Writing multiple output files based on number of threads then concatenating them

[jelber2]

Hi,
I am assembling the huge axolotl genome (estimated between 21-48GBp) from about 2 terabytes of Illumina 100bp paired-end reads on a node with 1.5TB RAM and 48 cores. I am noticing a load of 1 and CPU usage of 100% (instead of 4800%) for at least a day now as LightAssembler slowly writes the output to the contigs file. I looked at the code, and it appears that the output is written to the contigs file within lines 389-424 of GraphTraversal.cpp. I was wondering if instead of writing the output to a single contigs file in a serial fashion, if it would be possible to write the output to multiple temporary contigs files (as many as there are threads), concatenating the temporary contigs files into a single output file, and deleting the temporary contigs files. This seems like it would be a performance increase, granted I don't know the code well enough to see if this can be implemented.