Microbenchmark - Fix precision loss in kernel launch overhead benchmark

[alephpiece]

Description

The original code uses integer division in the time calculation: (end_tv.tv_usec) / 1000. Since an empty kernel launch overhead is typically very small (often less than 1ms), this integer division truncates the microsecond component to 0, leading to significant precision loss and incorrect results.

Major Revision

• Changed the divisor from 1000 to 1000. (floating-point literal). This forces floating-point arithmetic, preserving the microsecond precision required for accurate overhead measurement.

[guoshzhao]

Thanks @alephpiece for pointing out this issue. We will verify and review this PR.

[Copilot]

Pull request overview

This PR fixes a precision loss bug in the kernel launch overhead microbenchmark where integer division was causing microsecond components to be truncated to zero for sub-millisecond measurements.

Changes:

• Refactored time calculation from expanding then subtracting to subtracting then expanding, using floating-point arithmetic throughout

---

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

[guoshzhao]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 3 pipeline(s).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.69%. Comparing base (036c471) to head (0557744).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #773   +/-   ##
=======================================
  Coverage   85.69%   85.69%           
=======================================
  Files         103      103           
  Lines        7890     7890           
=======================================
  Hits         6761     6761           
  Misses       1129     1129           

Flag	Coverage Δ
cpu-python3.10-unit-test	70.42% <ø> (ø)
cpu-python3.7-unit-test	69.85% <ø> (ø)
cuda-unit-test	83.60% <ø> (ø)

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

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[guoshzhao]

@alephpiece could you please run python3 setup.py format and python3 setup.py lint to fix the lint check?

[alephpiece]

@guoshzhao
I may be missing something, but I’m not sure why python3 setup.py format and python3 setup.py lint would help for this PR. The change here is only a one-line fix in a .cu file, and there are no Python file changes. Also, CPP code lint is already passing.