compiler: move launch check injection to later in compilation pipeline

[EdCaunt]

Other half in PRO

[codecov]

Codecov Report

❌ Patch coverage is 18.18182% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.41%. Comparing base (ad2febe) to head (85603ea).

Files with missing lines	Patch %	Lines
devito/passes/iet/errors.py	18.18%	27 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2903      +/-   ##
==========================================
- Coverage   83.46%   83.41%   -0.05%     
==========================================
  Files         248      248              
  Lines       51570    51602      +32     
  Branches     4441     4447       +6     
==========================================
+ Hits        43045    43046       +1     
- Misses       7774     7804      +30     
- Partials      751      752       +1     

Flag	Coverage Δ
pytest-gpu-aomp-amdgpuX	68.74% <18.18%> (-0.05%)	⬇️
pytest-gpu-gcc-	78.04% <18.18%> (-0.04%)	⬇️
pytest-gpu-icx-	77.64% <18.18%> (-0.03%)	⬇️
pytest-gpu-nvc-nvidiaX	69.27% <18.18%> (-0.06%)	⬇️

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.

[mloubout]

Not sure this needs to be split in three functions. And if so the def/macro should be in the actual iet pass and this should just check the option

[EdCaunt]

It doesn't need to be three functions, but it is more readable that way imo. The def/macro is the same across all IETs so I don't see why it would need to go inside the IET pass itself?

[mloubout]

I don't think it's more readable to split a pass this small into three functions that don't have any reuse

[Copilot]

Pull request overview

This PR introduces an IET pass to inject a kernel-launch error check macro later in the compilation pipeline (gated by errctl), aiming to stop execution early when a device kernel launch fails.

Changes:

• Added a new check_launch pass that wraps KernelLaunch nodes found under Iteration nodes with an additional “check launch” statement.
• Added generation of a CHECK_LAUNCH macro definition (based on langbb error-peeking support) and returns it via the pass headers byproduct.
• Exported check_launch from devito.passes.iet.errors.

---

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