Merge pull request #131 from clEsperanto/ADD--map/unmap-functionnality

Add  map/unmap functionnality

benchmark/CMakeLists.txt

@@ -36,3 +36,8 @@ add_executable(benchmark_maximum_projections benchmark_maximum_projections.cpp)
 target_compile_features(benchmark_maximum_projections PUBLIC cxx_std_17)
 target_link_libraries(benchmark_maximum_projections benchmark_kernel)
 set_target_properties(benchmark_maximum_projections PROPERTIES FOLDER "Benchmark")
+
+add_executable(benchmark_push_pull benchmark_push_pull.cpp)
+target_compile_features(benchmark_push_pull PUBLIC cxx_std_17)
+target_link_libraries(benchmark_push_pull benchmark_kernel)
+set_target_properties(benchmark_push_pull PROPERTIES FOLDER "Benchmark")

benchmark/benchmark_push_pull.cpp

@@ -0,0 +1,62 @@
+#include <iostream>
+#include <string>
+#include <vector>
+
+#include "clesperanto.hpp"
+
+#include <benchmark_base.cpp>
+
+class PushPullBenchmark : public BenchmarkBase
+{
+protected:
+  cle::Clesperanto      cle;
+  std::vector<float>    inputData;
+  std::array<size_t, 3> dim;
+  auto
+  Setup() -> void override
+  {
+    inputData = std::vector<float>(dataWidth * dataWidth * dataWidth);
+    dim = std::array<size_t, 3>{ { dataWidth, dataWidth, dataWidth } };
+
+    cle.GetDevice()->WaitForKernelToFinish();
+  }
+
+  auto
+  Iteration() -> void override
+  {
+    auto gpuInput = cle.Push<float>(inputData, dim);
+    auto output = cle.Pull<float>(gpuInput);
+  }
+
+  auto
+  Teardown() -> void override
+  {}
+
+public:
+  size_t dataWidth = 0;
+
+  PushPullBenchmark()
+    : cle(cle::Clesperanto())
+  {}
+
+  ~PushPullBenchmark() = default;
+};
+
+auto
+main(int argc, char ** argv) -> int
+{
+  PushPullBenchmark d;
+  d.dataWidth = 1 << 10;
+
+  d.iterationWarmupCount = 5;
+
+  if (argc >= 2)
+  {
+    d.dataWidth = std::stoi(argv[1]);
+    std::cout << "using " << d.dataWidth * d.dataWidth * d.dataWidth * sizeof(float) << " bytes memory" << std::endl;
+  }
+
+  d.Run();
+
+  return 0;
+}