Made tests checks more exact and added some comments

tudasc · Aug 14, 2024 · f0e5f1a · f0e5f1a
1 parent fd48bd5
commit f0e5f1a
Show file tree

Hide file tree

Showing 38 changed files with 45 additions and 46 deletions.
diff --git a/testsuite/cuda-only/_cudaHostAllocBlock_user_malloc__nok.c b/testsuite/cuda-only/_cudaHostAllocBlock_user_malloc__nok.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// HECK-DAG: data race
+// HECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_cudaHostAlloc_user_malloc__nok.c b/testsuite/cuda-only/_cudaHostAlloc_user_malloc__nok.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// HECK-DAG: data race
+// HECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_cudaMallocBlock_user_malloc__nok.c b/testsuite/cuda-only/_cudaMallocBlock_user_malloc__nok.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// CHECK-DAG: data race
+// CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_cudaMalloc_user_malloc__nok.c b/testsuite/cuda-only/_cudaMalloc_user_malloc__nok.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// CHECK-DAG: data race
+// CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: [Error] sync
 #include <cstdio>
 #include <cuda_runtime.h>

diff --git a/testsuite/cuda-only/_cudafreeBlock_user_malloc_w_.c b/testsuite/cuda-only/_cudafreeBlock_user_malloc_w_.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// CHECK-NOT: data race
+// CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: [Error] sync
 
 #include <cstdio>
@@ -48,7 +48,7 @@ int main() {
   write_kernel_delay<<<blocksPerGrid, threadsPerBlock, 0, stream1>>>(data, size, 1316134912);
   // do a free which implicitly syncs/blocks
   cudaFree(data2);
-
+  //so this memcpy is safe
   cudaMemcpyAsync(h_data, data, size * sizeof(int), cudaMemcpyDefault, stream2);
   cudaStreamSynchronize(stream2);
   for (int i = 0; i < size; i++) {

diff --git a/testsuite/cuda-only/_cudafreeHostBlock_user_managed_w_.c b/testsuite/cuda-only/_cudafreeHostBlock_user_managed_w_.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// HECK-NOT: data race
+// HECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_cudafreeHost_user_managed_w_.c b/testsuite/cuda-only/_cudafreeHost_user_managed_w_.c
@@ -4,7 +4,7 @@
 
 // clang-format on
 
-// HECK-NOT: data race
+// HECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_cudafree_user_malloc_w_.c b/testsuite/cuda-only/_cudafree_user_malloc_w_.c
@@ -3,8 +3,8 @@
 // RUN: %tsan-options %cutests_test_dir/%basename_t.exe 2>&1 | %filecheck %s
 
 // clang-format on
-
-// CHECK-NOT: data race
+ 
+// CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: [Error] sync
 
 #include <cstdio>

diff --git a/testsuite/cuda-only/_ds_defuser_managed_w_.c b/testsuite/cuda-only/_ds_defuser_managed_w_.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-only/_ds_defuser_managed_w__nok.c b/testsuite/cuda-only/_ds_defuser_managed_w__nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-only/_mc_4_def_managed_w__nok.c b/testsuite/cuda-only/_mc_4_def_managed_w__nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 // REQUIRES: mca-rules
 

diff --git a/testsuite/cuda-to-mpi/send_ds_def_malloc_r_r.c b/testsuite/cuda-to-mpi/send_ds_def_malloc_r_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ds_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_ds_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ds_def_malloc_w_r_nok.c b/testsuite/cuda-to-mpi/send_ds_def_malloc_w_r_nok.c
@@ -49,10 +49,9 @@ int main(int argc, char* argv[]) {
   cudaDeviceSynchronize();
 
   if (world_rank == 0) {
+    //this kernel reads d_data
     kernel<<<blocksPerGrid, threadsPerBlock>>>(d_data, size);
-#ifdef cusan_SYNC
-    cudaDeviceSynchronize();  // FIXME: uncomment for correct execution
-#endif
+    //mpi aswell so we get a race
     MPI_Send(d_data, size, MPI_INT, 1, 0, MPI_COMM_WORLD);
   } else if (world_rank == 1) {
     MPI_Recv(d_data, size, MPI_INT, 0, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);

diff --git a/testsuite/cuda-to-mpi/send_ds_def_managed_r_r.c b/testsuite/cuda-to-mpi/send_ds_def_managed_r_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 
@@ -52,8 +52,9 @@ int main(int argc, char* argv[]) {
   if (world_rank == 0) {
     int* d_sum;
     cudaMalloc(&d_sum, size * sizeof(int));
+    //kernel only reads d_data
     kernel<<<blocksPerGrid, threadsPerBlock>>>(d_sum, d_data, size);
-
+    //same with mpi as such theres no race
     MPI_Send(d_data, size, MPI_INT, 1, 0, MPI_COMM_WORLD);
     cudaFree(d_sum);
   } else if (world_rank == 1) {

diff --git a/testsuite/cuda-to-mpi/send_ds_def_managed_w_r.c b/testsuite/cuda-to-mpi/send_ds_def_managed_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_functor.cpp b/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_functor.cpp
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_functor_nok.cpp b/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_functor_nok.cpp
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 
@@ -53,10 +53,9 @@ int main(int argc, char* argv[]) {
 
   if (world_rank == 0) {
     const auto lamba_kernel = [=] __host__ __device__(const int tid) { d_data[tid] = (tid + 1); };
+    //the kernel indirectly writes the d_data via the lambda function
     kernel_functor<decltype(lamba_kernel)><<<blocksPerGrid, threadsPerBlock>>>(lamba_kernel);
-#ifdef cusan_SYNC
-    cudaDeviceSynchronize();  // FIXME: uncomment for correct execution
-#endif
+    //and so we get a race with mpi
     MPI_Send(d_data, size, MPI_INT, 1, 0, MPI_COMM_WORLD);
   } else if (world_rank == 1) {
     MPI_Recv(d_data, size, MPI_INT, 0, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);

diff --git a/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_nok.c b/testsuite/cuda-to-mpi/send_ds_def_managed_w_r_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_eq_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_eq_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_es_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_es_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_mc_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_mc_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_mca_4_user_malloc_w_r.c b/testsuite/cuda-to-mpi/send_mca_4_user_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_mca_4_user_malloc_w_r_nok.c b/testsuite/cuda-to-mpi/send_mca_4_user_malloc_w_r_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ms_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_ms_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ms_def_malloc_w_r_nok.c b/testsuite/cuda-to-mpi/send_ms_def_malloc_w_r_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ms_pinned_def_malloc_w_r.c b/testsuite/cuda-to-mpi/send_ms_pinned_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ms_pinned_defuser_malloc_w_r.c b/testsuite/cuda-to-mpi/send_ms_pinned_defuser_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_msa_pinned_defuser_malloc_w_r_nok.c b/testsuite/cuda-to-mpi/send_msa_pinned_defuser_malloc_w_r_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 // CHECK-NOT: [Alloc]
 
 // REQUIRES: msa-rules

diff --git a/testsuite/cuda-to-mpi/send_msa_pinned_user_malloc_w_r_nok.c b/testsuite/cuda-to-mpi/send_msa_pinned_user_malloc_w_r_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 // REQUIRES: msa-rules
 

diff --git a/testsuite/cuda-to-mpi/send_sq_user_malloc_w_r.c b/testsuite/cuda-to-mpi/send_sq_user_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ss_user_malloc_r_r.c b/testsuite/cuda-to-mpi/send_ss_user_malloc_r_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/cuda-to-mpi/send_ss_user_malloc_w_r.c b/testsuite/cuda-to-mpi/send_ss_user_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/mpi-to-cuda/irecv_ds_def_malloc_w_w.c b/testsuite/mpi-to-cuda/irecv_ds_def_malloc_w_w.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/mpi-to-cuda/irecv_ds_def_malloc_w_w_nok.c b/testsuite/mpi-to-cuda/irecv_ds_def_malloc_w_w_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/mpi-to-cuda/irecv_ds_def_managed_w_w.c b/testsuite/mpi-to-cuda/irecv_ds_def_managed_w_w.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/mpi-to-cuda/irecv_ds_def_managed_w_w_nok.c b/testsuite/mpi-to-cuda/irecv_ds_def_managed_w_w_nok.c
@@ -5,7 +5,7 @@
 
 // CHECK-DAG: ThreadSanitizer: data race
 // CHECK-DAG: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-DAG: [Error]
+// CHECK-DAG: [Error] sync
 
 #include "../support/gpu_mpi.h"
 

diff --git a/testsuite/mpi-to-cuda/isend_ds_def_malloc_w_r.c b/testsuite/mpi-to-cuda/isend_ds_def_malloc_w_r.c
@@ -5,7 +5,7 @@
 
 // CHECK-NOT: ThreadSanitizer: data race
 // CHECK-NOT: Thread T{{[0-9]+}} 'cuda_stream'
-// CHECK-NOT: [Error]
+// CHECK-NOT: [Error] sync
 
 #include "../support/gpu_mpi.h"