| name | test-kerncap |
| description | Test local kerncap changes end-to-end by profiling an application, extracting a kernel, and validating the reproducer. Use when the user asks to test kerncap against any HIP or Triton workload, or wants to validate extraction on a real GPU application. |
Test kerncap Against an Application
Test local kerncap changes end-to-end by extracting and validating a kernel from any application.
Parameters
| Parameter | Required | Description |
|---|
app_cmd | Yes | Full command to run the application (binary + arguments), e.g. $WORK/dev/llama.cpp/build/bin/llama-bench -m model.gguf -p 512 -n 32 |
conda_env | No | Conda environment to activate before running commands (e.g. llama_cpp). If not provided, use the current environment. |
kernel_name | No | Name of the kernel to extract (e.g. mul_mat_q). If not provided, profile the application first and select the top kernel by execution time. |
Paths
| Item | Path |
|---|
| kerncap source | kerncap/ (relative to IntelliKit repo root) |
| Output directory | /tmp/kerncap-test/<kernel_name> |
Environment Setup
If conda_env is provided, activate it before any other step:
conda activate <conda_env>
If already in a different environment, switch explicitly. Do not assume the current shell environment is correct.
If conda_env is not provided, proceed with the current environment as-is.
Workflow
Step 1: Reinstall kerncap
Ensure the correct environment is active (if applicable), then uninstall and reinstall to pick up local changes:
pip uninstall kerncap -y && pip install kerncap/
Step 2: Profile to identify target kernel
If kernel_name was provided: Skip this step and proceed to Step 3.
If kernel_name was not provided: Run profiling to discover the top bottleneck kernel:
kerncap profile -- <app_cmd>
Select the kernel with the highest total execution time from the profile output. Use its name as kernel_name for all subsequent steps. Tell the user which kernel was selected and why.
Important: Use a sufficiently long substring from the profile output as kernel_name so that kerncap extract matches the intended kernel, not a different instantiation. For example, templated kernels like mul_mat_q have many instantiations differing only by template parameters; passing just mul_mat_q will capture the first dispatch that matches, which may not be the top-ranked one. Prefer including template parameters in the substring (e.g. mul_mat_q<(ggml_type)39 instead of mul_mat_q).
Step 3: Extract the kernel
kerncap extract --help
Use the help output to construct the appropriate kerncap extract command for the application. Key flags to determine:
--cmd — the application command (app_cmd)--source-dir — where the kernel source lives (ask the user if unclear)--output — /tmp/kerncap-test/<kernel_name>--language — hip or triton depending on the workload- Any additional flags (
-D defines, --dispatch, etc.)
If extraction fails or produces errors: Stop here and report the full error output. This indicates the local kerncap changes have a bug that needs fixing.
If extraction succeeds: Inspect the output directory for expected files (metadata.json, argument dumps, source files). If the output looks reasonable, proceed to compile and run.
Step 4: Compile and run the reproducer
Navigate to the output directory and build/run the reproducer:
cd /tmp/kerncap-test/<kernel_name>
make run
If make run fails: Stop here and report the full compiler or runtime error output. This is the primary signal that kerncap generated an incorrect reproducer.
If make run succeeds: Proceed to validation.
For Triton reproducers, there may be no make run target. Instead run:
cd /tmp/kerncap-test/<kernel_name>
python3 reproducer.py
This replays through Triton and, after edits, writes candidate.hsaco for byte-exact validation.
Step 5: Validate the reproducer
5a. Smoke test — confirm baseline replay works:
kerncap validate /tmp/kerncap-test/<kernel_name>
If no rebuilt HSACO is present, this is a smoke test only (VA-faithful captures). It confirms the replay runs without crashing but does not check numerical correctness. If candidate.hsaco or optimized.hsaco exists from a prior rebuild, kerncap auto-detects it and performs captured-vs-rebuilt byte-exact validation.
5b. Rebuild — build a baseline HSACO from the unmodified kernel source:
For HIP reproducers:
cd /tmp/kerncap-test/<kernel_name>
make recompile
This confirms the VFS-overlay recompile pipeline works. It produces optimized.hsaco from the unmodified kernel_variant.cpp.
For Triton reproducers:
cd /tmp/kerncap-test/<kernel_name>
python3 reproducer.py
This confirms the editable Python reproducer can re-JIT the kernel and write candidate.hsaco.
If the rebuild fails: Stop here and report the error. For HIP, this indicates an issue with the source finder or VFS overlay generation. For Triton, this indicates an issue with source copying, signature reconstruction, tensor layout reconstruction, or Triton re-JIT.
5c. Correctness validation — compare rebuilt HSACO against captured baseline:
kerncap validate /tmp/kerncap-test/<kernel_name>
This auto-detects optimized.hsaco (HIP) or candidate.hsaco (Triton), runs replay twice (captured HSACO vs rebuilt HSACO), and compares output memory regions byte-for-byte. Since the kernel source is unmodified, they should match exactly. A failure here indicates a recompilation or re-JIT fidelity issue.
Use kerncap validate -v /tmp/kerncap-test/<kernel_name> when you need the per-region PASS lines; by default kerncap prints a concise Result: PASS/FAIL (...) footer and suppresses noisy per-region success output unless validation fails.
Step 6: Report results
Summarize:
- Whether reinstall succeeded
- Whether profiling identified a kernel (if applicable, and which one)
- Whether extraction completed (and any warnings)
- Whether
make run compiled and executed successfully
- Whether smoke test passed (Step 5a)
- Whether rebuild succeeded (Step 5b)
- Whether correctness validation passed (Step 5c)
- Any errors or warnings encountered at each step
