Reformat a2asweep output to match gfxsweep style

[nileshnegi]

Motivation

Unify a2asweep output with the better-formatted gfxsweep output.
Also adds a summary table at the end.

Technical Details

• Replace printf with Utils::Print throughout the sweep table
• Use sep-character column separators (space vs comma for CSV mode)
• Move blocksize out of row prefix into the header line, print header per blocksize block like gfxsweep does
• Print summary block at end
• Default BLOCKSIZES changed from {256} to {256,512,768,1024}
• Single unified table: header printed once, blockSize is a row column
• USE_HIP_EVENTS controls timing mode (preset default: 1 = GPU-event-timed)
  • USE_HIP_EVENTS=1: GPU-event-timed per-executor minBw; SHOW_MIN_ONLY=0 adds a maxBw column per SE
  • USE_HIP_EVENTS=0: CPU wall-clock avgTotalBandwidthGbPerSec
• Banner and best-result summary reflect the active timing mode
• Results map keyed by (blockSize, numSes, unroll) for verbose output

Test Plan

Test Result

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 updates the a2asweep preset output formatting to align more closely with the gfxsweep style, including a unified table layout and an end-of-run “best result” summary, while also expanding the default blocksize sweep set.

Changes:

• Default USE_HIP_EVENTS behavior for a2asweep and reflect timing mode in the banner/output.
• Reformat sweep output into a single table with CSV-aware separators and a “best bandwidth” summary block.
• Expand default BLOCKSIZES from {256} to {256,512,768,1024}.

---

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