can't open dependency file

wblipp · 2010-10-07, 22:32

Pace's post processing failed with the following log file tail. What should we do next?

Code:

Thu Sep 23 19:34:33 2010  filtering completed in 2 passes
Thu Sep 23 19:34:36 2010  matrix is 13020423 x 13020600 (3907.0 MB) with weight 1133973655 (87.09/col)
Thu Sep 23 19:34:36 2010  sparse part has weight 880966222 (67.66/col)
Thu Sep 23 19:35:36 2010  read 13020600 cycles
Thu Sep 23 19:36:11 2010  matrix is 13020423 x 13020600 (3907.0 MB) with weight 1133973655 (87.09/col)
Thu Sep 23 19:36:11 2010  sparse part has weight 880966222 (67.66/col)
Thu Sep 23 19:36:11 2010  saving the first 48 matrix rows for later
Thu Sep 23 19:36:16 2010  matrix is 13020375 x 13020600 (3718.8 MB) with weight 907549158 (69.70/col)
Thu Sep 23 19:36:17 2010  sparse part has weight 844649260 (64.87/col)
Thu Sep 23 19:36:17 2010  matrix includes 64 packed rows
Thu Sep 23 19:36:17 2010  using block size 10922 for processor cache size 256 kB
Thu Sep 23 19:40:38 2010  commencing Lanczos iteration (4 threads)
Thu Sep 23 19:40:38 2010  memory use: 4135.9 MB
Thu Sep 23 19:41:43 2010  linear algebra at 0.0%, ETA 285h 8m
Mon Oct  4 02:32:45 2010  lanczos halted after 165911 iterations (dim = 10492215)
Mon Oct  4 02:32:45 2010  BLanczosTime: 890449
Mon Oct  4 02:32:45 2010  
Mon Oct  4 02:32:45 2010  commencing square root phase
Mon Oct  4 02:32:45 2010  reading relations for dependency 1
Mon Oct  4 02:32:45 2010  error: read_cycles can't open dependency file

Batalov · 2010-10-07, 23:41

You may want to restart with -ncr switch, because it is not done yet
(dim = 10492215 but should be around 13020375 to finish).
First check if there's a *.chk file in that directory...
In any case, back up the data before continuing.

jasonp · 2010-10-08, 01:09

What processor was this running on? The cache size reported is quite small, so maybe there's an L3 cache that was not detected properly. I have to fix the code every year or so to recognize new cache codes. Getting the size right can double the performance.

pinhodecarlos · 2016-11-24, 23:42

Quote:

Originally Posted by Batalov

You may want to restart with -ncr switch, because it is not done yet
(dim = 10492215 but should be around 13020375 to finish).
First check if there's a *.chk file in that directory...
In any case, back up the data before continuing.

Same issue here, trying your approach.

pinhodecarlos · 2016-11-25, 00:14

Quote:

Originally Posted by pinhodecarlos

Same issue here, trying your approach.

Still with problems.

Batalov · 2016-11-25, 01:21

If there are no dependencies, there will be no .dep file written.
Last time, there was an early abort (which is likely when CPU+memory+m/b memory lanes are not flawless.)

In your new case, the reason is different: lanczos halt was at the proper "end of the road" but you are getting only trivial deps. And this will not change by restarting from .chk file.

2010-10-07, 23:41	#2
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 10,061 Posts	You may want to restart with -ncr switch, because it is not done yet (dim = 10492215 but should be around 13020375 to finish). First check if there's a .chk file in that directory... In any case, back up the data before continuing. Last fiddled with by Batalov on 2010-10-07 at 23:43 Reason: paste error*

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
P95 - Windows 7 - Unable to open log file	ThomRuley	PrimeNet	2	2017-01-15 20:43
Unable to open SPOOL file	petrw1	PrimeNet	4	2015-11-24 16:06
Unable to open log file	Unregistered	Information & Answers	2	2011-06-07 16:15
Cannot open spool and log file	lidocorc	Software	4	2008-12-09 05:27
Can't open log file...	Bouchehog	Software	1	2002-10-26 22:52

2010-10-08, 01:09	#3
jasonp Tribal Bullet Oct 2004 3²·5·79 Posts	What processor was this running on? The cache size reported is quite small, so maybe there's an L3 cache that was not detected properly. I have to fix the code every year or so to recognize new cache codes. Getting the size right can double the performance.

2016-11-25, 01:21	#6
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 10011101001101₂ Posts	If there are no dependencies, there will be no .dep file written. Last time, there was an early abort (which is likely when CPU+memory+m/b memory lanes are not flawless.) In your new case, the reason is different: lanczos halt was at the proper "end of the road" but you are getting only trivial deps. And this will not change by restarting from .chk file.