Implement performance checking script: i8 vs f16 non-fused conv kernels

[dorde-antic]

Implement shell script that captures the performance difference between data types to validate expected kernel performance.
Resolves ROCm/rocMLIR-internal#1674

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #1727      +/-   ##
===========================================
- Coverage    78.52%   77.99%   -0.52%     
===========================================
  Files          100      100              
  Lines        29907    30057     +150     
  Branches      4452     4656     +204     
===========================================
- Hits         23482    23442      -40     
- Misses        4590     4601      +11     
- Partials      1835     2014     +179     

Flag	Coverage Δ
mfma	77.98% <ø> (-0.54%)	⬇️
navi3x	77.98% <ø> (?)
navi4x	77.99% <ø> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

see 34 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[djramic]

LGTM

[Copilot]

Pull Request Overview

This PR introduces a shell script to benchmark and compare FP16 vs INT8 performance for non-fused convolution kernels.

• Adds perf-test-i8-f16-conv.sh to compile, time, and record average runtimes for both data types.
• Supports configurable model name, ONNX path, and iteration count via flags.
• Outputs results into a directory named after the model with a times summary file.

Comments suppressed due to low confidence (1)

mlir/utils/performance/perf-test-i8-f16-conv.sh:5

• The usage comment references /performance-checking instead of this script's actual name (perf-test-i8-f16-conv.sh); updating it will avoid confusion.

# Usage: /performance-checking --d <model> --p <model_path> [--r <number_of_iterations>]"

[umangyadav]

Why this script is in rocMLIR ? Looks like better place is inside migraphx-benchmark-utils
https://github.com/ROCm/migraphx-benchmark-utils

[dorde-antic]

On MLIR Standup (in january) we talked to put it in /mlir/utils/performance. But I agree with what you have said. Should I open PR there and put this script then? @umangyadav

[dorde-antic]

And would it be in this directory: https://github.com/ROCm/migraphx-benchmark-utils/tree/main/scripts and to add it to readme table? @umangyadav

[umangyadav]

You can close this one

[dorde-antic]

Merged on migraphx-benchmark-utils repo - https://github.com/ROCm/migraphx-benchmark-utils/pull/70