Testing

Tests are an important part on preventing regressions and covering new feature functionality. There are multiple ways to run CIR tests.

The more aggresive (slower) one:

$ ninja check-all

CIR specific test targets using ninja:

$ ninja check-clang-cir
$ ninja check-clang-cir-codegen
  

Using lit from build directory:

$ cd build-release
$ ./bin/llvm-lit -a ../../clang/test/CIR
  

External

Currently we do not evaluate ClangIR against external test suites automatically, this, however, can be achieved with some manual work.

LLVM Test Suite

So far, we have tested ClangIR only against the SingleSource tests from this suite.

A table detailing ClangIR's status for each tests can be found here.

Currently, 51% (935/1824) of the SingleSource tests are passing. A good way to start contributing to ClangIR is to pick one of the NOEXE of FAIL tests and try to patch it!

How to Run

Build a release version of ClangIR with assertions enabled (prevents some false positives).
Apply the following patch to create a CMake flag that will enable ClangIR:
001-enable-clangir-on-singlesouce.patch

Create a build directory and build the tests with the release version of ClangIR:

cd <path-to>/test-suite
rm -rf ./build && mkdir build && cd build
cmake --no-warn-unused-cli ../ \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_C_COMPILER=<path-to-clangir-build>/bin/clang \
  -DCMAKE_CXX_COMPILER=<path-to-clangir-build>/bin/clang++ \
  -C=../cmake/caches/O3.cmake \
  -DTEST_SUITE_SUBDIRS=SingleSource \
  -DTEST_SUITE_CLANGIR_ENABLE=ON
  

Build tests (-k ensures it won't stop if a test fails to compile):
```
make -j -k
```
In the build directory, run the tests with:
```
lit --timeout=60 -v .
```

Generating the Comparison Table

Patch the utils/compare.py tool with:
002-generate-clangir-comparison-table.patch

Build the binaries and collect results for 15 runs of both baseline and ClangIR tests:

  # Set TEST_SUITE_CLANGIR_ENABLE=OFF to get the baseline tests.
  for i in {1..15}; do lit --timeout=60 -v -o "baseline_$i.json" .; done;
  for i in {1..15}; do lit --timeout=60 -v -o "clang_$i.json" .; done;
  

Create the comparison table using the patched utils/compare.py:

utils/compare.py \
--lhs-name baseline --rhs-name clangir --minimal-names \
--merge-average --all -m compile_time -m exec_time \
--csv-output results.csv \
baseline_1.json ... baseline_15.json vs clangir_1.json ... clangir_15.json
  