http://www.k-wave.org/forum/topic/extending-checkpoint-functionality Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 23:27:41 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/extending-checkpoint-functionality#post-6146 Mon, 09 Oct 2017 11:53:38 +0000 Jiri Jaros 6146@http://www.k-wave.org/forum/ <p>Hi Sebastian,</p> <p>h5dump -d 't_index' output_data_128_128_128_het_cuboid.h5 | grep '(0,0,0)' | cut -c13-<br /> h5dump -d 'Nt' output_data_128_128_128_het_cuboid.h5 | grep '(0,0,0)' | cut -c13-</p> <p>give you those two numbers. </p> <p>This is a better way than checking the timestamps. </p> <p>Best,<br /> Jiri </p> http://www.k-wave.org/forum/topic/extending-checkpoint-functionality#post-6145 Mon, 09 Oct 2017 11:34:13 +0000 bastlwastl 6145@http://www.k-wave.org/forum/ <p>Hi Jiri,</p> <p>well, that sounds reasonable, thanks for the advice. Still, a bit of bash scripting would be required, h5dump gives me a lot of information that needs to be cut away. So, the time solution is working fine for me.</p> <p>Best regards,<br /> Sebastian </p> http://www.k-wave.org/forum/topic/extending-checkpoint-functionality#post-6144 Mon, 09 Oct 2017 09:20:25 +0000 Jiri Jaros 6144@http://www.k-wave.org/forum/ <p>Hi Sebastian,<br /> There is an easy way to figure out the simulation has completely ended. When you open the output file, you can read datasets called t_index and Nt, where t_index says the time step the simulation was interrupted and Nt says the total number of time steps in the simulation.</p> <p>Thus, if (t_index == Nt) then the simulation has ended<br /> else you need to run another leg.</p> <p>You can parse the HDF5 files from bash scripts using h5dump -d "Nt" ... </p> <p>Best wishes,<br /> Jiri </p> http://www.k-wave.org/forum/topic/extending-checkpoint-functionality#post-6143 Tue, 03 Oct 2017 16:52:36 +0000 bastlwastl 6143@http://www.k-wave.org/forum/ <p>Dear all,</p> <p>I recently (had to) implement a checkpoint version of my K-wave simulation using the C++ binaries. As a small feedback, i would like to say that the manual is a bit weak at this point, at least for me. However, after i figured out the role of the input, output and checkpoint file in this scenario, i do miss some kind of feedback from the binaries telling me that the simulation was ended due to the wall time i defined. This would give the straight forward possibility to rerun from the saved step, if possible. I come up with some Bash magic i'm sharing with you, but i guess it would be nice to have a better solution clearly checking if the simulation is already finished. One clear disadvantage of my solution is that if by coincidence the simulation is finished but needs more time to save data,and thereby taking more time, my script will then assume that it need to run another loop. I guess not a big problem, but 'imperfect' ;)<br /> Anyway, find the script below, reduced to the important part. The required files has to be provided by the command line. </p> <p>Best Regards from Munich,<br /> Sebastian</p> <pre><code>#!/bin/bash -le #Setup files INPUTFILE=#1 OUTPUTFILE=#2 CHECKPOINT_FILE=#3 CPP_BIN_ADD_PARAM=&#39;--p_raw&#39; #one day in seconds CHECKPOINT_TIME=86400 loop_rerun=true while $loop_rerun; do #date +%s will give you the current time in seconds starting at the 1.1.1970 start_time=&lt;code&gt;date +%s&lt;/code&gt; echo Start time in seconds: echo $start_time #Running the Simulation with the c++ binaries kspaceFirstOrder3D-OMP_SSE4 -i $INPUTFILE -o $OUTPUTFILE $CPP_BIN_ADD_PARAM --checkpoint_interval $CHECKPOINT_TIME --checkpoint_file $CHECKPOINT_FILE echo End time in seconds: end_time=&lt;code&gt;date +%s&lt;/code&gt; echo $end_time #calculate the difference difference=&lt;code&gt;expr $end_time - $start_time&lt;/code&gt; echo Assumed Runtime: echo $difference if [ &quot;$difference&quot; -lt &quot;$CHECKPOINT_TIME&quot; ]; then loop_rerun=false echo &quot;End of Checkpoint simulation, finishing&quot; fi done</code></pre>