[rocm-libraries] ROCm/rocm-libraries#5249 (commit 2a114bb)

[CK] [CK_TILE] Improve build and test time of CI with smart
 dependency parser (#5249)

## Motivation

Existing dependency parser needs full build of tests to determine which
tests are affected by code changes in a PR. This still takes 2-4 hours
for building the tests which slows down the CI as the number of tests
grow. To resolve this issue we implemented a smart dependency parser
which uses CMake Configure to parse dependencies and build only the
affected test cases. We have ensured that two approaches are available
1) CMake pre-build analysis for each PR to ensure fast build and test.
2) Ninja post-build analysis to enable full build for nightly tests.

## Technical Details

```bash
### 1. Configure the project with CMake
cmake -G Ninja -DCMAKE_EXPORT_COMPILE_COMMANDS=ON ..

### 2. Analyze dependencies (no build required!)
python3 ../script/dependency-parser/main.py cmake-parse compile_commands.json build.ninja \
  --workspace-root .. --output cmake_dependency_mapping.json --parallel 8

### 3. Find tests affected by changes
python3 ../script/dependency-parser/main.py select cmake_dependency_mapping.json origin/develop \
  HEAD --test-prefix --output tests_to_run.json

### 4. Build only affected tests
ninja $(jq -r '.executables[]' tests_to_run.json | tr '\n' ' ')

### 5. Run affected tests
ctest -R "$(jq -r '.regex' tests_to_run.json)"
```

### Jenkins Integration
- Added `buildMode` to jenkinsfile to integrate both `selective` and
`full` build methods

### Known Limitations

### 1. Build-Time Generated Headers (HIGH RISK)

**Problem:** Files generated during the build process (e.g., via
`add_custom_command`) cannot be analyzed before building.

**Example:**
```cmake
add_custom_command(
  OUTPUT ${CMAKE_BINARY_DIR}/generated/config.hpp
  COMMAND generate_config.sh
  DEPENDS template.hpp.in
)
```

**Impact:** If a source file includes `generated/config.hpp`, the
dependency won't be detected until after building.

**Mitigation:**
- CK analysis shows **no generated headers** currently used
- If generated headers are added in the future, they must be built first
- Recommendation: Generate headers in CMake configure phase (not build
phase) when possible

## Test Plan
**1. Modified Files:**
```
include/ck_tile/ops/common.hpp
include/ck_tile/ops/gemm.hpp
include/ck_tile/ops/gemm/warp/warp_gemm.hpp
```
**2. Compare tests selected between `build.ninja` and `cmake-parse`
methods**

## Test Result
- 1. The test completed in 5-6 minutes finding about 8000+ executables
that should be built.
- 2. We selected a commit 5ccc1387ea which resulted in same 7 tests with
both legacy and new methods.
-

PR | Legacy tests | Smart tests | Notes
-- | -- | -- | --
5261 | 453 | 455 | Only 2 tests (test_amdgcn_mma and
test_amdgcn_sparse_mma)
5168 | 0 | 0 | Changes in   dispatcher only. No CK tests invoked.
5249 | 0 | 0 | Changes to   dependency parser. No CK tests invoked
5260 | 0 | 0 | Changes in   dispatcher only. No CK tests invoked.
5174 | 1 | 1 | One test from FMHA   affected by this PR in both cases
5383 | 0 | 0 | Changes are only in benchmark files. Did not trigger any
tests
5445 | 1 | 1 | Changes are only to tests/ck_tile/gemm_streamk. Only
triggered one streamk test in both cases.
5454 | 3 | 3 | Both methods identified same test_grouped_conv_bwd tests
5427 | 234 | 234 | Core infrastructure header changes. Detected exactly
same tests
5388 | 85 | 85 | modifies warp-level GEMM operations (warp_gemm.hpp,
warp_gemm_dispatcher.hpp). Correctly identified all the streamK gemm
tests

## Submission Checklist

- [x ] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
This commit is contained in:
Yaswanth Raparti
2026-03-19 05:31:35 +00:00
committed by assistant-librarian[bot]
parent 345a56c55e
commit 652d3456ca
13 changed files with 3585 additions and 210 deletions

109
Jenkinsfile vendored
View File

@@ -1,3 +1,29 @@
// Composable Kernel Jenkins Pipeline
//
// SMART BUILD SYSTEM:
// This pipeline uses intelligent dependency analysis to speed up PR builds while
// maintaining full validation on nightly runs.
//
// How it works:
// 1. PR Builds (Selective):
// - Configure: cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON (~30s)
// - Analyze: Parse compile_commands.json + clang -MM for dependencies (~2min)
// - Select: git diff to find affected tests (~1s)
// - Build: ninja <affected-tests> only (minutes vs hours)
// - Test: ctest -R <affected-pattern>
//
// 2. Nightly Builds (Full):
// - FORCE_CI=true from cron triggers full build
// - All targets built and tested for validation
//
// 3. Safety Checks:
// - Forces full build if CMake configuration changes
// - Forces full build if dependency cache stale (>7 days)
// - Manual override: set DISABLE_SMART_BUILD=true
//
// Benefits: PR builds 5h → 30min (typical), nightly builds unchanged
// See: script/dependency-parser/README.md for details
//
def rocmnode(name) {
return '(rocmtest || miopen) && (' + name + ')'
}
@@ -678,13 +704,21 @@ def cmake_build(Map conf=[:]){
}
setup_cmd = conf.get(
"setup_cmd",
"""${cmake_envs} cmake -G Ninja ${setup_args} -DCMAKE_CXX_FLAGS=" -O3 " .. """
)
build_cmd = conf.get(
"build_cmd",
"${build_envs} ninja -j${nt} ${config_targets}"
"""${cmake_envs} cmake -G Ninja ${setup_args} -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DCMAKE_CXX_FLAGS=" -O3 " .. """
)
// Smart-build: Only build if running all tests or forced
// Otherwise, smart-build will determine what to build after cmake configure
if (runAllUnitTests) {
build_cmd = conf.get(
"build_cmd",
"${build_envs} ninja -j${nt} ${config_targets}"
)
} else {
// Smart-build enabled: skip full build, only run cmake configure
build_cmd = ""
}
cmd = conf.get("cmd", """
${setup_cmd}
${build_cmd}
@@ -741,25 +775,44 @@ def cmake_build(Map conf=[:]){
//run tests except when NO_CK_BUILD is set
if(!setup_args.contains("NO_CK_BUILD")){
sh "python3 ../script/ninja_json_converter.py .ninja_log --legacy-format --output ck_build_trace_${arch_name}.json"
archiveArtifacts "ck_build_trace_${arch_name}.json"
sh "python3 ../script/parse_ninja_trace.py ck_build_trace_${arch_name}.json"
if (params.NINJA_BUILD_TRACE || params.BUILD_INSTANCES_ONLY){
if (params.NINJA_FTIME_TRACE) {
echo "running ClangBuildAnalyzer"
sh "/ClangBuildAnalyzer/build/ClangBuildAnalyzer --all . clang_build.log"
sh "/ClangBuildAnalyzer/build/ClangBuildAnalyzer --analyze clang_build.log > clang_build_analysis_${arch_name}.log"
archiveArtifacts "clang_build_analysis_${arch_name}.log"
}
// do not run unit tests when building instances only
if(!params.BUILD_INSTANCES_ONLY){
if (!runAllUnitTests){
sh "../script/launch_tests.sh"
// Smart Build: Run smart_build_and_test.sh
sh """
export WORKSPACE_ROOT=${env.WORKSPACE}
export PARALLEL=32
export NINJA_JOBS=${nt}
export ARCH_NAME=${arch_name}
export PROCESS_NINJA_TRACE=true
export NINJA_FTIME_TRACE=${params.NINJA_FTIME_TRACE ? 'true' : 'false'}
bash ../script/dependency-parser/smart_build_and_test.sh
"""
// Archive artifacts if they were generated
if (fileExists("ck_build_trace_${arch_name}.json")) {
archiveArtifacts "ck_build_trace_${arch_name}.json"
}
if (fileExists("clang_build_analysis_${arch_name}.log")) {
archiveArtifacts "clang_build_analysis_${arch_name}.log"
}
}
else{
sh "ninja check"
echo "Full test suite requested (RUN_ALL_UNIT_TESTS=true or develop branch)"
sh "ninja -j${nt} check"
// Process ninja build trace after full build
sh "python3 ../script/ninja_json_converter.py .ninja_log --legacy-format --output ck_build_trace_${arch_name}.json"
archiveArtifacts "ck_build_trace_${arch_name}.json"
sh "python3 ../script/parse_ninja_trace.py ck_build_trace_${arch_name}.json"
if (params.NINJA_FTIME_TRACE) {
echo "running ClangBuildAnalyzer"
sh "/ClangBuildAnalyzer/build/ClangBuildAnalyzer --all . clang_build.log"
sh "/ClangBuildAnalyzer/build/ClangBuildAnalyzer --analyze clang_build.log > clang_build_analysis_${arch_name}.log"
archiveArtifacts "clang_build_analysis_${arch_name}.log"
}
}
if (params.RUN_BUILDER_TESTS && !setup_args.contains("-DCK_CXX_STANDARD=") && !setup_args.contains("gfx10") && !setup_args.contains("gfx11")) {
sh 'ninja check-builder'
@@ -781,12 +834,24 @@ def cmake_build(Map conf=[:]){
}
else{
// run unit tests unless building library for all targets
// Note: This else block is when NINJA_BUILD_TRACE=false and BUILD_INSTANCES_ONLY=false
// So no ninja trace processing needed here
if (!params.BUILD_INSTANCES_ONLY){
if (!runAllUnitTests){
sh "../script/launch_tests.sh"
// Smart Build: Run smart_build_and_test.sh
sh """
export WORKSPACE_ROOT=${env.WORKSPACE}
export PARALLEL=32
export NINJA_JOBS=${nt}
export ARCH_NAME=${arch_name}
export PROCESS_NINJA_TRACE=false
export NINJA_FTIME_TRACE=false
bash ../script/dependency-parser/smart_build_and_test.sh
"""
}
else{
sh "ninja check"
echo "Full test suite requested (RUN_ALL_UNIT_TESTS=true or develop branch)"
sh "ninja -j${nt} check"
}
if (params.RUN_BUILDER_TESTS && !setup_args.contains("-DCK_CXX_STANDARD=") && !setup_args.contains("gfx10") && !setup_args.contains("gfx11")) {
sh 'ninja check-builder'
@@ -1241,6 +1306,10 @@ pipeline {
name: "USE_SCCACHE",
defaultValue: true,
description: "Use the sccache for building CK (default: ON)")
booleanParam(
name: "DISABLE_SMART_BUILD",
defaultValue: false,
description: "Disable smart build system and force full build/test (default: OFF). Smart build uses pre-build dependency analysis for selective testing on PRs, full builds on nightly runs.")
booleanParam(
name: "RUN_CPPCHECK",
defaultValue: false,