[rocprofiler-compute] Add pre-processor guards for rocflop

[vedithal-amd]

Motivation

Allow rocflop sample workload to be built on MI 200

Technical Details

Add pre-processor guards and runtime checks for rocflop.cpp for gfx90a

JIRA ID

Test Plan

Test Result

Ensure rocflop.cpp compiles on MI 200

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 adds preprocessor guards and runtime checks to exclude SMFMAC (Sparse MFMA) instructions for the gfx90a architecture (MI 200), where these instructions are not actually supported. The changes correctly restrict SMFMAC availability to gfx940 (MI300) and later architectures.

Key changes:

• Added !defined(__gfx90a__) to all SMFMAC-related preprocessor guards
• Updated the runtime architecture check from arch.minor > 0x4 to arch.minor >= 0x4 to correctly enable SMFMAC on gfx940 while excluding gfx90a
• Updated all related comments to accurately reflect that SMFMAC is only available on gfx940 and later, not on gfx906, gfx908, or gfx90a

---

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