Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Definition of possible run types #10

Open
makeclean opened this issue Apr 9, 2013 · 4 comments
Open

Definition of possible run types #10

makeclean opened this issue Apr 9, 2013 · 4 comments

Comments

@makeclean
Copy link
Contributor

There are a number of different run types that can be defined:

  • Serial: A normal serial calculation, jobs are simply spawned to the batch system and are run when computation nodes have availability. Could be used for parameters scans to vary input data.
  • HTC: High Throughput Computation, a normally serial calculation is split into a number of jobs. This is done by submitting the same input file N times, with each file running NPS/N particles. In maintain the RN sequence, the each run is seeded with the same RN seed, but each run starts on history such that the RN sequence isnt used twice
  • HPC: High Performance Computing, a standard mode of operation for most HPC cluster systems. N cpu's work collectively on the same problem, communicating initialisation and output data using MPI (or OMP).
  • HTPC: High Throughput Parallel Computation, like HPC but provisioned through an HTCondor scheduler on an opportunistic resource rather than a dedicated scheduler
  • HyPC - Hybrid Parallel Computation, a combination of HTC and HTPC, the granularity of the calculation is set such that a number of inputs are created which appropriately stride through the RN seed, however each input is operated upon in parallel using MPI

There are also a number of calculation modes that must be tested,

  • Nat. Geom - Nat(ive) Geometry, an MCNP input deck
  • DAG Geom - Direct Accelerated Geometry, tracking of particles upon faceted CAD models, can have more stringent RAM requirements over Native Geometry; also needs to transport more files

There are also a number of scoring modes that must be tested

  • Simple Tallies - A small number of uncomplicated tallies, several tallies of same types with no energy binning
  • Complex Tallies - A large number of complex tallies, energy dependence, fm card multipliers, dose response functions, traditional mesh tally (additional file handling)
  • Advanced Tallies - Newer style, using tetmesh and KDE tallies (additional file handling)
Par Mode Nat. Geom Simp. Tallies DAG Geom Simp Tallies Nat. Geom Comp Tallies DAG Geom Comp. Tallies DAG Geom Advanced Tallies
Serial beta beta beta beta #11 beta #17 #12
HTC beta beta #11 #3 beta #3 beta beta
HPC x x x x x
HTPC #4 #4 #4 #4 #4
HyPC x x x x x
@gonuke
Copy link
Member

gonuke commented Apr 10, 2013

It's probably important to continue to distinguish between HTC and HTPC. I'll edit the description to make that distinction.

gonuke pushed a commit that referenced this issue Apr 10, 2013
The computing modes and use cases enumerated in #10 are now on the README page for quick reference.
@gonuke
Copy link
Member

gonuke commented Apr 10, 2013

Also copied this to the README.md, but links to issues seems broken/tricky.

@makeclean
Copy link
Contributor Author

I'm updating this table rather than the README since we get nice issue hyperlinks, can we create "sticky issues" that would better serve this purpose or just leave this issue open for the duration?

@gonuke
Copy link
Member

gonuke commented Apr 19, 2013

I think we can go through a number of different issues, have them linked to each other in some sense, and put the current issue that is blocking progress in the table above? Too much work?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

2 participants