X-Git-Url: http://git.euphorik.ch/index.cgi?a=blobdiff_plain;f=WCudaMSE%2FStudent_Cuda%2Fsrc%2Fcpp%2Fcore%2F01_Hello%2F02_hello_add.cu;fp=WCudaMSE%2FStudent_Cuda%2Fsrc%2Fcpp%2Fcore%2F01_Hello%2F02_hello_add.cu;h=70f3908b8a24a025936e2fc7343c50af37d3a654;hb=8d08c12b29c2a14684f35c023ee39e694bb80d25;hp=0000000000000000000000000000000000000000;hpb=226de81f7e1f1fbf4ac79d0d089e8a05ec7159a0;p=GPU.git

diff --git a/WCudaMSE/Student_Cuda/src/cpp/core/01_Hello/02_hello_add.cu b/WCudaMSE/Student_Cuda/src/cpp/core/01_Hello/02_hello_add.cu
new file mode 100755
index 0000000..70f3908
--- /dev/null
+++ b/WCudaMSE/Student_Cuda/src/cpp/core/01_Hello/02_hello_add.cu
@@ -0,0 +1,104 @@
+// Attention : Extension .cu
+
+#include <iostream>
+#include <stdio.h>
+#include "cudaTools.h"
+#include "Device.h"
+
+using std::cout;
+using std::endl;
+
+/*----------------------------------------------------------------------*\
+ |*			Declaration 					*|
+ \*---------------------------------------------------------------------*/
+
+/*--------------------------------------*\
+ |*		Imported	 	*|
+ \*-------------------------------------*/
+
+/*--------------------------------------*\
+ |*		Public			*|
+ \*-------------------------------------*/
+
+__host__  bool isAddScalarGPU_Ok(void);
+
+/*--------------------------------------*\
+ |*		Private			*|
+ \*-------------------------------------*/
+
+__host__  static int addScalarGPU(int a, int b);
+__global__ static void addScalar(int a, int b, int* ptrC);
+
+/*----------------------------------------------------------------------*\
+ |*			Implementation 					*|
+ \*---------------------------------------------------------------------*/
+
+/*--------------------------------------*\
+ |*		Public			*|
+ \*-------------------------------------*/
+
+__host__  bool isAddScalarGPU_Ok(void)
+    {
+    cout << endl << "[Hello Cuda 2]" << endl;
+
+    int a = 2;
+    int b = 7;
+
+    int sumGPU = addScalarGPU(a, b);
+    int sumTheorique = a + b;
+
+    cout <<"\n[CPU] "<< a << " + " << b << " = " << sumGPU << endl;
+
+    return sumGPU == sumTheorique;
+    }
+
+/*--------------------------------------*\
+ |*		Private			*|
+ \*-------------------------------------*/
+
+__host__ int addScalarGPU(int a, int b)
+    {
+    int c;
+    int* ptrC=&c;	// on host (CPU)
+    int* ptrDev_c; 	// on device (GPU)
+
+    // Specifier nb thread : ici 1 thread au total !
+    dim3 dg = dim3(1,1,1);
+    dim3 db = dim3(1, 1, 1);
+
+    // Debug
+    //Device::print(dg, db);
+     Device::checkDimError(dg,db);
+
+    size_t size=sizeof(int);
+    HANDLE_ERROR(cudaMalloc((void**) &ptrDev_c, size)); // Device memory allocation (*)
+
+    addScalar<<<dg,db>>>(a,b,ptrDev_c); // asynchrone !!
+    Device::checkKernelError("addScalar"); // facultatif
+
+    //v1
+    Device::synchronize();// Pour printf sur  GPU
+
+    //v2
+   // cudaDeviceSynchronize(); // Pour printf sur  GPU
+
+    // memoryManagement => barrier de synchronisation
+    HANDLE_ERROR(cudaMemcpy(ptrC, ptrDev_c, size, cudaMemcpyDeviceToHost));// Device -> Host
+    HANDLE_ERROR(cudaFree(ptrDev_c)); // device dispose memory in (*)
+
+    return c;
+    }
+
+
+__global__ void addScalar(int a, int b, int* ptrC)
+    {
+    *ptrC = a + b;
+
+    // debug
+    printf("[GPU] %d + %d = %d",a,b, *ptrC);
+    }
+
+/*----------------------------------------------------------------------*\
+ |*			End	 					*|
+ \*---------------------------------------------------------------------*/
+