NDMETIS(1)

πŸ‘‘ NDMETIS πŸ‘‘

okay so bestie, you know how sometimes a graph is just... too much? like the fill-in is giving chaos demon and your solver is crying? πŸ’€ that's literally where ndmetis comes in. it's a User Command, it lives in the NDMETIS(1) manual, and what it does is compute a fill-reducing ordering of the vertices of the graph using multilevel nested dissection. iconic behavior honestly. ✨


🌸 SYNOPSIS 🌸

ndmetis [options] <filename>

πŸ’– DESCRIPTION πŸ’–

πŸŽ€ Required parameters πŸŽ€

  • filename β€” this is where the graph to be partitioned is stored. yeah. the whole point. you gotta give it the graph, bestie, no graph no party. 🫢

πŸ¦‹ Optional parameters πŸ¦‹

  • -ctype=string β€” picks the scheme used to match the vertices of the graph during the coarsening. your choices are:
    • rm β€” Random matching 🎲
    • shem β€” Sorted heavy-edge matching β€” this is the default, she's the moment πŸ’…
  • -iptype=string β€” applies only when -ptype=rb, just so you know. this one specifies the scheme used to compute the initial bisection of the graph. options are:
    • edge β€” Separator from an edge cut βœ‚οΈ
    • node β€” Separator from a greedy node-based strategy β€” the default, she ate πŸ”₯
  • -rtype=string β€” the scheme used for refinement. pick your fighter:
    • 1sided β€” 1-sided node-based refinement β€” the default 🌟
    • 2sided β€” 2-sided node-based refinement πŸ’«
  • -ufactor=int β€” the maximum allowed load imbalance between the left and right partitions during each bisection. the load imbalance is measured as the ratio of the 2*max(left,right)/(left+right), where left and right are the sizes of the respective partitions. a value of x means the allowed load imbalance is 1+x/1000. default is 200, which indicates a load imbalance of 1.20. so like, if you want it more balanced, lower it. balance queen πŸ‘‘
  • -pfactor=int β€” the minimum degree of the vertices that will be ordered last. if the value you give is x>0, then any vertices with a degree greater than 0.1*x*(average degree) get removed from the graph, an ordering of the rest of the vertices is computed, and then an overall ordering is computed by ordering the removed vertices at the end of the overall ordering. default value is 0, which means no vertices are removed at all, they all stay, nobody gets evicted 🏠
  • -no2hop β€” says that the coarsening will not perform any 2-hop matchings when the standard matching fails to sufficiently contract the graph. no extra hops, we're keeping it simple 🚫
  • -nocompress β€” says the graph should not be compressed by combining together vertices that have identical adjacency lists. twins stay separate, periodt πŸ’―
  • -ccorder β€” says whether the connected components of the graph should first be identified and ordered separately. do them one at a time, main character energy for each 🫧
  • -niter=int β€” the maximum number of iterations for the refinement algorithms at each stage of the uncoarsening process. default is 10 πŸ”
  • -nseps=int β€” the number of different separators that it will compute at each level of the nested dissection. the final separator that gets used is the smallest one. default is 1. shop around, pick the bestie that's smallest πŸ›οΈ
  • -nooutput β€” says that no ordering file should be generated. sometimes you just want the vibes, not the file πŸ™ˆ
  • -seed=int β€” selects the seed of the random number generator. for when you need your chaos to be reproducible chaos 🌱
  • -dbglvl=int β€” selects the dbglvl. that's it, that's the description, i'm not gonna lie to you bestie, i'm just relaying it fr fr πŸ”
  • -help β€” prints this message. your little rescue button when you forget everything πŸ†˜

⚠️ heads up bestie: -iptype only does anything when -ptype=rb. if you're not using rb, she's just sitting there, pretty but unemployed. πŸ’€

🌈 that's the whole page 🌈

this page is from ndmetis version 5.1.0.3, dated July 2025, and it's a User Commands page for NDMETIS(1). go order some vertices, bestie, you're gonna slay πŸŽ€πŸ’–βœ¨πŸ¦‹πŸ’…πŸ‘‘

πŸ“¦ where this page comes from