add Pad random input test

unit_tests/Test_PadImpl.cpp

@@ -10,17 +10,15 @@
  ********************************************************************************/
 
 #include <array>
+#include <numeric> // std::accumulate
+#include <random>  // std::random_device, std::mt19937, std::uniform_real_distribution
 
-#include <iostream>
 #include <catch2/catch_test_macros.hpp>
 
-#include "Test_cuda.hpp"
-
-#include "aidge/data/Tensor.hpp"
-
 #include "aidge/backend/cpu.hpp"
 #include "aidge/backend/cuda.hpp"
-
+#include "aidge/data/Tensor.hpp"
+#include "aidge/utils/TensorUtils.hpp"
 
 using namespace Aidge;
 
@@ -367,11 +365,6 @@ TEST_CASE("[gpu/operator] Pad(forward)", "[Pad][GPU]") {
 
         float* computedOutput   = new float[myOutput->size()]();
         cudaMemcpy(computedOutput, op->getOutput(0)->getImpl()->rawPtr(), sizeof(float) * myOutput->size(), cudaMemcpyDeviceToHost);
-        for(int i = 0; i < myOutput->size(); i++){
-            
-            const float targetOutput = *(static_cast<float*>(myOutput->getImpl()->rawPtr()) + i);
-            std::cout << "target: " << targetOutput << "  computed: " << computedOutput[i]<< std::endl;
-        }
         for(int i = 0; i < myOutput->size(); i++){
             const float targetOutput = *(static_cast<float*>(myOutput->getImpl()->rawPtr()) + i);
             REQUIRE(fabs(computedOutput[i] - targetOutput) < 1e-6);
@@ -621,4 +614,88 @@ TEST_CASE("[gpu/operator] Pad(forward)", "[Pad][GPU]") {
 
         delete[] computedOutput;
     }
+    SECTION("Random Input") {
+        constexpr std::uint16_t NBTRIALS = 10;
+        // Create a random number generator
+        std::random_device rd;
+        std::mt19937 gen(rd());
+        std::uniform_real_distribution<float> valueDist(
+            0.1f, 1.1f); // Random float distribution between 0 and 1
+        std::uniform_int_distribution<std::size_t> padTypeDist(std::size_t(0), std::size_t(1));
+        std::uniform_int_distribution<std::size_t> dimSizeDist(std::size_t(1), std::size_t(10));
+        std::uniform_int_distribution<std::size_t> padSizeDist(std::size_t(0), std::size_t(5));
+
+        // To measure execution time of 'forward()'
+        std::chrono::time_point<std::chrono::system_clock> start;
+        std::chrono::time_point<std::chrono::system_clock> end;
+        std::chrono::duration<double, std::micro> duration{};
+        for (std::uint16_t trial = 0; trial < NBTRIALS; ++trial)
+        {
+            const std::size_t nbDims = 4;
+            std::vector<std::size_t> dims, paddings;
+            for (std::size_t i = 0; i < nbDims; ++i) {
+                dims.push_back(dimSizeDist(gen));
+            }
+
+            const std::size_t nb_elements = std::accumulate(dims.cbegin(), dims.cend(), std::size_t(1), std::multiplies<std::size_t>());
+
+            const std::size_t borderType = padTypeDist(gen);
+            const std::size_t padding = padSizeDist(gen);
+            // Create Pad Operator CUDA
+            std::shared_ptr<Node> myPadCUDA = Pad<2>({padding, padding, padding, padding}, "mypadcuda",  static_cast<PadBorderType>(borderType));
+            auto op_cuda = std::static_pointer_cast<OperatorTensor>(myPadCUDA -> getOperator());
+            op_cuda->setDataType(DataType::Float32);
+            op_cuda->setBackend("cuda");
+
+            // Create Pad Operator CPU
+            std::shared_ptr<Node> myPadCPU = Pad<2>({padding, padding, padding, padding}, "mypadcpu", static_cast<PadBorderType>(borderType));
+            auto op_cpu = std::static_pointer_cast<OperatorTensor>(myPadCPU -> getOperator());
+            op_cpu->setDataType(DataType::Float32);
+            op_cpu->setBackend("cpu");
+
+            float* array0 = new float[nb_elements];
+            for (std::size_t i = 0; i < nb_elements; ++i) {
+                array0[i] = valueDist(gen);
+            }
+
+            // input CUDA
+            float* array0_d;
+            std::shared_ptr<Tensor> T0_cuda = std::make_shared<Tensor>();
+            T0_cuda->setDataType(DataType::Float32);
+            T0_cuda->setBackend("cuda");
+            T0_cuda->resize(dims);
+            op_cuda->associateInput(0, T0_cuda);
+            cudaMalloc(reinterpret_cast<void **>(&array0_d), sizeof(float) * nb_elements);
+            cudaMemcpy(array0_d, array0, sizeof(float) * nb_elements, cudaMemcpyHostToDevice);
+            T0_cuda->getImpl()->setRawPtr(array0_d, nb_elements);
+
+            // input CPU
+            std::shared_ptr<Tensor> T0_cpu = std::make_shared<Tensor>();
+            op_cpu->associateInput(0,T0_cpu);
+            T0_cpu->setDataType(DataType::Float32);
+            T0_cpu->setBackend("cpu");
+            T0_cpu->resize(dims);
+            T0_cpu -> getImpl() -> setRawPtr(array0, nb_elements);
+
+            // forward CUDA
+            start = std::chrono::system_clock::now();
+            op_cuda->forward();
+            end = std::chrono::system_clock::now();
+            duration += std::chrono::duration_cast<std::chrono::microseconds>(end - start);
+
+            const std::size_t outSize =  op_cuda->getOutput(0)->size();
+            float *computed_cuda = new float[outSize]();
+            cudaMemcpy(computed_cuda, op_cuda->getOutput(0)->getImpl()->rawPtr(), sizeof(float) * outSize, cudaMemcpyDeviceToHost);
+
+            // forward CPU
+            op_cpu->forward();
+            float *computed_cpu = static_cast<float*>(op_cpu->getOutput(0)->getImpl()->rawPtr());
+            REQUIRE(approxEq<float>(*computed_cuda, *computed_cpu));
+
+            delete[] array0;
+            delete[] computed_cuda;
+            cudaFree(array0_d);
+        }
+        std::cout << "total time: " << duration.count() << "μs" << std::endl;
+    }
 }
\ No newline at end of file