• Confluence is currently running with all non-system plugins disabled, some functionality may be missing
Child pages
  • Cooperative rendering. The Merging process
Skip to end of metadata
Go to start of metadata

The Cooperative mode is a special mode that allows selected nodes to work on the same frame individually and to merge the images they have created at the end of the rendering process. You can activate Cooperative mode by selecting it in the Wizard panel when a job is submitted.

While rendering, all involved nodes will render the same frame with a different starting “seed”. They will get almost identical images of the same scene, only the noise distribution is different between them (due the different random seed they used). When the job is finished, the Manager will collect these results and merge them into a single output file, resulting in one cleaner image.

Preview a cooperative job

It is possible to preview Cooperative jobs while rendering by selecting the job tree and pressing the Preview button. The Monitor will send a request to the Manager for merging the contribution of all the nodes and display the resulting image in the Preview window as soon as the Manager collects, merges and sends the image to the Monitor.

Merging Manually

If the network fails or crashes at some point and the merging of MXI files is not completed (you can check the Monitor nodes and manager for error messages), you can manually merge the cooperative MXI files created during the render process. All the Render Nodes save the current MXI file in their temp folder (accessible through File>Open temp folder from each Node). The Manager also stores all the MXI files from the nodes in its temp folder, creating a subfolder for each job so it is easier to find the set of MXI files you are looking for and merge them manually.
In order to make a cooperative render work, each MXI file must have a different starting “seed” so that each render has a slightly different noise pattern. This random seed value is given automatically by the -idcpu command; you should not specify it yourself.

There are two instances where automatic merging can fail:

  • The machine running the Manager crashes: In this case you have to gather all the MXI files from each nodes temp folder and manually merge them. Open Maxwell.exe and go to File>Merge MXI, browse to the folder where you copied all the separate MXI files and select them all. Next Maxwell will ask you where you wish to save the merged MXI.
  • One or more nodes crash, in which case Manager will gather all the MXI files from the nodes that didn't crash and store them in the Managers temp folder: In this case you can manually copy over the partial MXI files from the crashed nodes, to the Managers temp folder, and proceed to manually merge them as described above.

Always check the logs of the Manager and Nodes first of all if something went wrong during a network render. The Monitor logs are the least informative. You can set the level of logging from File>Preferences in Manager/Node




Related content:
Error formatting macro: contentbylabel: com.atlassian.confluence.api.service.exceptions.BadRequestException: Could not parse cql : null