"Turn" cuda and two-dimensional dynamic array

1 /*2 * Copyright Xu Hongzhi (northwest. School of Information Engineering). All rights reserved.3 * Data:2012-4-224 */5 //6 //This program is a demonstration of two-dimensional dynamic array of space applications and memory data with each other to copy both methods7#include <stdio.h>8 //#include <cutil_inline.h>9#include <iostream>Ten#include <cuda_runtime.h> One //#include <cutil.h> A using namespacestd; -  - intMainintargcChar**argv) the { -      -     //Cut_device_init (argc, argv); //Start CUDA - #if1 +     //Method 1. Progressive copy -     float**cpu_orign, **cpu_ret;//host-side raw data, copy-back data +     float**gpu;//device-side data A     intwidth =5, height =3;//the width and height of the array atsize_t size =sizeof(float) *width;//width of data in bytes -     intRow, col; -  -     //Request memory space, and initialize -Cpu_orign =New float*[height]; -Cpu_ret =New float*[height]; in      for(row =0; Row < height; ++row) -     { toCpu_orign[row] =New float[width]; +Cpu_ret[row] =New float[width]; -         //Initializing Data the          for(col =0; Col < width; ++Col) *         { $Cpu_orign[row][col] = (float) (Row +col);Panax NotoginsengCpu_ret[row][col] =0.0f; -         } the     } +  A     //Request memory space and initialize theGPU =New float*[height]; +      for(row =0; Row < height; ++row) -     { $(Cudamalloc (void* *) &Gpu[row], size)); $(Cudamemset (Gpu[row),0, size)); -     } -  the     //Copy host-side raw data to device side -      for(row =0; Row < height; ++row)Wuyi (cudamemcpy (Gpu[row], Cpu_orign[row], size, cudamemcpyhosttodevice)); the  -     //copy device-side data to the host side to return data Wu      for(row =0; Row < height; ++row) - (cudamemcpy (Cpu_ret[row], Gpu[row], size, cudamemcpydevicetohost)); About  $     //print host side return data -      for(row =0; Row < height; ++row) -     { -          for(col =0; Col < width; ++Col) Acout << Cpu_ret[row][col] <<" "; +cout <<Endl; the     } -     //free up memory and graphics space $      Free(cpu_orign); the      Free(Cpu_ret); the      for(row =0; Row < height; ++row) the (Cudafree (Gpu[row])); the #endif -  in  the #if0 the     //Method 2: Overall copy About     float**cpu_orign, **cpu_ret;//host-side raw data, copy-back data the     float**gpu;//device-side data the     intwidth =5, height =3;//the width and height of the array thesize_t size =sizeof(float) *width;//width of data in bytes + size_t pitch; -     intRow, col; the Bayi     //Request memory space, and initialize theCpu_orign =New float*[height]; theCpu_ret =New float*[height]; -      for(row =0; Row < height; ++row) -     { theCpu_orign[row] =New float[width]; theCpu_ret[row] =New float[width]; the         //Initializing Data the          for(col =0; Col < width; ++Col) -         { theCpu_orign[row][col] = (float) (Row +col); theCpu_ret[row][col] =0.0f; the         }94     } the  the     //Request memory space and initialize the(Cudamallocpitch (void* *) &AMP;GPU, &pitch, size, height));98(Cudamemset2d (GPU, Pitch,0, size, height)); About  -     //Copy host-side raw data to device side101 (Cudamemcpy2d (GPU, pitch, cpu_orign, size, size, height, cudamemcpyhosttodevice));102 103     //copy device-side data to the host side to return data104 (Cudamemcpy2d (cpu_ret, size, GPU, pitch, size, height, cudamemcpydevicetohost)); the 106     //print host side return data107      for(row =0; Row < height; ++row)108     {109          for(col =0; Col < width; ++Col) thecout << Cpu_ret[row][col] <<" ";111cout <<Endl; the     }113     //free up memory and graphics space the      Free(cpu_orign); the      Free(Cpu_ret); the (Cudafree (GPU));117 #endif118System"Pause");119     //cut_exit (argc, argv); //Exit Cuda -}

http://blog.csdn.net/xhz1234/article/details/7487767

"Turn" cuda and two-dimensional dynamic array