Merging MXI files

The merge button


The users can merge two or more mxi files together in order to join all the calculation jobs done on different machines into one single image. The purpose of this is to use several machines to compute one single image, so many machines contribute to the same render. This is basically what the network system does when launching a cooperative job: it gives a copy of the same scene (the MXS file) to each rendering machine, calculates which SL needs to be reached on each machine to get the desired SL.

Please, take into account that, when merging the MXI files, you are not adding up the SL but just the render time or the calculations made on different machines into a single image.

So, imagine the case you have 5 machines of the same power; it would be the same running a render on one machine for 5 hours than running the same render on 5 machines for 1 hour and then merge the result into one mxi file.


How is it done?

The process is very simple. It is done from Maxwell Render (light green icon). You can use the merge button from the toolbar or access the command from the File menu > Merge MXI…

Maxwell will first ask you to select all the MXI files you want to merge (for example, sceneA_1.mxi, sceneA_2, sceneA_3), then in a second dialogue, it asks you to choose the name of the resulting merged MXI file (for example, sceneA_merged.mxi). Then it will take some time to be calculated and saved and that would be all.


The same thing can be done for sequences of MXI files. For example, two computers have rendered the same animation sequence and you want to merge the resulting MXI files. In this case, Maxwell expects you to have each sequence in separate folders with the MXI files of each frame named the same, so it will merge the equally named MXI files of the different folders together and obtain a merged sequence.

After that, you can use the MXI Batch Processing tool you can find in the same menu to generate the output image of each sequence.


Of course, you cannot merge any file with any file. The MXI files you want to merge have to come from the same MXS file, show the same camera, have the same resolution, the same channels and the same number of Multilight sliders; the values in the Multilight sliders may differ, though.

One important detail is that the CPU Id has to be different for each MXI file. The CPU Id represents the noise seed and is assigned randomly to each render when it is started. If the CPU Id was the same, Maxwell would have calculated the same noise pattern and the same light rays, so the result wouldn’t be complementary.



Calculating the number of identical nodes needed to get a particular final SL.

Most of the times you won’t need to calculate anything or Maxwell will do it for you in a cooperative network job, but maybe this calculation is useful at some point; it lets you know the SL required for each node to get a particular final SL when merged, provided you know the number of nodes involved and that all the nodes have the same computing power:

SL_NODE = ( log(((exp(SL_TOTAL * log (1.5)) - 1.0 ) / nNodes ) + 1.0 )) / log(1.5)

log = neperian logarithm
exp = e^x

Which is the same as this in a more visual way: