Merge pull request #380 from guilhermeAlmeida1/patch-1

Ensure synchronisation with async copies

Author: stephenswat

device/common/src/make_prefix_sum_buffer.cpp

@@ -34,7 +34,7 @@ prefix_sum_buffer_t make_prefix_sum_buffer(
         vecmem::data::vector_buffer<prefix_sum_size_t> sizes_sum_buff(
             sizes_sum.size(), mr.main);
         copy.setup(sizes_sum_buff);
-        (copy)(vecmem::get_data(sizes_sum), sizes_sum_buff);
+        (copy)(vecmem::get_data(sizes_sum), sizes_sum_buff)->wait();
         vecmem::data::vector_view<prefix_sum_size_t> sizes_sum_view(
             sizes_sum_buff);
 

examples/run/cuda/seeding_example_cuda.cpp

@@ -26,7 +26,6 @@
 #include <vecmem/memory/cuda/device_memory_resource.hpp>
 #include <vecmem/memory/host_memory_resource.hpp>
 #include <vecmem/utils/cuda/async_copy.hpp>
-#include <vecmem/utils/cuda/copy.hpp>
 
 // ACTS include(s).
 #include <Acts/Definitions/Units.hpp>
@@ -60,11 +59,10 @@ int seq_run(const traccc::seeding_input_config& i_cfg,
 
     traccc::cuda::stream stream;
 
-    vecmem::cuda::copy copy;
-    vecmem::cuda::async_copy async_copy{stream.cudaStream()};
+    vecmem::cuda::async_copy copy{stream.cudaStream()};
 
-    traccc::cuda::seeding_algorithm sa_cuda{mr, async_copy, stream};
-    traccc::cuda::track_params_estimation tp_cuda{mr, async_copy, stream};
+    traccc::cuda::seeding_algorithm sa_cuda{mr, copy, stream};
+    traccc::cuda::track_params_estimation tp_cuda{mr, copy, stream};
 
     // performance writer
     traccc::seeding_performance_writer sd_performance_writer(
@@ -169,8 +167,8 @@ int seq_run(const traccc::seeding_input_config& i_cfg,
         // Copy the seeds to the host for comparisons
         traccc::seed_collection_types::host seeds_cuda;
         traccc::bound_track_parameters_collection_types::host params_cuda;
-        copy(seeds_cuda_buffer, seeds_cuda);
-        copy(params_cuda_buffer, params_cuda);
+        copy(seeds_cuda_buffer, seeds_cuda)->wait();
+        copy(params_cuda_buffer, params_cuda)->wait();
 
         if (run_cpu) {
             // Show which event we are currently presenting the results for.

examples/run/cuda/seq_example_cuda.cpp

@@ -30,7 +30,6 @@
 #include <vecmem/memory/cuda/host_memory_resource.hpp>
 #include <vecmem/memory/host_memory_resource.hpp>
 #include <vecmem/utils/cuda/async_copy.hpp>
-#include <vecmem/utils/cuda/copy.hpp>
 
 // System include(s).
 #include <exception>
@@ -72,13 +71,12 @@ int seq_run(const traccc::full_tracking_input_config& i_cfg,
 
     traccc::cuda::stream stream;
 
-    vecmem::cuda::copy copy;
-    vecmem::cuda::async_copy async_copy{stream.cudaStream()};
+    vecmem::cuda::async_copy copy{stream.cudaStream()};
 
     traccc::cuda::clusterization_algorithm ca_cuda(
-        mr, async_copy, stream, common_opts.target_cells_per_partition);
-    traccc::cuda::seeding_algorithm sa_cuda(mr, async_copy, stream);
-    traccc::cuda::track_params_estimation tp_cuda(mr, async_copy, stream);
+        mr, copy, stream, common_opts.target_cells_per_partition);
+    traccc::cuda::seeding_algorithm sa_cuda(mr, copy, stream);
+    traccc::cuda::track_params_estimation tp_cuda(mr, copy, stream);
 
     // performance writer
     traccc::seeding_performance_writer sd_performance_writer(
@@ -220,9 +218,9 @@ int seq_run(const traccc::full_tracking_input_config& i_cfg,
         traccc::seed_collection_types::host seeds_cuda;
         traccc::bound_track_parameters_collection_types::host params_cuda;
         if (run_cpu || i_cfg.check_performance) {
-            copy(spacepoints_cuda_buffer, spacepoints_per_event_cuda);
-            copy(seeds_cuda_buffer, seeds_cuda);
-            copy(params_cuda_buffer, params_cuda);
+            copy(spacepoints_cuda_buffer, spacepoints_per_event_cuda)->wait();
+            copy(seeds_cuda_buffer, seeds_cuda)->wait();
+            copy(params_cuda_buffer, params_cuda)->wait();
         }
 
         if (run_cpu) {

examples/run/sycl/seeding_example_sycl.sycl

@@ -168,8 +168,8 @@ int seq_run(const traccc::seeding_input_config& i_cfg,
         // Copy the seeds to the host for comparison.
         traccc::seed_collection_types::host seeds_sycl;
         traccc::bound_track_parameters_collection_types::host params_sycl;
-        copy(seeds_sycl_buffer, seeds_sycl);
-        copy(params_sycl_buffer, params_sycl);
+        copy(seeds_sycl_buffer, seeds_sycl)->wait();
+        copy(params_sycl_buffer, params_sycl)->wait();
 
         if (run_cpu && i_cfg.check_performance) {
             // Show which event we are currently presenting the results for.

examples/run/sycl/seq_example_sycl.sycl

@@ -242,9 +242,9 @@ int seq_run(const traccc::full_tracking_input_config& i_cfg,
         traccc::seed_collection_types::host seeds_sycl;
         traccc::bound_track_parameters_collection_types::host params_sycl;
         if (run_cpu || i_cfg.check_performance) {
-            copy(spacepoints_sycl_buffer, spacepoints_per_event_sycl);
-            copy(seeds_sycl_buffer, seeds_sycl);
-            copy(params_sycl_buffer, params_sycl);
+            copy(spacepoints_sycl_buffer, spacepoints_per_event_sycl)->wait();
+            copy(seeds_sycl_buffer, seeds_sycl)->wait();
+            copy(params_sycl_buffer, params_sycl)->wait();
         }
 
         if (run_cpu) {