AcademySoftwareFoundation / AcademySoftwareFoundation/OpenCue

RQD - reserve the right number of cores.

Closed
#1,295 0 comments 0 reactions 0 assignees View on GitHub
enhancement
Dominant language
Python
Stars
957
Forks
259
Avg merge
3d 6h
Merged PRs (30d)
16

Description

**The problems**
- RQD reserves too much cores and is caped by the number of P-cores.
- a frame with 5 cores on an hyperthreaded system will reserve at least 6 cores up to 10 cores.
- a system listed with 24 cores (8 P-cores(x2) + 8 E-cores(x1)) will not accept frames with more than 16 minCores.
- the core assignation will randomly pick cores and count them as 1, even if they have multiple threads.

**Describe the enhancement**
- During reservation, list cores in order and reserve only the required amount, counting their threads.
- ex : minCores=5 -> reserve 2 P-cores(x2) and 1 E-core
- ex : minCores=7 -> reserve 3 P-cores(x2) and 1 E-core
- ex : minCores=8 -> reserve 4 P-cores(x2)

**Screenshots**
minCores = 5 -> uses 7 cores
![image](https://github.com/AcademySoftwareFoundation/OpenCue/assets/5556461/7f792544-a2e0-423f-bd04-c0e2e9d23ed7)

same job on another rqd process,
minCores = 5 -> uses 8 cores
![image](https://github.com/AcademySoftwareFoundation/OpenCue/assets/5556461/e25fd417-4e90-4977-b1ab-ac2fd360810e)

**Version Number**
0.22

Contributor guide

Open the contributing guide

Research direction

Start by tracing RQD's core reservation path for minCores and inspect how physical cores and threads are counted and selected. Reproduce the examples for minCores=5, 7, and 8 on a hybrid hyperthreaded system. Done means reservation is ordered, counts threads correctly, and does not cap usable capacity at the number of P-cores.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
backend
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.