General

General

How profile data is handled in parallel?

    • FAQFAQ
      Participant

      The host reads the profile, and then this profile data is copied to each of compute cores. Since each core holds a copy, memory requirement per node will be equal to the number of cores on the node times the memory taken by the profile. For large profiles, this could pose a memory problem. Sometimes profile is written from outside Fluent. Such files may be large. In these cases, please read it serial and initialize the case in Fluent. Then write profile file from Fluent. This profile file is much smaller which you can use in Fluent parallel. Please also see solution # #2039888 for reorienting profile. Keywords: profile, parallel