When I was processing the HCP dataset, I found that the mrconvert command was executed very slowly. By checking the operation of system resources, it is found that the CPU utilization rate is only about 100% when mrconvert is executed. My CPU is i7-12700H, which has 14 cores and 20 threads, so the maximum utilization rate should be 2000%. Is there a multi-thread method to execute mrconvert?
Hi @zzcpage,
All MRtrix commands are already multi-threaded, this should already be using all available cores – assuming this hasn’t been over-ridden by command-line arguments (via the -nthread option), the MRTRIX_NTHREADS environment variable, or the NumberOfThreads config file variable. You can verify how many threads are actually being used with the -debug option, which should report something like this (along with a lot of other information):
mrconvert: [DEBUG] launching 32 threads "loop threads"...
Assuming this isn’t the issue, this wiki article may be relevant?
All the best,
Donald.
1 Like
Thank you very much for your help. I solved the problem I encountered by modifying the dirty page threshold of Linux given by you. As you said, the mrconvert command is multi-threaded. When you execute - debug, it shows that 20 CPU threads are enabled