Problem with moving window

[Fettpet]

I used the current dev branch 5e781594a02322f855eb181fbd573b623caa8279. The modules are

• gcc/4.9.2
• cuda/8.0
• cmake/3.7.2
• boost/1.62.0
• openmpi/1.8.6.kepler.cuda80
• pngwriter/0.7.0
• hdf5-parallel/1.8.15
• libsplash/1.6.0

I follow the basics guideline. The only difference was the usage of /etc/picongpu/0008gpus.cfg instead of /etc/picongpu/0016gpus.cfg. The problem with this configuration is, the outputted pngs doesn't look right. The matter disappear to the left of the image. If I remove the option -m in TBG_movingWindow the output looks like it should. I think there is a bug in the moving window.

[PrometheusPi]

@Fettpet thanks for the report. Could you please post the two png outputs here, so we can directly see the results you mentioned?

I suspect that the default setup for 8 GPUs might be a bit too short to keep the entire first plasma cavity when used with moving window. I will check this.

[Fettpet]

I created two videos. The first use moving window. In the second one, I remove the sliding window option. By the way, I didn't tested other default settings, like 8, 16 or 32 GPUs.

[PrometheusPi]

The default length of the 8 GPU example is too short for a moving window simulation. The window starts sliding before the laser entered the simulation box completely. Thanks for pointing this out. I will fix the setup.

[PrometheusPi]

@Fettpet Does the pull request resolve your issue?