removing secondary reordering

[AtlantaPepsi]

Motivation

Quick fix for missed grouping change

Technical Details

Grouping should be based directly on StrideGenerate output, previous vestige of a second layer of reordering is in poda2a and caused wrong grouping results.

Test Plan

Example:
With 3 nodes, each with 4 devices, NUM_GROUPS=4, GROUP_STRIDE=4 should result in
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11] => [0, 4, 8, 1, 5, 9, 2, 6, 10, 3, 7, 11] => [0, 4, 8], [1, 5, 9], [2, 6, 10], [3, 7, 11]

Test Result

A2A group 0: R0:G0, R1:G0, R2:G0
A2A group 1: R0:G1, R1:G1, R2:G1
A2A group 2: R0:G2, R1:G2, R2:G2
A2A group 3: R0:G3, R1:G3, R2:G3

Submission Checklist

• Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

[Copilot]

Pull request overview

This PR fixes incorrect device grouping in the PodAllToAll preset by removing an unintended second reordering step and basing group construction directly on the Utils::StrideGenerate() permutation.

Changes:

• Builds the devices list in the same order as the StrideGenerate output (instead of reindexing into a separately permuted layout).
• Ensures group membership corresponds to contiguous chunks of the strided device order (fixing wrong grouping results in poda2a).

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.