In GitLab by Keith Bennett (@keithbennett) on 18 Nov 2014, 00:00 UTC (GitLab issue GL#1190)
In the current implementation, each process reads an arbitrary section of the file and then redistributes the data to the correct processor. This typically doubles the amount of memory required. A better approach would be for each processor to read in its own data section from file.
In GitLab by Keith Bennett (@keithbennett) on 18 Nov 2014, 00:00 UTC (GitLab issue GL#1190)
In the current implementation, each process reads an arbitrary section of the file and then redistributes the data to the correct processor. This typically doubles the amount of memory required. A better approach would be for each processor to read in its own data section from file.