AMReX-Astro / AMReX-Astro/Microphysics

loop over cells in order by cost to reduce divergence

Open
#1,683 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

enhancement GPU reactions
Dominant language
C++
Stars
43
Forks
46
Avg merge
2d 18h
Merged PRs (30d)
15

Description

It may be possible to reduce load imbalance across a warp by sorting cells based on their cost (rhs evaluations or substeps), and then looping through them based on this sorting. Then, cells with similar costs to burn are computed together across the same warp, reducing SIMD divergence.

This is mostly intended for chemistry networks, where the cost is mostly a function of density (which sets how close to equilibrium the network is).

Contributor guide

No contributing guide indexed for this repository

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

No files, tests, or entry points are identified. Start by locating the chemistry-network cell loop and the rhs-evaluation or substep cost paths, then determine how cells could be ordered by cost without changing results. Done would require demonstrating reduced SIMD divergence and load imbalance on representative chemistry workloads.

Written by the indexing model from the issue text.

Assessment

Tech stack
cpp
Domain
hpc, performance
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.