Skip to content

feat: compute-sanitizer in build #1735

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
vgvassilev merged 2 commits into from
Feb 23, 2026
+41 −26
Merged

feat: compute-sanitizer in build #1735

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
f531327
add compute-sanitizer
fogsong233 Feb 20, 2026
9ed2502
fix a cuda mem error.
fogsong233 Feb 21, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion .github/workflows/ci.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -106,7 +106,7 @@ jobs:
clang-runtime: '19'
coverage: true
cuda: true
extra_cmake_options: '-DCLAD_ENABLE_ENZYME_BACKEND=On'
extra_cmake_options: '-DCLAD_ENABLE_ENZYME_BACKEND=On -DCLAD_CUDA_TEST_USE_SANITIZER=On'
extra_packages: ' libzstd-dev '
#clang-format: true

Expand All @@ -127,6 +127,7 @@ jobs:
os: [self-hosted, cuda, heavy]
runs-on: cuda
compiler: clang-16
extra_cmake_options: '-DCLAD_CUDA_TEST_USE_SANITIZER=On'
clang-runtime: '18'
cuda: true

Expand Down
4 changes: 4 additions & 0 deletions test/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,11 +84,15 @@ if(NOT LLVM_MAIN_SRC_DIR)
set(LLVM_MAIN_SRC_DIR ${LLVM_BUILD_MAIN_SRC_DIR})
endif()

option(CLAD_CUDA_TEST_USE_SANITIZER "Run Clang cuda tests with compute-sanitizer" OFF)
option(CLAD_TEST_USE_VG "Run Clang tests under Valgrind" OFF)
set(CLAD_TEST_EXTRA_ARGS --verbose --show-skipped --show-unsupported)
if(CLAD_TEST_USE_VG)
set(CLAD_TEST_EXTRA_ARGS ${CLAD_TEST_EXTRA_ARGS} "--vg --vg-arg=-q")
endif ()
if(CLAD_CUDA_TEST_USE_SANITIZER)
set(CLAD_TEST_EXTRA_ARGS ${CLAD_TEST_EXTRA_ARGS} "--param" "cuda_sanitizer=1")
endif ()

add_lit_testsuite(check-clad "Running the Clad regression tests"
${CMAKE_CURRENT_BINARY_DIR}
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ForwardMode.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oForwardMode.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ForwardMode.out | %filecheck_exec %s
// RUN: %cudarun ./ForwardMode.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/GradientCuda.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -8,7 +8,7 @@
// RUN: %cladclang_cuda -I%S/../../include --cuda-gpu-arch=%cudaarch \
// RUN: --cuda-path=%cudapath %cudaldflags -oGradientCuda.out %s
//
// RUN: ./GradientCuda.out | %filecheck_exec %s
// RUN: %cudarun ./GradientCuda.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
20 changes: 10 additions & 10 deletions test/CUDA/GradientKernels.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
// RUN: %cladclang_cuda -Xclang -plugin-arg-clad -Xclang -disable-tbr -I%S/../../include --cuda-path=%cudapath \
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oGradientKernels.out %s
//
// RUN: ./GradientKernels.out | %filecheck_exec %s
// RUN: %cudarun ./GradientKernels.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime

Expand Down Expand Up @@ -1019,11 +1019,11 @@ __global__ void injective_reassignment_loop(int *a) {

int main(void) {
int *a, *d_a;
cudaMalloc(&a, sizeof(int));
cudaMalloc(&d_a, sizeof(int));
cudaMalloc(&a, 2 * sizeof(int));
cudaMalloc(&d_a, 2 * sizeof(int));

TEST(kernel, dim3(1), dim3(1), 0, false, a, d_a, 1); // CHECK-EXEC: 10
TEST(kernel, dim3(1), dim3(1), 0, true, a, d_a, 1); // CHECK-EXEC: 10
TEST(kernel, dim3(1), dim3(1), 0, false, a, d_a, 2); // CHECK-EXEC: 10
TEST(kernel, dim3(1), dim3(1), 0, true, a, d_a, 2); // CHECK-EXEC: 10

auto error = clad::gradient(fake_kernel);
error.execute_kernel(dim3(1), dim3(1), a, d_a); // CHECK-EXEC: Use execute() for non-global CUDA kernels
Expand Down Expand Up @@ -1146,12 +1146,12 @@ int main(void) {
TEST_2(indices_lin_comb, dim3(1), dim3(5, 1, 1), 0, false, "out, in", dummy_out, dummy_in, d_out, d_in, 5); // CHECK-EXEC: 20, 25, 45, 15, 15

int *n, *d_n;
cudaMalloc(&n, sizeof(int));
cudaMalloc(&d_n, sizeof(int));
cudaMalloc(&n, 2 * sizeof(int));
cudaMalloc(&d_n, 2 * sizeof(int));

TEST(kernel_device_injective, dim3(1), dim3(1), 0, false, n, d_n, 1); // CHECK-EXEC: 4
TEST(injective_reassignment, dim3(1), dim3(1), 0, false, n, d_n, 1); // CHECK-EXEC: 1
TEST(injective_reassignment_loop, dim3(1), dim3(1), 0, false, n, d_n, 1); // CHECK-EXEC: 1
TEST(kernel_device_injective, dim3(1), dim3(1), 0, false, n, d_n, 2); // CHECK-EXEC: 4
TEST(injective_reassignment, dim3(1), dim3(1), 0, false, n, d_n, 2); // CHECK-EXEC: 1
TEST(injective_reassignment_loop, dim3(1), dim3(1), 0, false, n, d_n,2); // CHECK-EXEC: 1

cudaFree(dummy_in);
cudaFree(dummy_out);
Expand Down
8 changes: 4 additions & 4 deletions test/CUDA/RunCudaDemos.C
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@
// CHECK_VECTOR_ADDITION: clad::custom_derivatives::thrust::reduce_pullback
// CHECK_VECTOR_ADDITION: clad::custom_derivatives::thrust::transform_pullback

// RUN: ./VectorAddition.out | FileCheck -check-prefix CHECK_VECTOR_ADDITION_EXEC %s
// RUN: %cudarun ./VectorAddition.out | FileCheck -check-prefix CHECK_VECTOR_ADDITION_EXEC %s
// CHECK_VECTOR_ADDITION_EXEC: Running vector addition demo.
// CHECK_VECTOR_ADDITION_EXEC: Gradients of sum wrt initial x: 1 1 1 1 1 1 1 1 1 1

Expand All @@ -17,7 +17,7 @@
// CHECK_PARTICLE_SIMULATION: clad::custom_derivatives::thrust::copy_pullback
// CHECK_PARTICLE_SIMULATION: clad::custom_derivatives::thrust::transform_pullback

// RUN: ./ParticleSimulation.out | FileCheck -check-prefix CHECK_PARTICLE_SIMULATION_EXEC %s
// RUN: %cudarun ./ParticleSimulation.out | FileCheck -check-prefix CHECK_PARTICLE_SIMULATION_EXEC %s
// CHECK_PARTICLE_SIMULATION_EXEC: Running particle simulation demo.
// CHECK_PARTICLE_SIMULATION_EXEC: Gradients of final x-pos sum wrt initial vx: 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5

Expand All @@ -44,7 +44,7 @@
// CHECK_LINEAR_REGRESSION-NEXT: }
// CHECK_LINEAR_REGRESSION-NEXT: }

// RUN: ./LinearRegression.out | FileCheck -check-prefix CHECK_LINEAR_REGRESSION_EXEC %s
// RUN: %cudarun ./LinearRegression.out | FileCheck -check-prefix CHECK_LINEAR_REGRESSION_EXEC %s
// CHECK_LINEAR_REGRESSION_EXEC: Running linear regression demo.
// CHECK_LINEAR_REGRESSION_EXEC: Gradients of loss wrt weights (w): -9 -18 -27 -36 -45 -54 -63 -72 -81 -90

Expand All @@ -53,7 +53,7 @@
// CHECK_BOW_LOGREG: void logistic_loss_batch2_prepared_l2_grad(
// CHECK_BOW_LOGREG: clad::custom_derivatives::thrust::inner_product_pullback

// RUN: ./BoWLogisticRegression.out | FileCheck -check-prefix CHECK_BOW_LOGREG_EXEC %s
// RUN: %cudarun ./BoWLogisticRegression.out | FileCheck -check-prefix CHECK_BOW_LOGREG_EXEC %s
// CHECK_BOW_LOGREG_EXEC: Running minimal logistic regression demo.
// CHECK_BOW_LOGREG_EXEC: Loss:
// CHECK_BOW_LOGREG_EXEC: Gradient wrt w:
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustAdjacentDifference.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustAdjacentDifference.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustAdjacentDifference.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustAdjacentDifference.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustCopy.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustCopy.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustCopy.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustCopy.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustReduce.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustReduce.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustReduce.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustReduce.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustReduceByKey.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustReduceByKey.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustReduceByKey.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustReduceByKey.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustScan.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustScan.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustScan.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustScan.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustScanByKey.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustScanByKey.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustScanByKey.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustScanByKey.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustSortByKey.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustSortByKey.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustSortByKey.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustSortByKey.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustTransform.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustTransform.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustTransform.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustTransform.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
2 changes: 1 addition & 1 deletion test/CUDA/ThrustTransformReduce.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@
// RUN: --cuda-gpu-arch=%cudaarch %cudaldflags -oThrustTransformReduce.out \
// RUN: -Xclang -verify %s 2>&1 | %filecheck %s
//
// RUN: ./ThrustTransformReduce.out | %filecheck_exec %s
// RUN: %cudarun ./ThrustTransformReduce.out | %filecheck_exec %s
//
// REQUIRES: cuda-runtime
//
Expand Down
10 changes: 10 additions & 0 deletions test/lit.cfg
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -323,6 +323,15 @@ if clang_repl_path:
#if loadable_module:
# config.available_features.add('loadable_module')

cuda_run_prefix = ''
if lit_config.params.get('cuda_sanitizer'):
cuda_tool_path = config.environment['PATH']
san_path = lit.util.which('compute-sanitizer', os.path.join(config.cuda_path, 'bin'))
if san_path is None:
lit_config.fatal("cuda_sanitizer is enabled but compute-sanitizer was not found")
cuda_run_prefix = san_path + ' --tool memcheck --error-exitcode=1 --print-limit=0'
config.substitutions.append(('%cudarun', cuda_run_prefix))

libcudart_path = lit.util.which('libcudart.so', config.cuda_libdir)
if libcudart_path is not None:
config.available_features.add('cuda-runtime')
Expand All @@ -337,6 +346,7 @@ if libcudart_path is not None:
config.environment['CUDA_VISIBLE_DEVICES'] = os.environ['CUDA_VISIBLE_DEVICES']
config.substitutions.append(('%cudaarch', config.cuda_test_arch))


if(config.have_enzyme):
config.available_features.add('Enzyme')

Expand Down
Loading