect不到Server，忽發奇想...我把Client.cfg內的Proxy=yes
設為no，並把Terminal profile langouste內command由8880改回8080，出現以下結果....

[09:45:35] Trying to send all finished work units

[09:45:35] Project: 6903 (Run 0, Clone 3, Gen 37)
[09:45:35] + Attempting to get work packet
[09:45:35] + Attempting to send results [December 27 09:45:35 UTC]
[09:45:35] - Reading file work/wuresults_01.dat from core
[09:45:38] (Read 222413385 bytes from disk)
[09:45:38] - Connecting to assignment server
[09:45:38] Connecting to http://130.237.232.237:8080/
[09:45:38] Connecting to http://assign.stanford.edu:8080/
[09:45:39] Initial: ED82; - Successful: assigned to (130.237.232.237).
[09:45:39] Loaded queue successfully.
[09:45:39] Sent data
[09:45:39] Connecting to http://130.237.232.237:8080/
[09:45:53] Initial: 0000; - Receiving payload (expected size: 57247471)
[09:48:09] Initial: 0000; - Uploaded at ~1438 kB/s
[09:48:09] - Averaged speed for that direction ~1438 kB/s

[09:48:09] - Server has already received unit.

[09:48:09] + Sent 0 of 1 completed units to the server

[09:48:09] - Autosend completed

[09:51:37] - Downloaded at ~162 kB/s
[09:51:37] - Averaged speed for that direction ~358 kB/s
[09:51:37] + Received work.

之後另一個6903又開始跑了，再看回Stanford官方Folding Stat(那EOC是日休業...)，
又好似由11116xxxx跳到11152xxxx，是增加了300,000績分，似乎是成功了！;0a155537;

但是這卻又令我產生另一個疑惑了...如此一來...那麼langouste.sh豈不是被廢了武功，Proxy
一定要行8880這個Port嗎？若然將之完全停用會否產生其他反效果呢？langouste除了是負責
Send & Receive同時進行之外，她是否只擔當成中介人的角色而已？Thanks.

 

[labaking]

忽然驚醒夢中人.....
現在即時已將-advmethods參數全刪以防「中招」~~

不知道您指的「中招」為何？？
這個參數不是產生詳細的LOG檔嗎？